NATO ASI Series Advanced Science Institutes Series A series presenting the results of activities sponsored by the NATO Science Committee, which aims at the dissemination of advanced scientific and technological knowledge, with a view to strengthening links between scientific communities.

The Series is published by an international board of publishers in conjunction with the NATO Scientific Affairs Division

A Life Sciences B Physics

C Mathematical and Physical Sciences

D Behavioural and Social Sciences

E Applied Sciences

F Computer and Systems Sciences

G Ecological Sciences H Cell Biology I Global Environmental

Change

Plenum Publishing Corporation London and New York Kluwer Academic Publishers Dordrecht, Boston and London

Springer-Verlag Berlin Heidelberg New York London Paris Tokyo Hong Kong Barcelona Budapest

NATO-PCO DATABASE

The electronic index to the NATO ASI Series provides full bibliographical references (with keywords and/or abstracts) to more than 30000 contributions from international scientists published in all sections of the NATO ASI Series. Access to the NATO-PCO DATABASE compiled by the NATO Publication Coordination Office is possible in two ways:

- via online FILE 128 (NATO-PCO DATABASE) hosted by ESRIN, Via Galileo Galilei, I-00044 Frascati, Italy.

- via CD-ROM "NATO Science & Technology Disk" with user-friendly retrieval software in English, French and German (© WTV GmbH and DATAWARE Technologies Inc. 1992).

The CD-ROM can be ordered through any member of the Board of Publishers or through NATO-PCO, Overijse, Belgium.

Series F: Computer and Systems Sciences, Vol. 133

The ASI Series F Books Published as a Result of Activities of the Special Programme on ADVANCED EDUCATIONAL TECHNOLOGY

This book contains the proceedings of a NATO Advanced Research Work­shop held within the activities of the NATO Special Programme on Advanced Educational Technology, running from 1988 to 1993 underthe auspices of the NATO Science Committee.

The volumes published so far in the Special Programme are as follows (further details are given at the end of this volume):

67: Designing Hypermedia for Learning. 1990. 76: Multimedia Interface Design in Education. 1992. 78: Integrating Advanced Technology into Technology Education. 1991. 80: Intelligent Tutoring Systems for Foreign Language Learning. 1992. 81: Cognitive Tools for Learning. 1992. 84: Computer-Based Learning Environments and Problem Solving. 1992. 85: Adaptive Learning Environments: Foundations and Frontiers. 1992. 86: Intelligent Learning Environments and Knowledge Acquisition in Physics. 1992. 87: Cognitive Modelling and Interactive Environments in Language Learning. 1992. 89: Mathematical Problem Solving and New Information Technologies. 1992. 90: Collaborative Learning Through Computer Conferencing. 1992. 91: New Directions for Intelligent Tutoring Systems. 1992. 92: Hypermedia Courseware: Structures of Communication and Intelligent Help. 1992. 93: Interactive Multimedia Learning Environments. 1992. 95: Comprehensive System Design: A New Educational Technology. 1993. 96: New Directions in Educational Technology. 1992. 97: Advanced Models of Cognition for Medical Training and Practice. 1992.

104: Instructional Models in Computer-Based Learning Environments. 1992. 105: Designing Environments for Constructive Learning. 1993. 107: Advanced Educational Technology for Mathematics and Science. 1993. 109: Advanced Educational Technology in Technology Education. 1993. 111: Cognitive Models and Intelligent Environments for Learning Programming. 1993. 112: Item Banking: Interactive Testing and Self-Assessment. 1993. 113: Interactive Learning Technology for the Deaf. 1993. 115: Learning Electricity and Electronics with Advanced Educational Technology. 1993. 116: Control Technology in Elementary Education. 1993. 117: Intelligent Learning Environments: The Case of Geometry. 1994. 119: Automating Instructional Design, Development, and Delivery. 1993. 121: Learning from Computers: Mathematics Education and Technology. 1993. 122: Simulation-Based Experiential Learning. 1993. 125: Student Modelling: The Key to Individualized Knowledge-Based Instruction. 1994. 128: Computer Supported Collaborative Learning. 1994. 129: Human-Machine Communication for Educational Systems Design. 1994. 132: Design of Mathematical Modelling Courses for Engineering Education. 1994. 133: Collaborative Dialogue Technologies in Distance Learning. 1994.

Collaborative Dialogue Technologies in Distance Learning

Edited by

M. Felisa Verdejo Departamento de Ingenieria Electrica, Electrönica y de Control Escuela Tecnica Superior de Ingenieros Industriales Universidad Nacional de Educacion a Distancia Apdo 60149, E-28080 Madrid, Spain

Stefano A. Cerri Dipartimento di Scienze dell' Informazione Universitä di Milano Via Comelico 39,1-20135 Milano, Italy

Springer-Verlag

Berlin Heidelberg GmbH

Proceedings of the NATO Advanced Research Workshop on Collaborative Dialogue Technologies in Distance Learning, held in Segovia, Spain, April 24-27, 1993

CR Subject Classification (1991): K.3, H.5

ISBN 978-3-642-63394-2 ISBN 978-3-642-57899-1 (eBook) DOI 10.1007/978-3-642-57899-1

CIP data applied for

This work is subject to copyright. All rights are reserved, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcast­ing, reproduction on microfilms or in any other way, and storage in data banks. Duplication of this publication or parts thereof is permitted only under the provisions of the German Copyright Law of September 9, 1965, in its current version, and permission for use must always be obtained from Springer-Verlag. Violations are liable for prosecution under the German Copyright Law.

© Springer-Verlag Berlin Heidelberg 1994 Originally published by Springer-Verlag Berlin Heidelberg New York in 1994 Softcover reprint of the hardcover 1 st edition 1994

Typesetting: Camera-ready by authors/editors SPIN: 10130750 45/3140 - 5 4 3 2 1 0 - Printed on acid-free paper

Preface

In April 1993, an interdisciplinary NATO Advanced Research Workshop on "Collaborative dialogue technologies in distance learning" was held in Segovia, Spain. The workshop brought together researchers in fields related to distance learning using computer-mediated communication. The statement of justification of the NATO ARW follows hereafter.

Justification of the NATO Advanced Research Workshop on Collaborative Dialogue Technologies in Distance Learning

Computer Mediated Communication (CMC) systems have features that reduce some temporal, physical and social constraints on communication. Theories of communication have shifted from viewing communication as a linear transmission of messages by a sender to a receiver, to viewing it as a social paradigm, where individuals are actors in a network of interdependent relationships embedded in organizational and social structures. Recent research focuses on models of information-sharing to support not only the activities of individuals but also the problem-solving activities of groups, such as decision-making, planning or co­writing. This area of research is called Computer Supported Cooperative Work (CSCW). The Artificial Intelligence (AI) approach uses knowledge-based systems to enhance and facilitate all these processes, including the possibility of using natural language.

The traditional model of distance education places a strong emphasis on indepen­dent study, supported by well developed learning materials. This model can be characterized as one-way media. However, the potential of CMC to provide better guidance to the student in Higher Distance Education has been quickly recognized for at least two kind of activities: information sharing and interaction. This is a first step towards a change from stand-alone study to a more cooperative situation.

The shift from individual to collaborative learning is also an emerging approach in recent learning theories and experiments. Distance learning is one setting in particular that deserves more and more attention (as, for example, in the current CEC DELTA programme). The scenario may include: a) a group of students performing different kinds of cooperative learning activities such as discussion or problem solving, using conventional or multimedia-based material; b) a group of teachers, organizing, facilitating and controlling the cooperative learning; c) a group of authors creating courseware; and d) a computer environment supporting all these activities.

The proposed workshop is organized around themes related to three main activities involved in collaborative distance learning scenarios: learning, coordination and co-authoring. The title "collaborative dialogue technologies" stresses the focus on conversational models of group interaction. Leading lecturers have been carefully selected as experts in at least one of the topics, reaching a balance from theoretical

VI Preface

and practical background, covering the fields mentioned above, and coming from industry and academy. Other participants, to be chosen by means of a limited call for papers, will enrich the presentations and discussions.

Group activities commenting on each other's ideas will be an important component of the workshop. Three groups will be organized to facilitate active discussions. Their work will be presented in plenary meetings. Participants will be assigned to a specific group, in order to facilitate the sharing of different profiles. An in-depth synthesis will close the workshop. Presentations, a summary of the group's discus­sions and the synthesis will form the basis of the book.

Maria Felisa Verdejo, November 1992

Eighteen papers were presented at the workshop: thirteen were invited and five selected from responses to an electronic call for contributions. Presentations and group discussions were interleaved. Three working groups were organized to discuss further issues concerning:

• Pedagogical framework • Learning experiences • Models and systems

This volume comprises an edited and refereed version of those papers, plus three additional papers that synthesize the discussions of the three working groups. The introduction presents the papers and points out some of the links between them.

April 1994 Maria Felisa Verdejo, Stefano Alessandro Cerri

Acknowledgments VII

Acknowledgments

The workshop was sponsored by NATO Advanced Educational Technology Programme. Additional funding was provided by the Spanish Distance Learning University (U.N.E.D) through Media & Technologies and Continuum Education Vice-Rectories, the Engineering School, and the Associated Center of U.N.E.D in Segovia. Our thanks to these organizations.

We would like to acknowledge the contribution of our co-organizers, Morten Flate Paulsen and Cristina Simon, the speakers and participants in making the workshop a shared success.

Local arrangements, apart from the weather, were the effort of many people, from U.N.E.D and the Centro Asociado de Segovia. Let us specially mention Miguel Padilla, M. Soledad Salafranca, Miguel Angel Sebastian, as well as other Spanish colleagues: M.Teresa Abad, Beatriz Barros, Ana Garda Serrano, Encarna Pastor.

This book is an outcome of the presentations and group work that took place at the workshop. Pre-proceedings were circulated at the workshop and each paper has been later reviewed by at least two reviewers who were asked to collaborate directly with the author, preferably via e-mail, in order to help himlher to enhance the paper content and formulation. The paper review process has been in this respect quite uncommon, as the taSk was not to accept the papers but to criticize constructively papers already accepted. In most cases this brought significant improvements with respect to the originally circulated versions. To all of the authors, as well to the referees we express our sincere gratitude for their time and efforts.

In producing this volume, Bruna Zonta handled the formatting and lexical check and Gordon Davies made a linguistic proofreading. Their essential and profes­sional work is gratefully acknowledged.

We would like to mention the special contribution of Cristina Simon, from the very beginning of the workshop organization to the last stage of the book preparation. Her availability and assistance was a great personal and professional support.

Referees

Monique Baron, Nur Erol, Camillo Falcone, Peter Goodyear, Guy Gouarderes, Monique Grandbastien, Roger Hartley, Gordon McCalla, Angus McIntyre, Jean Francois Nicaud, Helen Pain, Morten Flate Paulsen, Domenico Rotondi, John Self, Cristina Simon, Kris Van Marcke, Martial Vivet, Radboud Winkels

Foreword

This book is an important contribution to the growing literature and research on the use of technology-mediated interaction in distance education and open learning. It may seem paradoxical to some people that technology can be used successfully to bring a strong element of inter-personal interaction, dialogue, and social networking into educational programmes, but this in fact is the case in what we recognise now as 'third generation' distance education. The first generation of distance education is represented by the traditional, relatively artisanal, correspondence course, where the learner is isolated, and the only communication is by letter, with one's correspondence tutor. The second generation is represented by institutions such as the British Open University or Spain's Universidad Nacional de Educaci6n a Distancia, which operate more on a broadcast and mass production model, with economies of scale due to large numbers of students following individual courses; tuition by written correspondence is still an important element of interaction, but this is supplemented by periodic face-to-face tutorials and residential sessions. Third generation distance education courses are characterised by the combination of well­prepared course materials designed for independent study (print, audio-visual, computer-based ... ), with the possibility of various forms of technology-mediated interaction amongst learners, tutors, and resource people dispersed in space and time The technology is thus used as a medium for conducting conversations and dialogues around the courseware, allowing for the co-construction of new knowledge and understandings amongst groups of learners and tutors, who might otherwise have little opportunity to interact with each other.

Over a third of the papers in this book are concerned in one way or another with the use of computer conferencing and electronic mail in education; this relatively low cost technology for deferred time, text-based communication, is proving to be a powerful and increasingly popular medium - either for totally 'online' courses based on the virtual classroom model, or as an adjunct to more conventional, second generation, distance education courses. Computer conferencing has in fact the potential to blur many of the traditional distinctions between face-to-face and distance education methods, and to help promote a resource-based approach to learning which no longer emphasises the teacher as the main source of knOWledge. Conferencing supports many to-many communication, and conferencing software includes features specifically designed to help in the organisation, structuring, and retrieval of messages. Messages can be linked to each other (eg as 'comments'), organised in different 'branches' or 'topics' of a conference, and search commands can rapidly identify messages with particular key words in their titles, or in the body of the text. Support is also provided for tracking the activity of individual group members: for example, it is generally possible to see which messages in a conference a given member has read, or which messages have been written by a given member. Special commands are available to the person responsible for a conference (the 'moderator' or 'organiser') which can help in defining the membership of the conference, in keeping the discussion on track, and in scheduling the opening and closing of discussion topics.

X Foreword

Computer conferencing is by no means a new or advanced technology (the fIrst systems were developed twenty years ago), but its use in education on a signifIcant scale, along with other computer networking posibilities, is a relatively recent phenomenom associated with the more widespread availability of personal computers, modems, and communications software. Over the next few years, the appearance on the consumer market of 'teleputers' (personal computers with built-in features for both real time and deferred time multimedia communication) and of access to ISDN facilities, will no doubt lead to an increase in the use of real time document sharing, audio graphic conferencing, and video-conferencing in distance education. Several of the papers in this book explore the potential for educational dialogue, and the associated research issues, which will arise from the new and very rich communicative environments which these technological developments will make available.

Perhaps the strongest potential in terms of the development of third generation distance education lies in the area which has come to be known as computer supported collaborative learning (CSCL) - an emerging fIeld of research and practice to which this book makes an important contribution. There are three classes of technology which, combined, can provide groupware environments suitable for CSCL:

- communication systems (synchronous text, audio, audio graphics, and video communication; asynchronous electronic mail, computer conferencing, voice­mail, and fax), - resource sharing systems (synchronous screen-sharing and electronic white­boards, concept mapping tools; asynchronous access to fIle systems and databases), - group process support systems (project management systems, shared calendars, co-authoring tools, voting tools, ideas generation and brainstorming tools).

Previously, these systems tended to be used independently, and the classic audio, audio graphic, and videoconferencing systems, of course, were in use long before the advent of multimedia personal computers. However, it is the bringing together of these three classes of technology into one computer-supported environment, on the desktop, or in educational resource centres, that will make a major qualitative difference in the educational potential of computer support for collaborative learning, multiple dialogue structures, and group work.

Why is the interest in collaborative learning, and its support and mediation through computers and telecommunications, so strong at the moment? This quote from another book* in the NATO ASI Series provides a succinct answer:

" ... Computers can provide a conversational environment in which the learner can apply knowledge to problems and consider their actions as reusable events. Learners can control their learning, learn from others, and develop reflection on actions as metacognitive skills .... We believe that learning environments should support collaborative construc­tion of knowledge involving both teachers and students. [ ... J Collaborative knowledge construction environments enable all members of a class or learning

Foreword XI

group the opportunity to contribute their interpretation. It is important for advanced knowledge acquisition that learners realise that there exist multiple interpretations for every event or object. Those interpretations may be dissonant or consonant, but they reflect the natural complexity that defines most advanced knowledge domains. Collaborative environments enable learners to identify and reconcile those multiple perspectives in order to solve problems."

* from: Jonassen, D., Mayes, T., and McAleese, R. (1993) A manifesto for a constructivist approach to uses of technology in higher education. In: Duffy, T.M., Lowyck, J., Jonassen, D.H. (eds.) Designing Environments for Con­structive Learning. NATO ASI Series F, Vol. 105, Berlin, Springer-Verlag, pp. 231-247.

The effectiveness of collaborative learning approaches can be attributed to the fact that participants simultaneously experience the active construction of knowledge; that they engage in peer teaching and develop oral, written, and graphical explanation skills; that they learn from each other through exposure to different models for problem-solving and interaction; all this in a context which can provide motivating feedback from others. This is not to suggest that collaborative learning is a panacea to be applied to any subject, to any target group, and to any situation. Participants need some prior level of competence in the domain, they need to share and understand common goals, to respect and trust each other, and to accept that decisions do not have to be based on consensus. If these conditions can be fulfilled amongst a group of dispersed learners, then it is highly probable that groupware technologies, with their range and variety of communication modes and group support tools, can provide the necessary infrastructure for effective collaborative learning.

The pressures underlying the development of flexible education, training, and re­training programmes for school-Ieavers and adults, and the increasing emphasis on the importance of team work in many organisations, provide a favourable socio­economic climate for the development of collaborative learning approaches. Technological developments in computer networking and in design of software for support of group work provide an increasingly accessible infrastructure for social networking and exchange of knowledge and skills. In an era when the growth of users of the InterNet has become exponential (under 2 million in January 1991 to over 10 million by the end of 1992), and when corporate mergers between telecoms operators, electronics manufacturers, software companies, and broadcasting and cable organisations are becoming commonplace in an international market, this book will have an important role to play in raising basic questions about the effective use of interactive technologies in education and training at both local and global levels.

Anthony R. Kaye

Institute of Educational Technology, Open University, Milton Keynes, UK. UK Representative, NATO Science Committee Special Programme Panel on

Advanced Educational Technologies, 1992-93

Table of Contents

Introduction .... ..... ... ...... .... ...... ...... ......... ..... ..... ... ........... ..... ......... ... .......... ....... 1 M. Felisa Verdejo, Stefano A. Cerri

Part 1 Pedagogy and Learning Experiences

Pedagogy

1 A Pedagogical Framework for CMC Programmes ..................................... 11 Morten Flate Paulsen, Beatriz Barros, Peter Busch, Benita Compostela, Mirabelle Quesnel

2 Computer Support for Collaboration .......................................................... 21 Thore Danielsen

3 Some Pedagogical Techniques for Computer-Mediated Communication.. 33 Morten Flate Paulsen

4 Learning in Groups: Some Experiences of Online Work David McConnell

Learning Experiences

5 Learning Experiences with Collaborative Working Technologies:

46

Some Critical Factors ...................................................................................... 60 Cristina Simon

6 Towards Collaborative Learning at a Distance Gary Alexander, Paul Lefrere, Steve Matheson

7 An Experimental Network-Mediated Study Support System in

65

Higher Distance Education .............................................................................. 78 Henk Ellerman, Ad Schellekens, Willibrord Huisman

8 Innovative Support Technologies for Tele- and Team-Work at Universities 97 Ralph Seitz, Freimut Bodendoif

9 Remote Collaboration in Medicine: Some Tools and Experiences in Portugal 109 A. Sousa Pereira, I. Amaldo Martins, 1.A. Veiga Pires

Part 2 Models and Systems

10 Models and Systems for Collaborative Dialogues in Distance Learning.. 119 Stefano A. Cerri

XIV Table of Contents

Models

11 Cooperative Open Systems Architecture .................................................. 126 Christian Lemaitre, Victor German Sanchez, Cristina Loyo

12 The Role of Knowledge Based Systems for Automatic Coordination in Distance Learning ........................................................................................... 142 Jose Cuena, Ana Garcia-Serrano, M. Felisa Verdejo

13 Supporting Collaborative Dialogues in Distance Learning Carla Simone

14 Towards Models of Interaction Between an Artificial Agent and a

156

Human One ..................................................................................................... 170 Dominique Guin

15 The "Natural Laboratory" Methodology Supporting Computer Mediated Generic Dialogues ...... .......... ......... ........ ........ ...... ...... ......... .................. ........... 181 Stefano A. Cerri

16 Building an Evolving Knowledge-Base from Computer Teleconferencing 202 Andre Boder, Christine Gardiol

17 Educational Scenarios for Telecommunication Applications Peter Zorkoczy

Systems

18 Developing a Tool to Support Collaborative Dialogues and Graphical

210

Representation of Ideas ................................................................................... 219 M.B. Twidale, T. Rodden, l. Sommerville

19 Real-Time Multimedia Conferencing System and Collaborative Learning 236 A. C. Derycke, C. Vieville

20 Distributed Multimedia Environment for Distance Learning Encarna Pastor, Gonzalo Sanchez, Javier Alvarez

258

21 Production of Flexible and Modulized Course Material in COSYS 270 Mette Ringsted, Finn Groenbaek, Peter Busch, Hanne Shapiro

22 OSCAR: A System for Collaborative Distributed Authoring of Multimedia Training Materials ....................................................................... 282 Antonio Ulloa, Antonio De Girolamo, Stephen Delaney

List of Contributors ................................. ........................................................ 291

Subject Index .. .......................... ............. ... .................................... ......... .......... 293

Introduction

M. Felisa Verdejo and Stefano A. Cerri

What this book is about

We announced in the justification for the workshop that the title collaborative dialogue technologies stresses the focus on conversational models of group interaction. Our intention was to indicate our priorities in the list of the potential topics to be covered under the more generic heading of distance learning. In this sense, the choice about the book title has influenced the outcome. Differently from other edited publications, the book presents work in progress at the moment with a special emphasis on modelling dialogues in general and learning dialogues in particular, at a distance through a computer network.

Research, dialogue technologies, learning

First of all, the book is the outcome of a Research Workshop. Secondly, it is about dialogue technologies. Thirdly, it is about learning. Let us analyze a bit more critically the three components with respect to what one may expect or not expect today from such a book; i.e. from the viewpoint of the user of the book.

A book emerging from a research workshop cannot include an established corpus of knowledge organized as an introduction to a consolidated domain. If research is needed, and there are people investing time and efforts in performing research, that means that there are problems in the area, that some problems have been identified but not yet fully solved, and other problems are expected, even if they have not yet been identified. As we all know, it is much easier to solve a problem that has been identified and formally described, than to identify a problem from a situation. In this sense, the book is a real research output: it attempts to define problems and to contextuaIize them with respect to situations that are likely to profit from a possible solution of those problems. It is not a book of recipes, it is not a book of answers to consolidated questions. We do not need answers now, because we are even not sure if the questions are correct. Research is needed first of all to find out what are the correct questions.

Secondly, the book is about dialogue technologies, i.e. computer technologies that support dialogues among humans and between humans and computers. These technologies are emerging but are far from being consolidated or available for applications. For instance, the editors of this book wanted to use one of these technologies (formatted file transfer) in order to edit the book collaboratively at a distance. We have not been able to do it this way. Everything is "in principle" available, but in fact one has to spend time and energy to do it, so that it does not emerge as an effective priority. This experience of ours supported once more our general view about the specificity of dialogue technologies, that is outlined in the following pages.

2 M.F. Verdejo, S.A. Cerri

Dialogue technologies cannot be evaluated as other technologies: their intrinsic value is not in what they can eventually in principle do, but in what they effectively do now for me. They are technologies intrinsically linked with human factors. The technical component of the technology is just one part; the rest is the link with the human, with the real human situation. It is important to stress this notion, because one would be tempted to forget that the combination of the human and the artifact does not have the sum of the properties of both, so the evaluation criterion for technologies dedicated to support the human activity will heavily depend on the human acceptance of the outcomes of the combination.

Let us try to make this concept even more clear by means of an analogy in another domain: the conception of the content of Computer Science. If you ask a theoretical computer scientist what computer science is about, (s)he will answer that computer science is about what is computable. If you ask an experimental computer scientist what Computer Science is about, (s)he will answer that Com­puter Science is about what people compute. The difference is relevant. We do not want that problems are shown to be eventually solved in principle, we want to solve them now effectively and efficiently. Therefore, for us (the editors of the book) Dialogue Technologies offered not even a remote formatted file transfer facility, because even if that facility was offered in principle we considered - after a few efforts - that it would not be an efficient solution for us .

This concrete attitude is not to be confused with a purely applicative one. In applications, one assumes 1) to have identified the problem and 2) to have the tools for assembling the solution. Most often, in our CDTDL domain, we are not in this situation. Historically, in the domain of technologies for Education and Training there have been several failures exactly for this reason. People assumed the two propositions above before checking if they were eventually false. A success occurred only in the relatively few cases that - once those propositions were shown to be verified - the application was correctly build and used.

It is not so unexpected that people with an applicative attitude forget to check either of the two propositions (or sometimes both). People who have an educational need often forget to estimate the availability (now at affordable costs) of the tools; people willing to apply the tools often forgets that the problem has to be first identified by the potential user, and often offer the most advanced technology to somebody that is not willing to search for a problem.

The third aspect of the book is learning. Learning is crucial and strategic; there is much more demand that can be met, everywhere. Unfortunately,learning (and its associated initiatives: education, training, retraining, etc.) absorbs approximately 10% of the resources in the industrialized countries but 99% of that 10% is historically devoted to the salaries or to the buildings of the institutions committed to provide the teaching.

One important outcome of the workshop is that learning better ...i a direct effect of teaching, can be considered a side effect of many activities. If that is correct, then the result would be to diminish the resources dedicated to teaching and augment those that facilitate learning. Even a small redirection of those huge resources would have an expected tremendous impact. However that shift in the investment is unlikely. First of all due to the socio-economic crisis that affects our countries, it is difficult to increase the resources available, and even more difficult

Introduction 3

to modify the organization of labour (the human resources around the teaching). Even if the companies active in the Information and Telecommunication Tech­nologies, including those active in the publishing sector would claim and perhaps demonstrate that such a shift would augment the available working places (from the teaching sector to their own) it is not likely that the change will take place massively in the next future. The transition will occur, but will be very slow with respect to the urgency ofthe demand. Nevertheless, we feel that it is our own duty, as researchers and practitioners, to point out to decision makers, the opportunities offered today by the technologies, and their potential cost/benefit in order to influence their priorities.

Value of the book and expected audience

We would like to point out that there are three main positive aspects of the book: it is a real overview of the state of the art today; it presents a genuine integration of viewpoints emerging from the educationalists - providers of training - users of the technologies and of the scientists - engineers - producers of the technologies and finally it offers a reasonable balance between the issues in fundamental research that emerge from the area and the applicative concepts. The value of the book is not in any of these aspects alone, but in the integration of all of them.

Therefore we believe that it will be useful mainly for three classes of readers: decision makers who need a fast reference to the area (particularly facilitated by the work on the subject index); practitioners who should evaluate more in depth what each issue entails (and wish to look at the references cited by each of the authors of the papers) and students as well as more established researchers who need ideas for further investigation. The book is not an introduction to the domain, but rather a reference for the domain, even if a few chapters present a quite clean overview of the available conceptual and computational options.

The organization of the book

Papers in this collection are grouped in two parts. Each part contains papers broadly discussing issues concerning the nature of collaboration, mediated com­munication and distance learning, and other ones presenting a concrete proposal or experience focusing the discussion on a particular aspect of the learning process.

Part I is more oriented to didactic issues and applications presenting on the one hand, a number of pedagogical approaches to carry out collaborative learning for a variety of learner groups, and on the other hand, pointing out key issues involved in the setting of collaborative learning environments using computer mediated technologies.

Part II groups papers presenting models or systems to carry out a range of tasks and activities involved in the learning/teaching processes. Their presentation stresses the dialogue situations, and how this point of view is taken into account for the software design of their particular proposal.

A detailed description of each part follows.

4 M.F. Verdejo, S.A. Cerri

Part I: Pedagogical Issues and Learning Experiences

The papers in Part I are grouped in two sections. The first one (Chapters 1-4) addresses general concerns about the nature of group collaboration and distance learning both from the point of view of the supporting technologies and the educational aspects. The second section, Chapters 5-9, includes papers presenting a range of applications involving adult distance education, concerning either students at the university level or professional training.

Pedagogical Issues

Chapter 1, by Paulsen et al., report on the group discussion carried out during the workshop about didactic related topics. An attempt is made at characterizing a CMC-based learning environment. Resources, constraints, demands and choices are considered as relevant dimensions for establishing a general framework in terms of which pedagogical techniques can be established. Danielsen, in Chapter 2, revisits groupware and CSCW technologies to discuss the elements of a platform for collaboration pointing at those aspects relevant for distance learning. Chapter 3, by Paulsen, reviews a variety of pedagogical techniques that have been proposed for learning through computer mediated communication. Each one is described with references to the literature, pointing at the applications where they have been experimented. The conferencing environment, as compared to the face-to-face situation, is analyzed by McConnell in Chapter 4, looking at transcripts of conference discussions. He focuses on the view of learning as a social process and points out at elements in the group dynamics which differ when using computer mediated communication.

Learning Experiences

Chapter 5, by Simon, summarizes the workshop group discussions about critical factors for the setting of distance learning applications. The four following chapters detail particular experiences, specifying the group communication technology and the software facilities provided. The cases represent a variety of settings: the first and second ones concern Open Universities, while the third one relates to a traditional University and the fourth illustrates an example of professional training. Alexander et al. discuss in Chapter 6 a learning approach to provide collaboration for distance learning. Their ideas are illustrated through the design of an experi­mental course, based on an Interactive Learning Support System developed in the framework of a project to create "an electronic open university". On the other hand, Ellerman et al. review in Chapter 7 the current state of affairs in the practice of distance education. As a way to overcome the observed difficulties they propose to start with a network-mediated study support system, a first step to introduce a more advanced technology in the future. Chapter 8, by Seitz and Bodendorf, outlines a pilot project to explore tele and team work for students at the University, in order to provide them with tools to participate in decentralized courses and learn to work in a collaborative way. A range of in-service training involving healthcare is described by Sousa et al. in Chapter 9, the final paper of this part.

Introduction 5

Part II: Models and Systems

Part II covers work on models and systems to provide computer supported collaboration for distance learning, facilitate computer mediated dialogues and study how this combination can promote efficient learning. Most of the papers focus on learning processes but other related activities such as group coordination, authoring and course production are also described. Chapter 10, by Cerri, introduces this part, looking at questions representing

controversial themes throughout the participant's discussions such as the need for specific models for learning and teaching. He also reports on the main points of convergence reached at the workshop.

The Models section follows, consisting of Chapters 11-17, while the Systems section comprises Chapters 18-23.

Models

In Chapter 11, Lemaitre et al. develop a computational model for a cooperative network integrating humans and artifacts, the model follows ideas from distributed artificial intelligence, and is based on an explicit representation of the cooperation strategies established by the agents integrated into the network. The proposed architecture is outlined and some components related to dialogue activities are further discussed through an example.

Chapter 12, by Cuena et at., focuses on the organizational model of distance education to propose a model actively supporting coordination and collaborative activities. A knowledge-based architecture is presented and illustrated in two case studies. The importance of a knowledge-based level is stressed in order to accomplish two main requirements in these kind of environments: adaptability of the system to new group collaboration criteria and explanation of the system's proposals.

Chapter 13, by Simone, deals with the automatic management of dialogues based both on the idea of semi-structured messages and the language/action perspective for modelling collaboration. She discusses why this approach is suitable for distance learning settings and shows how CHAOS, a generic cooperation system can be used for supporting collaboration in distance learning.

Chapter 14, by Guin, focuses on modelling dialogues in Interactive Learning Environments (ILE) using machine learning techniques. She applies a previous formalism used to build a learner model, to represent now the didactic interactions between an artificial agent (a teacher in the case of and ILE) and a learner. The aim is to provide a precise way for describing interactive educational situations distinguishing the agents, the roles they can play during the process and the management of the role interplay.

In Chapter 15, Cerri presents the NOBILE methodology for managing computer mediated dialogues. Starting from tutoring situations, he first describes how to use the method for student model acquisition through dialogues and then, he proceeds to generalize the method in various dimensions, for instance the type of problem solving task, or the type of dialogue situation.

A method for knowledge acquisition, based on the reification of interactions between professionals when solving a case, is presented by Boder in Chapter 16.

6 M.F. Verdejo, S.A. Cerri

The purpose is to build a knowledge base in such a way that knowledge can be easily retrieved and accessed "just in time".

In the last paper of this section, Zorkoc'Y elaborates the idea of "scenario" for the analysis and design of technology supported systems, pointing at ongoing industrial projects where this framework is been applied.

Systems

In Chapter 17, Twidale et al. describe the Designer's NotePad, a tool to support the refinement of ideas for design processes. They discuss the potential for supporting collaborative dialogues, and stress the importance of the interface, proposing a rapid prototyping approach.

In Chapter 18, by Derycke et al., the CO-LEARN project is presented: their final aim is to support various collaborative learning situations. The paper focuses on the real-time multimedia conferencing system: first the software architecture is detailed, secondly an example is presented, and finally the main issues for the design of a multimedia collaborative learning environment are discussed.

In Chapter 19, Pastor et aL report on the EDUBA project where they claim the adequacy of the unstructured communication paradigm to approach the design of a distributed collaborative multimedia educational environment.

The last two chapters are dedicated to course production and co-authoring respectively. Ringsted et al. discuss a new approach to the production of flexible and multimedia learning material, while Ulloa et al. outline OSCAR, a system supporting collaborative and distributed authoring of multimedia materials. The system supports coordination, codecision and re-use, these services being integrated in the co-authoring space.

Conclusions and Trends

There are a few clear trends that emerge as an outcome of the book. 1. Learning as a side effect. This means that CDTDL are, in fact, Collaborative

Dialogue Technologies and the side effect of those technologies (learning) has to be achieved and verified by those that are interested in learning (the educators, managers of education etc.) while the impact of the scientists and engineers is necessarily wider that only for learning. For instance: authoring systems are built for authoring in general, and cooperation in Computer-Mediated Communication is relevant for any collaborative intellectual activity.

2. Dialogue technologies cannot provide for purely "passive" (e.g. file transfer) support but need to become "active". In order to have that, the syntactic level of description of messages is not sufficient, a semantic level is needed. In order to design and develop that level, the first issue is represented by standards (conceptual and computational).

3. The computer is used more as a communication medium and less as a computational device: communication IS computation + a carrier (syntactic). It is important to note here briefly that the computational part of communication has to do with coding and decoding, i.e. with the semantic and syntactic level.

Introduction 7

4. The contribution of the discipline of Computer Science is more that of a methodology for conceiving and representing new organizational processes than one that claims simply realizing the corresponding support tools for managing them. Here concepts such as reflection and mediation, corne into focus.

Each of these conclusions can be controversial with respect to the most common understanding of the area. Both traditional "educationalists" and "computer scientists or engineers" should wonder about these results that, maybe, redefine the main focuses of research, development and implementation initiatives: we have to assume, in the future, that it is not necessary nor sufficient to design systems that teach in order for people to learn, to propose networking technologies in order for people to communicate, to have powerful computational resources in order to solve the most difficult problem of the effectiveness of the human-computer interface and the induced reorganization of human activities. This is not bad news; on the contrary, taken to the extreme, down to the fundamental principles of common sense in the design and application of technologies guided by an adequate under­standing of the problems to be solved. If and how these conclusions will be shared by the readers of the book, is a question that will be answered in a few months by your reactions. For the moment we are aware of the opportunities, difficulties and pitfalls of the domain and are quite happy to have served the community of readers to enhance the first. overcome the second and avoid the third.

Part One

Pedagogy and Learning Experiences

1 A Pedagogical Framework for CMC Programmes

Morten Plate Paulsen 1, Beatriz Barros2, Peter Busch3, Benita Compostela4, and Mirabelle Quesnel5

1 NKI, P.O.B. 111, 1341 Bekkestua, Oslo, Norway, e-mail: morten@nkLno 2 Esc.de Informatica, Km7 Carretera de Valencia, Univ. Polit., E-28031 Madrid, Spain 3 Danish Technological Institute, Teknologiparken, Aarhus, DK-8000 Denmark 4 Depto Psicologfa Social, Univ. Complutense, Madrid, Spain 5 Grenoble Pole Europeen, B.P 52, F-38402 Saint-Martin d' Heres, France

Abstract. This paper presents a pedagogical framework for computer-mediated communication that comprises the programme environment, the universe of CMC resources, the programme's CMC environment, pedagogical techniques, and pedagogical styles. The programme environment is defined by the programme's constraints, demands, and choices; the universe of CMC resources comprises all external and internal CMC resources. As a common subset of these realms, a programme's CMC environment must be established. Within this CMC environment, programme planners and teachers can utilize a set of pedagogical techniques and pedagogical styles to facilitate learning.

Keywords. Computer-mediated communication, pedagogical framework, pedagogical techniques, pedagogical styles.

1 Introduction

This paper presents the results from a working group intending to establish a framework for planning and analysis of programme design and teaching with regard to computer-mediated communication (CMC). It is the group's opinion, that it is impossible to provide ideal models for every conceivable CMC programme. It is, however, the group's intention to provide a broad overview of issues to be considered in educational CMC programmes and to support practitioners with some basic guide-lines for programme planning and teaching in CMC networks. To this end, the group has chosen to present and categorize a series of factors that can influence the effectiveness of the learning process in CMC environments. The framework is meant to provide guide-lines to issues that need to be addressed during planning and implementation of CMC programmes. By discussing these issues, the group hopes that teachers and programme planners will acknowledge more of the opportunities and constraints related to their specific CMC programmes.

The structure of the framework is not an indication that some of the elements will exclude others. On the contrary, it is the opinion of the group, that it is possible to apply many different pedagogical techniques and styles in one specific programme, as it will be feasible to apply many different programmes to a set of

12 M.F. Paulsen et al.

chosen techniques and pedagogical styles. In other words, the group perceives CMC pedagogy as a very complex and comprehensive construct, having no paragon application, but a lot of feasible implementations. It is to be hoped, though, that this framework may help teachers and programme planners to provide better CMC programmes in the future.

2 The Framework

This paper presents a pedagogical framework for CMC programmes. The framework is illustrated in Fig. 1 and is elaborated on in the following text. The programme environment defines the programme's opportunities and limitations and the universe of CMC resources comprises all existing CMC resources. As a common subset of these realms, a programme's CMC environment must be established. Within this CMC environment, programme planners and teachers can utilize a set of pedagogical techniques and pedagogical styles to facilitate learning.

The universe of CMC sources: People: Individual experts and online interest groups Information: Online joumaJs and online databases Applications: OnUne applications, software libraries

The program The program's CMC environment:

environment: Pedagogical techniques: one-aIone

Constraints one-to-one Demands one-to-many Choices many-to-many

Pedagogical styles:

philosophical orientation moderator role

facilitation techniques

Fig. 1. A pedagogical framework for CMC programmes

A Pedagogical Framework for CMC Programmes 13

2.1 The Programme Environment

The choice of CMC resources, pedagogical techniques and pedagogical styles derives from the actual programme environment. Donaldson discusses the environ­ment in view of constraints, demands and choices. He argues that programme administrators should seek to:

"push back constraints, and work to have demands relaxed, thereby increasing the quantity, improving the quality, and expanding the types of choices available to them". (Donaldson 1990).

Constraints. There are many constraints that limit the opportunities for CMC utilization. Among these are the type of institution, geographical issues, equipment, resources, course time frame, course workload, communication pattern, and the financial health of the programme provider. The programme must comply to the rules, regulations, and policies of the institution and these may vary considerably from a private to a public institution and from secondary education to universities. A programme is often confined to a certain geographical area such as a campus, a community, a state, or a nation. With regard to CMC, these restrictions are more often due to policies and legislation than to technical limitations. Lack of computer resources, such as hardware, software, and communication networks, is though, an important limitation for many CMC programmes. The institutions scheduling policy could pose several restrictions on a CMC course. It is not always convenient for a CMC course to follow a university semester or term plan. In the same way, requirements of a weekly course load could constrain a programme. Some institutions may also require some sort of synchronous communication that further constrain a programme.

Demands. National legislation and parent organizations form policies and procedures that cannot be ignored. There are demands for quality programmeming and healthy finances. Further, the local community, faculty, staff, and students all have demands to the programme. These demands may, of course, be more or less rigid, but together they are an important element of the programme environment.

Choices. Some programme choices could be the programme's actual target group, its objectives, and to a certain degree the available resources. The use of CMC may depend on the programme's actual target group. Important target group variables are group size, educational level, CMC expertise, computer aptitUde, motivation, and financial status. The group size may vary from one to more than one thousand participants, but very few CMC courses have more than fifty participants. CMC programmes are available at primary level, secondary level, undergraduate level, graduate level, and continuing education level. In addition, the prospective students' motivation, financial status, and experience with CMC and computers vary considerably.

A programme's use of CMC will further depend on the programme objectives. Important aspects of the objectives are the programme discipline and subjects, such as science, arts, or business related subjects. Of further importance is the formal character of the programme, i.e. whether it is a diploma course, a vocational course, or a purely informational programme.

14 M.F. Paulsen et al.

The institutional resources are of vital importance to a CMC programme. The resources include employed and affiliated staff and faculty, CMC technology and proficiency, and, of course, the institution's financial status and its willingness to invest in CMC.

2.2 The Universe of CMC Resources

The universe of CMC resources is constituted by all the resources accessible via every existing CMC system. The resources could be internal - provided by the institution, or external - made available from other institutions. Whether they are internal or external, these resources can be labelled people, information, and applications. It is from this set of resources that programme planners may choose when they establish a programme's CMC environment.

People. Millions of individual experts and thousands of online interest groups are reachable via external CMC networks. These constitute a tremendous resource for lifelong learning. Individual experts can be consulted and interviewed via e-mail. An online interest group (OIG) is a group of people with a common interest who convene via CMC. There are thousands of OIGs that can be accessed via international CMC networks and it can be argued that they all have some sort of educational purpose. Howse (1992) states that more than 1,000 scholarly lists are distributed via Listserv on Internet and that over 1,000 international newsgroups, carrying more than 250,000 items every day, can be accessed at Murdoch University in Australia. Internally, the institution can choose to provide access to local conferences and individual teachers, peer students, and support staff.

Information. A growing number of databases and electronic journals are available through external CMC networks. They probably cover most of the subjects that are taught in online courses. Online databases are organized collections of data that can be accessed via CMC. Utilizing these external resources, a course provider could maintain local databases of relevance to the courses. An easier solution, though, could be to provide access to international databases. Online journals are periodicals that are distributed to subscribers via CMC networks. They are increasingly important resources for information and learning. Supporting this statement, Strangelove (1992) has compiled a directory of about 35 electronic journals and 90 newsletters that are available via Internet. Local bulletin boards could be used to redistribute online journals and to disseminate local information.

Applications. An enormous number of software applications is available via external and internal CMC networks. Online applications are software programs that can be executed on a remote computer via a computer network. They include a range of applications from software development tools; via specific applications for statistics, economical analysis, etc.; to computer aided instruction applications. A related, but slightly different approach, is to establish a software library for software applications that allows remote users to download software applications from a host computer so that they later can execute the programs on local microcomputers. Such files are available from a number of host computers.

A Pedagogical Framework for CMC Programmes 15

Internet provides a standardized file transfer protocol (FTP) for this purpose and a large number of PC based bulletin board systems have software exchange as their main service.

2.3 The Programme's CMC Environment

Each programme has a CMC environment. This environment is a subset of the universe of CMC resources and it comprises the programme's internal and external CMC resources. It is more or less judiciously included to support the educational process. It is further influenced by the programme's constraints and demands. The programme's CMC environment defines opportunities and constraints for the pedagogical techniques and the pedagogical styles. The environment could be described in The CMC Environment Matrix as shown in Table 1.

Table 1. Example of a CMC environment matrix

Resources Students Faculty Staff External online interest groups Access Access Access External individual experts Access Access Access Internal conferences Access Access Access Internal students, faculty, staff Access Access Access External online databases External online journals Access Access Access Internal online databases Internal bulletin boards Access Access Access External online applications Access External online software libraries Access Internal online applications Access Access Internal online software libraries Access Access

The CMC Environment Matrix should be further detailed with regard to internal conferences and bulletin boards that are established for educational purposes. As shown in Table 2, this could be done using the Framework Matrix devised for a goal-oriented method for the establishment of an electronic college (paulsen 1992). After deciding which user groups to serve and what resources are to be provided, an Environment Matrix can be established. The user groups are placed along the horizontal axis and the selected resources along the vertical axis. A determination must be made for each cell in the matrix as to the extent of access for each user group to the corresponding information. User privileges are designated "no access", "read only access", and "read and write access". Often, more subtle access delineations such as "censored" and "mandatory" are useful.

16 M.F. Paulsen et al.

Table 2. Example of a framework matrix

Conferences/

Bulletin Boards

Curriculum board User manual board Published articles Admission office Faculty club Administrative board Exam board Course boards Course conferences Cafe conference Presentation board

Students

Read Only Read Only Read Only No Access No Access Read Only Read Only Read Only ReadIWrite ReadIWrite ReadIWrite

2.4 Pedagogical Techniques

Faculty

Read Only Read Only Read Only Read Only ReadIWrite Read Only Read Only ReadIWrite ReadIWrite ReadIWrite ReadiWrite

Staff

ReadiWrite ReadIWrite ReadiWrite ReadIWrite ReadIWrite ReadiWrite ReadIWrite Read Only ReadIWrite ReadIWrite ReadiWrite

A pedagogical technique is a manner of accomplishing teaching objectives. Pedagogical techniques may be organized according to the four communication paradigms that are frequently used in CMC. The four paradigms correspond to the elements of CMC--information retrieval, electronic mail, bulletin boards, and computer conferencing. A similar model is presented in a paper by Harasim (1989). Presenting "the Collaborative Learning Horizon", she distinguishes between one­to-one, one-to-many, and many-to-many learning approaches. In addition, she points out that the many-to-many approach could be moderated by a teacher or it could be unmoderated, i.e. lead by the students themselves without much intervention by the teacher. The techniques do not, however, change significantly whether the teacher is directly involved in the group interaction or not. The foregoing considerations result in a framework of four classes of techniques. First, the techniques classified as one-alone techniques are characterized by retrieval of information from online resources and the fact that a student can perform the learning task without communication with the teacher or other students. Second, the techniques classified as one-to-one communication can be conducted via e-mail applications. Third, the techniques discussed as one-to-many techniques will typically be conducted via bulletin boards or distribution lists for e-mail. Finally, the techniques presented as many-to-many communication can be organized within a computer conferencing system. Several techniques could be included in each of the four classes as shown in Table 3.

2.5 Pedagogical Styles

In this article, pedagogical style is perceived as a multifaceted construct. It is suggested that the facets of special importance to moderation of educational computer conferences are the moderator's philosophical orientation, the moderator's role and the facilitation techniques applied by the moderator.

A Pedagogical Framework for CMC Programmes l7

Table 3. Some pedagogical CMC techniques

One-alone Online databases Online journals Online applications Software libraries Online interest groups Interviews

One-to-one Learning contract Mentorship Apprenticeship Internship Correspondence In-basket

2.5.1 Philosophical Orientation

One-to-many Lecture Symposium Panel Skits

Many-to-many Debate Simulation Role play Case study Discussion Transcripts Brainstorming Delphi Nominal Group Observation Forum Project Group Hearing Cognet Jigsaw

The moderators' pedagogical styles are based on their philosophical orientations and theories toward education. Discussing adult education philosophies, Zinn (1991) argues that a teacher's philosophy of education may be unrecognized, inconsistent and just partially formulated, but that it still provides a basis for the teacher's facilitation of learning. She further distinguishes between liberal, behaviorist, progressive, humanistic, and radical philosophies. These and other philosophies in adult education are presented in selected writings edited by Merriam (1984) and Jarvis (1987). With regard to distance education, Keegan (1988) identifies three theoretical positions; theories of autonomy and independence, theories of industrialization, and theories of interaction and communication. Discussing these theoretical positions, Paulsen (1992b) presents "the Theory on Cooperative Freedom" which is a first attempt to establish a distance education theory attuned to CMC. So, summing up, moderators will perceive their role in educational computer conferencing in the light of their basic theories and philosophies toward education.

2.5.2 Moderator Roles

Discussing group dynamics in general, Forsyth (1990) classifies roles in two basic functions, task roles and socioemotional roles. More detailed and focusing on computer conferencing, Mason (1991) identifies three roles functions that computer conferencing moderators must possess. These functions are:

Organizational role. One of the first duties of an online tutor is to set the agenda for the conference: the objectives of the discussion, the timetable, procedural rules and decision-making norms. Managing the interactions with strong leadership and direction is considered a sine qua non of successful conferencing ....

18 M.F. Paulsen et al.

Social role. Creating a friendly, social environment for learning is also seen as an essential moderator skill. Sending welcoming messages at the beginning and encouraging participation throughout are specific examples, but providing lots of feedback on students' inputs, and using a friendly, personal tone are considered equally important .... Intellectual role. The most important role of the online tutor, of course, is that of educational facilitation. As in any kind of teaching, the moderator should focus discussions on crucial points, ask questions and probe responses to encourage students to expand and build on comments .... (Mason 1991)

Brochet (1989) discusses these roles: • The goal-setter, who makes plans for the conference and decides whether the

plans should be changed during the conference. • The discriminator, who differentiates between useful and useless ideas. • The host, who creates the feeling of trust and motivates contributions. • The pace setter, who removes communication barriers and promotes

cooperation. • The explainer, who relays overlooked messages and raises questions that have

remained unanswered. • The entertainer, who evaluates the conference mood and ensures that participants

are relaxed.

Additional roles could be classified as lecturer, tutor, facilitator, mediator, mentor, assistant, provocateur, observer, participant, etc. Each of the roles will have some organizational, social, and intellectual elements as suggested in Table 4.

Table 4. Some pedagogical role models

Role

goal-setter discriminator host pace setter explainer entertainer lecturer tutor facilitator mediator mentor assistant provocateur observer participant

Organizational function

high

high

high high high low

low low

Social

function

high

high low

high high

low

Intellectual function

high

high high

high high high

high

high low

A Pedagogical Framework for CMC Programmes 19

2.5.3 Facilitation Techniques

A few authors (Eisley, 1991; Hiltz, 1988; and Feenberg, 1989) discuss moderation of educational computer conferences in particular. In table 5, their recommended facilitation techniques are compiled and organized according to the role functions. The advice is mostly based on personal experience in specific contexts, so, it may or may not be relevant to other moderators in different contexts.

Table 5. Some recommended facilitation techniques

Organizational facilitation

Spur participation when lagging Require regular participation Use response activities Move misplaced content Handle tangents appropriately Vary participation Have students conduct discussions Give decisive ends to discussions Invite visiting experts Don't overload Read the status report daily Don't lecture Prompt frequently Use simple assignments Be clear Set up student interaction Synchronize and resynchronize Take procedural initiatives Remember the law of proportionality Be patient

3 Conclusion

Social facilitation

Reinforce good discussant behaviors Request change in poor discussant behaviors Hang loose Be responsive Request meta-comments

Intellectual facilitation

Summarize the discussion Write weaving comments Respond to student contributions Make the material relevant Present conflicting opinions Simulate an agent provocateur Be objective Expect less Don't rely on off-line materials Request responses

The framework presented in this paper is not meant to provide ideal models for every conceivable CMC programme. It is, however, intended to provide a broad overview of issues to be considered in educational CMC programmes and to support practitioners with some basic guide-lines for programme planning and teaching in CMC networks.

20 M.F. Paulsen et a1.

References

Donaldson, J. F. (1990) Managing credit programmes in continuing higher education. Urbana-Champaign, Illinois: The University of Illinois at Urbana-Champaign

Eisley, M. (1991) Guidelines for conducting instructional discussions on a computer conference. In: A. Miller (ed) Applications of computer conferencing to teacher education and human resource development. 35-39. Proceedings from an international symposium on computer conferencing at the Ohio State University, June 13-15

Feenberg, A. (1989) The planetary classroom: International applications of advanced communications to education. In: E. Stefferud, O. J. Jacobsen, and P. Schicker (eds.) Message handling systems and distributed applications. 511-524. Amsterdam: North-Holland

Forsyth, D. R. (1990) Group dynamics. Second edition. Pacific Grove, California: Brooks/Cole Publishing Company

Harasim, L. (1989) On-line education: A new domain. In: R. Mason and A. Kaye (eds.) Mindweave: communications, computers, and distance education. 50-62. Oxford: Pergamon Press

Howse, W. J. (1992). The Internet: Discoveries of a distance educator. EDU Magazine, Spring/Summer 1992, Issue Number 58

Hiltz, S. R. (1988) Teaching in a virtual classroom. Volume 2 of A virtual classroom on EIES: Final evaluation report, Newark, New Jersey: New Jersey Institute of TeChnology

Jarvis, P. (1987) Twentieth century thinkers in adult education. London: Routledge Keegan, D. (1988) On defining distance education. In: D. Sewart, D. Keegan, and B.

Holmberg (eds.) Distance education: International perspectives, 6-33. London: Routledge

Mason, R. (1991) Moderating educational computer conferencing. DEOSNEWS 1(19) Merriam, S. B. (1984) Selected writings on philosophy and adult education. Malabar,

Florida: Krieger Paulsen, M. F. (1992a) GO MEEC! A goal-oriented method for the establishment of an

electronic college. In: M. F. Paulsen. From bulletin boards to electronic universities: Distance education, computer-mediated communication, and online education. 38-45. University Park, Pennsylvania: The American Center for the Study of Distance Education

Paulsen, M. F. (1992b) The hexagon of cooperative freedom: A distance education theory attuned to computer conferencing. In: M. F. Paulsen. From bulletin boards to electronic universities: Distance education, computer-mediated communication, and online education. 56-64. University Park, Pennsylvania: The American Center for the Study of Distance Education

Strange love, M. (1992) Directory of electronic journals and newsletters. Ottawa, Ontario: Listserv@acadvml.uottawa.ca.

2 Computer Support for Collaboration

Thore Danielsen

Norwegian Telecom Research Sector Specific Solutions, P.O. Box 1156, N-9001 Tromsoe

Abstract. Since the early years of computers we have seen an ever increasing usage of this tool in a number of both work and "off time" activities. And ever since the first computer-based tools for mediating information between people were built, there has been a quest for the computer-based tools for establishing, maintaining and strengthening the interpersonal relations - providing computer support for collaboration. This paper is a journey through the enthusiasm towards groupware and CSCW along with the set-backs and scepticism towards the computer-support. It ends up with a discussion on the elements of a platform for collaboration through the use of computers. The discussion will be based on applications and experiences from NTRs search for sector specific solutions which may be of relevance for distance learning.

Keywords. Computer-mediated communication, groupware, CSCW, collaborative design, AMIGO, cooperative technology, Telemedicine, distance learning, local government, integration.

1 Looking Back into the Eighties - Dreams of a Bright Future

During the work with my thesis some time ago I came across a book of promises which was to become my looking glass for the new reality of fast, reliable and universal communication. The name of the book was a promise in itself - we should all get together in a networked nation (Hiltz & Turoff 1978). From this idea grew the specification of a system through which people could get together and exchange information as if they were involved in "just another" meeting. And the hopes of a new era with more power to the individual was fuelled by the ongoing discussions on the impacts of computers and computing which also found its way to Norway (Engebretsen & Jensen 1979). It seemed as a favourite pastime throughout Europe was discussing the relations between computers and computing, telecom­munications and the processing of information.

We felt that EIES was history, and COM was a way of communicating which belonged to the future. The Norwegian universities joined forces and established their own plan for introducing computer-mediated communication and the sharing of information across departments and national borders. And we continued pushing

22 T. Danielsen

towards the interaction among systems handling infonnation through, for instance, the Community sponsored project called GILT - Get Interconnected Local Text systems. GILT ended with a prototype and then a sudden death caused by an international standard called X.400. Suddenly we found computer-based messaging on more lips than ever. The bright future of COM and the like fell apart with the rapid introduction of electronic mail through the EAN networks. On a European basis, Norway "bought" the idea of a standard messaging service very early. The notion of groups interacting lost momentum. Computer-mediated communication was interpersonal and distributed, and groups were only addressed through personal distribution lists. But networking - and internetworking - gave hopes for a new­born life of computer-supported collaboration. And along came the idea of relieving the user of more tasks in the game of interacting through the specification and implementation of groupware systems. In the name of progress, we believed in the coding of interpersonal relations and behaviour in to the hard and softer parts of the computer. AMIGO for one, another Community sponsored project, worked out a model for describing activities - units of collaboration (Pankoke-Babatz 1989). Some works found their group descriptions given through theories on how we, people, interact verbally. This gave rise to systems (e.g. Coordinator (Flores et al. 1988)), projects and prototypes (e.g. Cosmos (Bowers et aI1988)) and more projects (e.g. Chaos (De Cindio et al. 1988)), neither of which proved a success with respect to numbers of implementations and installations.

Suddenly I found myself questioning whether this approach - implementing the group structure and representing the mechanisms supporting interaction among group members within the computer - would lead us closer to what we actually needed. It became apparent that there would have to be some discussions on how we reach the descriptions we use as a basis for our implementations, and what impacts do the descriptions themselves have on the products we deliver (Holand & Danielsen 1991). Another set-back, groupware could be running into another winter - the rise and fall of computer-supported collaboration? Intrigued by this question we could for instance throw a glance at studies on the use and impact of computer-mediated communication (CMC). And what we find (Danielsen 1993) are statements like:

use of CMC may lead to a reduction in the costs of coordinating groups of communicators, but may also delay decision making asynchronous interaction can never be incremental, which means that asynchronous communication is inappropriate for various types of activity that rely on negotiation or shared meaning because these require incremental communication users rate CMC as just as asynchronous as letter-writing even though electronic transfer is that much faster & E-mailing was construed to be equally spontaneous as face-to-face and telephone conversation some experiments also show that electronic groups maintain a significantly higher degree of contact and they have considerably less communication isolation than non-electronic groups working on comparable tasks

The studies from which these statements were adopted from (Eveland & Bikson 1988, Hahm & Bikson 1989, Hiltz 1988, Hiltz 1989, Lea 1991, Sproull and Kiesler 1991, Whittaker 1991, McConnick and McConnick 1992, Motiwalla & Nunamaker 1992) looked at the use of electronic mail and conferencing systems.

Computer Support for Collaboration 23

They do, however, provide valuable infonnation for requirements towards more juicy groupware to be designed.

A most promising future for CSCW could be found provided that we did not forget that "the design and construction of any computer-based artifact takes place in a process of cooperation. We do have to consider how we describe these processes, and how we employ the descriptions in the cooperation itself' (Holand & Danielsen 1991). And whatever fonnalisms and representations we may choose for our system supporting collaboration - or specification of such - we must always consider the limitations of these fonnalisms. Not only through asking what they are of, but also what they do - not through seeing them as outside what they represent, but as inside (Bowers 1991). We should carefully investigate the environments in which we aim at providing support for collaboration. And we should "keep in mind that cooperation may be viewed differently by the partners involved, and that the descriptions may change over time" (Holand & Danielsen 1991)1. "More so than any other computer discipline, CSCW has to evaluate why a system solution is needed and integrate the computer into the work setting ( ... ) the entire social setting of group work must be taken into account" (Henninger 1991). I suddenly found myself approaching a "user environment" in search of acts of collaboration, knowing that my mere existence in their environment caused them to have a different view of what I could potentially supply them with. Collaborative design could be "dangerous" in itself. Collaborative design of systems supporting collaboration even more dangerous. Fonnalisms would be needed to get any system working, but fonnalisms could strike back. One thing was learned: applications should be tailored to the collaboration in question, and applications should at least be adaptable and changeable. Whether we may reach a true adaptive environment (Danielsen et al1991) remains to be seen.

2 Descriptions Revisited: A First Look at Sector-Specific Solutions

In modern thought, (if not in fact) Nothing is that doesn't act,

So that is reckoned wisdom which Describes the scratch but not the itch.

(McLuhan, 1964)

On returning from a journey that took almost ten years I found that groupware was to be designed like the glove which could fit anyone, anywhere without disturbing what they were actually doing. Groupware was to be both simple and complex at the same time. And we had to head for something other than what we were heading for. I disliked the feeling of marching backwards2• Although we could learn aspects of

1 The paper presents three perspectives on cooperation: the strategy perspective, the coordination perspective, and the reflections and creativity perspective. This led us to descriptions of contexts for understanding cooperation. 2 "The past went that-a-way. When faced with a totally new situation, we tend to attach ourselves to the objects, to the flavor of the most recent past. We look at the present

24 T. Danielsen

cooperation through the use of formalisms like the ones given in the AMIGO work (Pankoke-Babatz 1989, Smith et al 1989), I felt certain that we would have to approach descriptions of user environments more slowly. And computers would have to be investigated on a broader scope than merely the support of interaction between users. Norwegian Telecom Research (NTR) has for a number of years attempted to investigate end-user-needs and supply end-user applications. This activity was strengthened in 1988 when the Telemedicine project was initiated. Today the Telemedicine project belongs to a research programme heading for "sector specific solutions" - the use of technology in specific sectors of the society. In a way, Telemedicine became a prototype on how we ought to approach the development of computer-support for collaboration. Surveys, observations and participatory design did all lead towards groupware or methods of applying existing technology in collaboration among health care workers.

Consequently, applications were not found by looking for where to apply what technology, but instead looking for where to work on cooperation and collaboration. And none of the descriptions given for generic situations - again, like AMIGO -would necessarily fit in any environment where the specified activity is considered to take place. What could have been learned from the AMIGO was the concept of roles, because roles could be the descriptions our future users longed for3• But roles are not as static and generic as we would like them to be, if we are to provide a full-blown copy of their descriptions in a computer-oriented manner. And more important than the explicit representation of the role which each user acted was the context in which the user acted. With such indications, permissive rather than prescriptive technology4 should be chosen.

3 CSCW and Groupware - State of the Art?

Systems for CSCW, or groupware, are still in their infancy except for the more basic conference systems, i.e. vanilla flavoured groupware. Some may claim it to be more of a success since more than two hundredS "groupware" applications, prototypes or projects may be listed. But a closer look at these show that just a few of these are commercially available systems with group support. A number of prototypes do also lack essential parts of their group support. It may also seem as if the concept of groupware itself needs clarification. A more likely method in achieving that is through an extensive classification of applications, prototypes or projects (Malm 1993). The scheme by which systems are classified is in itself of interest. The classification could for instance look at the technical, time and space, group structure or group activity aspects of systems (ibid.) or any combination of these.

through a rear-view mirror. We march backwards into the future. Suburbia lives imaginatively in Bonanza-land." (McLuhan & Fiore, 1967). 3 "The young today reject goals. They want roles - R-O-L-E-S. That is, total involvement, They do not want fragmented, specialized goals or jobs." (McLuhan & Fiore, 1967). 4 These concepts are discussed by, for instance, Galegher and Kraut (1990). S For lists of applications, please confer (Greenberg 1991 or MaIm 1993)

Computer Support for Collaboration 25

Any use of groupware is bound to impose changes on interpersonal relations and interorganizational behaviour. As soon as we introduce any new medium to the organization, a change in its relations may be recognized. Cooperative technology could be introduced to users from at least three different starting points:

technology in search of application: already there: technology and some knowledge on how it works could be there: an idea of usage problem: needs users technology with an urge for "upgrading": already there: users and some technology could be there: an urge for more technology problem: needs groupware organization looking for changes in behaviour: already there: people already cooperating could be there: knowledge on technology and means for providing more problem: who needs it?

The first could be tricky, but a non-tested case, i.e. specification, could always be given. The second would be close to consultancy which could payoff, but which could also lead to communicative disaster in the given environment. The third would be the most difficult to achieve and the one which would require most from the potential users. In a "user-participation" perspective, the former of these three would be the one with the least need for participation and the latter would be the one with the most need for participation. In arguing for a participatory design, Bowers (1991) states that "cooperative design should be relational, reflexive, critical and anti­idealist". Less than pushing a specific technology for collaborative work, we should investigate the nature of the collaborative efforts and later provide technology which we find feasible.

4 Just Doing it (CSCW): the Telemedicine Project

Computer supported collaboration has turned out to be a part of the Telemedicine­project6 even though it has not been a goal or activity in itself. The Telemedicine activities could be categorized into the three following topics (Danielsen 1991): - remote consultancy (accessing expert knowledge) - competence distribution and sharing - distance learning and supervision Applications have all been delivered as answers to problems, and not as mere changes for the sake of changing the environment. Problems and questions we have been looking for have all been related to communication between people. It turns out that the Telemedical applications could all be described as technology supporting collaboration.

6 Initiated by NTR in 1988. An overview is given in (ValvAg 1992).

26 T. Danielsen

4.1 Computer Support for Remote Consultancy

An important objective for the public health service in Norway is to secure equal rights to access to the services regardless of geographical or social differences. An uneven distribution of personnel resources is further strengthened by the medico­technical development moving towards centralization and division of work operations (Nymo & Engum 1990). Remote consultation of expertise would therefore be most advantageous. The teleradiology experiment (Sund et al. 1991), for instance, gave birth to the following CSCW application:

X-rays are scanned at a local (smaller) clinic, stored and copied to a workstation at a regional hospital The radiologist at the regional hospital studies the X-rays on his diagnostic workstation, whereupon he dictates the diagnosis. The diagnosis is recorded on the diagnostic workstation and copied to the workstation at the local clinic. The diagnosis received from the radiologists workstation is typed at the local workstation.

Teleradiology has now been implemented for servicing local clinics from the University Hospital in Tromsf/t (UHT), on a routine basis. The prototype application will be further developed and commercialized in the very near future.

4.2 Computer Support for Distance Learning and Supervision

During 1989 two speech therapists performed an experiment with therapy for six persons suffering from aphasia (Eidsvik 1990). The equipment for each user was a PC, a data/voice modem, telephone and a blackboard system. Through this blackboard system the therapists could see what their clients were doing on their computers 7• They used the system for transmitting new exercises to their clients, and for supervising clients when they attempted at solving the exercises. A simple groupware application - a shared-screen facility - proved very valuable. Clients could receive parts of their treatment at home instead of staying in institutions for a longer period of time. And, "in some respects it turned out to be an advantage to separate the therapist and the client ( ... ) the fact that they could not see each other forced the clients to express themselves through written or oral language" (ibid.) Computer supported collaboration proved to have positive side effects.

5 Aiming at It (CSCW): A Project Concerning Use of IT in Local Government

The next sector to be investigated by the NTR in its quest for evaluating technology and its usage is the local government. I shall now discuss just a couple of the ongoing projects in our action concerning local government.

7 Both users saw both their own and the remote users cursor. This enabled them to communicate using pointing techniques as well.

Computer Support for Collaboration 27

5.1 The Traditional Educational System and Distance Learning

Local government is facing a time where a strong growth in the resources available is a thing of the past. Instead, local government will have to provide better, more efficient and more available services to its users - people. Strengthened education would require reaching a broader "market" through the use of less resources. Distance learning could turn out to be a most required technology helping us in gathering even more knowledge8• Technology supporting distance learning could help us in establishing broader networks and stronger relationships (confer for instance (Eveland & Bikson 1988)). But the educational tasks are merely parts of the activities found in modem schools. Collaboration among tutors, students and pupils are all more common nowadays. Students often participate in process-oriented collaboration in solving problems or participating in discussing specified tasks. Tutors learn from each other through questioning experiences made by peers and by explicitly discussing specified pedagogical problems. A network of tutors might also supply a richer educational environment for students.

For some areas of education it would be a most fruitful and necessary approach to cooperate with other parts of the society. We have thus initiated a project aiming at discussing scenarios of future usage of communication- and information­technology. The project concerns some twenty graduate schools9 and six centres providing further education. The scenarios should put forth suggestions for a new technological reality which takes into account both pedagogical requirements, participation10 by the audiencell and the effects of the use of a combination of media12• Because technology may facilitate a new kind of learning experience, involvement in the active construction and use of knowledge. Groupware may span across classrooms and schools and provide a new network/or learning.

The scenarios will be based on experiences on the use of communication technology in different educational "environments". One of these is the application of distance education for nursing students. Currently some 40 students at 11 different locations are involved in a four year study in nursing. As we would like to

8 "(But) with electricity and automation, the technology of fragmented processes suddenly fused with the human dialogue and the need for over-all consideration of human unity. Men are suddenly nomadic gatherers of knowledge, nomadic as never before, informed as never before, free from fragmentary specialism as never before - but also involved in the total social process as never before; since with electricity we extend our central nervous system globally, instantly interrelating every human experience." (McLuhan, 1964) 9 These schools had approximately 1000 teachers in full-time positions, 300 teachers in part-time positions and more than 10 000 students in 1990. 10 What I hear, I might forget. What I see, I might remember. What I do, I might understand. (i.e. the more you involve me, the more I might learn) 11 "Any hot medium allows of less participation than a cool one, as a lecture makes for less participation than a seminar, and a book for less than a dialogue." (McLuhan, 1964) 1 2 "The hybrid or the meeting of two media is a moment of truth and revelation from which new form is born. For the parallel between two media holds us on the frontiers between forms that snap us out of the Narcissus-narcosis. The moment of the meeting of media is a moment of freedom and release from the ordinary trance and numbness imposed by them on our senses." (McLuhan, 1964)

28 T. Danielsen

investigate the pros and cons of different technologies, we have started with a simplest possible design: videotaped lectures, telefax transmission for additional written material and audio-conferences for discussions among groups of students and their tutors.

5.2 Political Work

Local government politicians engage in their political activities after they have finished their daily work routines. Thus, they are active after the administrative staff have left their offices and after the administration is closed. A key issue in political activity is the discussion and dialogues between "political colleagues" and discussions with the people they are to serve. Most parts of these would have to take place in the traditional setting: the face-to-face meeting or the telephone conversation. But political activity does also include studying large amounts of written material. Our project will investigate the use of information technology for providing means for communicating with fellow politicians in manners suitable for the political work, and providing access to stored information, either at the local public administration or other public institutions. In other words, we will investigate the use of groupware in local government. The application has to provide an integrated collection of services in a simplest possible manner.

6 Learning Through the Network

A new medium is never an addition to an old one, nor does it leave the old one in peace. It never ceases to oppress the older media until it finds

new shapes and positions for them. (McLuhan, 1964)

CMC has given a new dimension to the world of distance learning. With CMC available it turns out that students in distance learning may more easily establish and maintain interpersonal networks and they are given easier and more convenient "access" to their tutors and fellow students (Fjuk & Jenssen 1989, Castro 1991, Rasmussen et al 1993). It seems that CMC fills the need for a way of exchanging questions and answers, problems and suggested ways of dealing with those in an asynchronous manner. And, taking into account the situation of the distance learner who holds down a job or has a family to look after, CMC is particularly suited (Castro 1991). In this manner, CMC turns out to be a powerful addition to the pre­packaged instruction material.

CMC has been used for constructing virtual classrooms (Hsu & Hiltz 1991) and has proven to be extremely helpful in teaching collaborative skills to students. Through the virtual classroom students were grouped into teams carrying out role playing, event-oriented scenario games. A control group was given the same tasks to perform, but were not given access to the virtual classroom. Not only did the students with access to the virtual classroom gain better results in the games, but they did also establish a closer social network and "camaraderie" (ibid).

Computer Support for Collaboration 29

In another study (Harasim 1991) it was found that "delivery of education through computer communication alters the relationship of the instructor, the students, and the course content" - students engaged in collaborative learning. The same study does however point to the importance of the interface of the CMC and the integration with other computer-based tools for handling information, including the information exchanged through CMC. Integration and the structuring of information may further strengthen the position of CMC in distance education. One such system is based on the use of hyper-structures as means for communicating in a local area network (Romiszowski 1990). This experiment came about after some initial experiments with plain CMC in distance learning. The integration of CMC with access to databases has also proven as a powerful means of gathering information in the process of learning.

7 The Challenge of CSCW

The Ratatosk environment (Danielsen et al1991) is designed to give users access to a number of information and communication services. Ratatosk is further designed to provide support for any number of services of the different categories, i.e. Ratatosk is adaptable. Instead of designing another mailer, we have designed a user interface integrating a number of services through access to a broadest possible range of applications. In CSCw, integration of communication services is essential.

Furthermore, we have to deliver applications well-suited for their specific tasks. The application has to have some knowledge of the user. Any communication and cooperation support should be "hidden" underneath the application in which the user finds functionality directed towards his or her tasks. And, if we ever are to reach beyond the highly-interested, computer-knowledgeable users, we have to provide systems with extensive user support (Wasson & Akselsen 1992). In CSCW, integration with the user environment is essential.

Our users move from task to task, from one work process to another. They even move from group context to another group context all during the day. They change roles and their roles evolve. No groupware system may survive if it sees its users as a standalone group. In CSCw, integration of services providing group support is essential. Furthermore, groupware should not limit itself to embodying services which in the traditional setting is provided by computers. Media would all have to melt into one single unit - one tool - for the user in question. In CSCW, integration of networks is essential.

No standalone groupware for specific tasks will ever become a common commodity such as the telephone. Nor will this ever reach through even if the groupware is too open-ended and general. Computer support for collaboration would have to cover these specific functions:

- integration and cooperative use of communicative channels13

- total integration with work- and activity-related applications14

1 3 Some might use the phrase multimedia in this situation. 14 Instead of yet another work-oriented application, add network to existing ones.

30 T. Danielsen

adaptability towards networks and network-provided services adaptability towards communicative processes15

The use of groupware and the design of systems supporting collaboration may help us interconnect the information islands. Although groupware may be taken as having been there all the time - actively considering cooperation changes the whole situation. No user is supposed to be left on an isolated island.

The challenge of CSCW is to prove itself as the highly context- and problem­specific tool in daily work routines, and not to present itself as merely another general-purpose method for communicating.

References

Bowers, J.M., J. Churcher & T. Roberts (1988). Structuring computer-mediated communication in COSMOS. In: Research into networks and distributed applications. Schicker, P. & E. Stefferud (eds.) Amsterdam, North-Holland.

Bowers, J.M. (1991). The Janus faces of design: some critical questions for CSCW. In: Studies in computer supported cooperative work. Bowers, J.M. & S.D. Benford (eds.) Amsterdam, Elsevier Science Publ.

Castro, A.S. (1991). Using computer-mediated communication to support the australian adult distance learner. Interactive learning international, vol. 7, no. 1, pp. 67-74.

Danielsen, T. (1991). Health care through any plug or the accumulation of bits and medical knowledge (in Norwegian: HeIse i hver plugg eller opphopning av bits og medisinsk kunnskap). In: The network of telematics between daily life and society / The electronic connection from home to society (in Norwegian). Hallingby, H.S. (ed.) Kjeller, Norwegian Telecom Research. (NTR-report TF R 5/91).

Danielsen, T., W. Finnset, F. Flregstad, G. Hartvigsen & R. Steen (1991). Ratatosk - an adaptive user interface to computer-mediated communication. In: Message handling systems and application layer communication protocols. Schicker, P. & E. Stefferud (eds.) Amsterdam, North-Holland. pp. 345-354.

Danielsen, T. (1993). The challenge of computer-mediated communication in health care. Telektronikk, vol. 88, pp. 72-77.

De Cindio, F., G. De Michelis & C. Simone (1988). Computer-based tools in the language/action perspective. In: Research into networks and distributed applications. Schicker, P. & E. Stefferud (eds.) Amsterdam, North-Holland.

Eidsvik, A.K. (1990). Distance learning for people suffering from aphasia. Paper presented at the "World conference on computers in education" in Sidney, July 1990. Kjeller, Norwegian Telecom Research. (NTR-Iecture TF F 18/90).

Engebretsen, J.E. & W. Jensen (1979). Computer power - to whom? (in Norwegian: Datamakt - for hvem). Oslo, Gyldendal.

Eveland, J.D. & T.K. Bikson (1988). Work group structures and computer support: a field experiment. ACM Transaction on office information systems, vol. 6, no. 4, pp. 354-379.

Fjuk, A. & A.E. Jenssen (1989). EDB-basert kommunikasjon i fjernundervisningen (CMC in distance learning). Oslo, University of Oslo. (Thesis for the degree of cand.scient in Informatics).

15 Group processes and collaboration are all subject to changes throughout their life­time. The groupware have to provide facilities for supporting this.

Computer Support for Collaboration 31

Flores, F., M. Graves, B. Hartfield & T. Winograd (1988). Computer systems and the design of organizational interaction. ACM Transactions on Office Information Systems, vol. 6, no. 2 pp. 153-172.

Galegher, J. & RE. Kraut (1990). Technology for intellectual teamwork; perspectives on research and design. In: Intellectual teamwork. Galegher, J., R.E. Kraut & c. Egido (eds.) Hillsdale, New Jersey, Lawrence Erlbaum Associates.

Greenberg, S. (1991). An annotated bibliography of computer supported cooperative work. SIGCHI bulletin, vol. 23, no. 3, pp. 29-62.

Hahm, W. & T.K. Bikson (1989). Retirees using email and networked computers. International journal of technology and aging, vol. 2, no. 2, pp. 113-123.

Harasim, L. (1991). Designs & tools to augment collaborative learning in computerized conferencing systems. Proceedings of the 24th annual Hawaii international conference on system sciences (HICSS), 1991, pp. 379-385.

Henninger, S. (1991). Computer systems supporting cooperative work: a CSCW '90 trip report. SIGCHI bulletin, vol. 23, no. 3, pp. 25-28.

Hiltz, S.R & M. Turoff (1978). Network nation: Human communication via computer. Reading, Mass., Addison-Wesley.

Hiltz, S.R. (1988). Productivity enhancement from computer-mediated communication: a systems contingency approach. Communications of the ACM, vol. 31, no. 12, December 1988, pp. 1438-1454.

Hiltz, S.R (1989). Computer-mediated communication systems: dropouts versus users. Proceedings of the IFIP 11th World Computer Congress. Amsterdam, North-Holland, 1989, pp. 1089-1094.

Holand, U. & T. Danielsen (1991). Describing cooperation - the creation of different psychological phenomena. In: Studies in computer supported cooperative work. Bowers, J.M. & S.D. Benford (eds.) Amsterdam, Elsevier Science Publ.

Hsu, E.Y.P. & S.R. Hiltz (1991). Management gaming on a computer mediated conferencing system: a case of colloborative learning through computer conferencing. Proceedings of the 24th annual Hawaii international conference on system sciences (HICSS), 1991, pp. 367-371.

Lea, M. (1991). Rationalist assumptions in cross-media comparisons of computer­mediated communication. Behaviour & information technology, vol. 10, no. 2, pp. 153-172.

MaIm, P. (1993). Classification of cooperative systems from a technological perspective. Groupware for local government administration. Tromsp, University of Tromsp. (Thesis for the degree of cand.scient in Informatics, in preparation).

McCormick, N.B. & lW. McCormick (1992). Computer friends and foes: content of undergraduates' electronic mail. Computers in human behavior, vol. 8, pp. 379-405.

McLuhan, M. (1964). Understanding media: the extensions of man. New York, Mentor. McLuhan, M. & Q. Fiore (1967). The medium is the massage. London, Penguin Books. Motiwalla, L.F. & J.F. Nunamaker (1992). MAIL-MAN: A knowledge-based mail

assistant for managers. Journal of organizational computing, vol. 2, no. 2, pp. 131-154.

Nymo, B.J. & B. Engum (1990). Telemedicine to improve the quality, availability and effectiveness of the health service in rural regions. Paper presented at "Seminar on the regional impact of advanced telecommunication services" in Kiruna, June 1990. Kjeller, Norwegian Telecom Research. (NTR-lecture TF F 10/90).

Pankoke-Babatz, U. (ed.) (1989). Computer based group communication. The AMIGO activity model. London, Ellis Horwood.

Rasmussen, T. ,P. Helmersen, J. Bang, K. Lundby & B. Skov (1993). PROFF: Computer­mediated communication in Nordic research education: a social feasibilty study. Kjeller, Norwegian Telecom Research. (NTR-report TF R 7/93).

32 T. Danielsen

Romiszowski, AJ. (1990). Computer mediated communication and hypertext: building structure into distance seminars. In: Proceedings of Pacific telecommunications: weaving the technology and social fabric. Wedemeyer, DJ. & M.D. Lofstrom (eds.) Honolulu, Jan. 1990, pp. 540-547.

Smith, H., J. Onions & S. Benford (eds.) (1989). Distributed group communication. The AMIGO information model. London, Ellis Horwood.

Sproull, L. & S. Kiesler (1991). Two-Level Perspective on Electronic Mail in Organizations. Journal of Organizational Computing, vol. I, no. 2, April-June 1991, pp. 125-134.

Sund, T., E. Rinde & J. Stlolrmer (1991). The TMS teleradiology experiment. Paper presented at the "2nd Japan-Nordic PACS symposium" in Tampere, Junel991. Kjeller, Norwegian Telecom Research. (NTR-lecture TF F 12191).

ValvA-g, H. (1992) (ed.). Applied telemedicine. Kjeller, Norwegian Telecom Research. Wasson, B. & S. Akselsen (1992). An overview of on-line assistance: from on-line

documentation to intelligent help and training. The knowledge engineeringreview, vol. 7, no. 4, pp. 289-322.

Whittaker, S. (1991). Towards a theory of mediated'communication. Bristol, HP Laboratories, report no. HPL-91-78, June 1991.

3 Some Pedagogical Techniques for Computer-Mediated Communication

Morten Plate Paulsen

NKI, P.O. Box 111, N-1341 Bekkestua, Oslo, Norway, E-mail:morten@nki.no

Abstract. This paper presents a comprehensive set of examples showing the gamut of pedagogical techniques that are available for programme planners and designers of computer-mediated communication courses.

Keywords. Computer-mediated communication, pedagogical techniques, computer conferencing, distance education.

1 The Framework

There is an abundance of literature describing methods and techniques that can be used to facilitate adult learning. The following books, though, have been especially useful for the identification of the techniques included in this literature review: "Adult Learning Methods" (Galbraith 1990), "Helping Adults Learn" (Knox 1987), "Approaches to Training and Development" (Laird 1985), and "Effective Strategies for Teaching Adults" (Seaman and Fellenz 1989). Articles that discuss pedagogical techniques for computer-mediated communication are much more scarce. Several descriptive articles, presenting experiences from CMC courses, include some information about the pedagogical techniques, but the issue is rarely pivotal. These articles are referred to in the review of each technique. A few articles have attempted to give an overview of educational use of CMC systems (McCreary and Van Duren 1987, Harasim 1991 and 1992, Rekkedal and Paulsen 1989, Rekkedal 1990, and Kaye 1992). A closer examination of these articles reveals, though, that they are just preliminary attempts to cover the gamut of pedagogical techniques that are available in CMC systems. The techniques discussed here are organized according to four communication paradigms that are frequently used in computer-mediated communication. The four paradigms correspond to the elements of CMC-­information retrieval, electronic mail, bulletin boards, and computer conferencing. The classification is inspired by Rapaport (1991) who uses it in the title of his comprehensive book on CMC. First, the techniques classified as one-alone techniques are characterized by retrieval of information from online resources and the fact that a student can perform the learning task without communication with the teacher or other students. Second, the techniques classified as one-to-one communication can be conducted via e-mail applications. Third, the techniques discussed as one-to-many techniques will typically be conducted via bulletin boards or distribution lists for e-mail. Finally, the techniques presented under many-to­many communication can be organized within a computer conferencing system.

34 M.F. Paulsen

1.1 Online Databases

An online database is an organized collection of data that can be accessed via CMC. Utilizing these resources, a course provider could maintain local databases of relevance to the courses. An easier solution, though, could be to provide access to international databases. A growing number of such databases are now available via CMC networks. At Murdoch university, Howse (1991) states, the printed list of addresses for library catalogues that are accessible via Internet is seventy pages long. Students and faculty at Ohio State University have, for example, access to at least nine major libraries and a number of databases via Internet (Dixon 1991). CompuServe users can, as another illustration, access the Academic American Encyclopedia, Dissertation Abstracts, ERIC, Magazine Database Plus, and Peterson's College Database. A sample of an indepth review of one specific database is available for the ICDL Database for distance education at the Open University, UK (Paulsen 1991).

Table 1. Some pedagogical CMC techniques

One-alone Online databases Online journals Online applications Software libraries Online interest groups Interviews

1.2 Online Journals

One-to-one Learning contract Mentorship Apprenticeship Internship Correspondence In-basket

One-to-many Lecture Symposium Panel Skits

Many-to-many Debate Simulation Role play Case study Discussion Transcripts Brainstorming Delphi Nominal Group Observation Forum Project Group Hearing Cognet Jigsaw

Online journals are periodicals that are distributed to subscribers via CMC networks. They are increasingly important resources for information and learning. Howse (1991) states that the Association of Research Libraries compiles a directory of electronic journals, newsletters, and scholarly discussion lists. He further states that the directory to be released in June 1991 will list some thirty journals, about sixty newsletters, and over one thousand scholarly lists. In CMC programs, teachers can encourage and help students to subscribe to these journals and use them as an integrated part of a course or as a supplement to the course work.

Some Pedagogical Techniques for Computer-Mediated Communication 35

1.3 Remote Access Applications

Remote software applications are programs that can be executed on a remote computer via a computer network. They include a range of application from software development tools; via specific applications for statistics, economical analysis, etc.; to computer aided instruction applications. The following two examples present experiments from the NKI Electronic College (paulsen 1992).

As a part of the Introduction to Computer Science course, Lindland used the EKKO online multiple choice database (EKKO-base) for the first time in the fall 1989. The students could first download a number of multiple choice questions, then spend the time they needed to figure out the answers, and finally upload their suggestions for automatic scoring. EKKO-base is further described in an article by Quale (1990).

In another experiment, B!IIrsum taught a programming course that allowed the students to access the host computer's Cobol compiler. Although it was more convenient to use a local microcomputer compiler, the experiment showed that remote students can access host computer applications such as compilers, software for statistics, and tools for development of databases.

Hiltz and Turoff (1978) believe that "most work in computer-assisted instruction (CAl) has suffered from a lack of the incorporation of structured communications" ... and that it " ... will require the integration of CAl systems with (computer conferencing) systems to allow the encompassing of the total educational process."

1.4 Interviews

An interview is described by Seaman and Fellenz (1989) as: "a presentation in which an interviewer asks questions of one or possibly two resource persons before an audience. The resource persons are knowledgeable about a previously determined topic of interest to the audience and should have been informed about the kinds of questions that will be asked, especially those that will open the interview. Questions may be prepared in advance, improvised by the interviewer as the activity progresses, submitted to the interviewer on small cards by members of the audience, or a combination of the above."

In an article, Etzkowitz (1989) reviews e-mail as a communication and interviewing medium. The paper discusses an experiment in using e-mail in qualitative social research where the face-to-face focused interview is adapted for electronic use. The article presents the concept of the focused interview and its electronic adaption as well as a comparison between e-mail and face-to-face focused interviews.

An example of an online interview is presented by Paulsen (1992). In the introduction to the interview he describes how the interview was conducted via Internet:

This interview with Bruce Scriven, programme chair of the 16th ICDE World Conference, was conducted via e-mail. The first set of questions was posted in January. In February, a second round of questions was posted to clarify and elaborate on the questions and answers. Finally, after the additional information from the second announcement flier was included, the interview was

36 M.F. Paulsen

dispatched to Bruce Scriven for approval. Except for some technical problems with lost messages, this was an interesting experience using an interview technique that can be recommended.

A more concrete example of an educational application of an interview could be to give the students the following assignment; Each of the small groups should conduct an e-mail interview focusing on a topic that is relevant to the course, and post the interview to the class.

1.5 Learning Contract

A learning contract is "a formal agreement written by a learner which details what will be learned, how the learning will be accomplished, the period of time involved, and the specific evaluation criteria to be used in judging the completion of the learning. The learning contract is a technique that is used to individualize the learning process." (O'Donnell and Caffarella 1990)

Marantz and England describe their experiences at Empire State College with a learning contract carried out via CMC:

Not only was none of the value of face-to-face contract mode lost, but much more was added. Online together, we developed a learning contract, "Telecommunication in Education: exploring the future," and carried the study to completion using all capabilities of the medium -- email, BITNET communication, database file transfer, a CAUCUS conference, and PHONE "chat" -- in a way that enhanced learning for each of us. It involved close reading, intensive discussions and critical argument, a broadly ranging survey of users, and the development of three substantive papers. We found that by maximizing CMC strengths and identifying potential shortcomings, this electronic "distance" study achieved at least as much, and often more, than what face-to-face tutorials provide by way of "close" collaboration and meaningful mutual learning. (Marantz and England, 1992)

1.6 Apprenticeship

An apprentice is a learner of a trade who has agreed to work for a number of years in return for being taught. Levin, Haesun, and Riel state that "Patterns that we've observed in instructional electronic network interactions resemble those described in face-to-face apprenticeships .... Thus we may see emerging a new pattern, 'teleapprenticeships', with some ofthe properties offace-to-face apprenticeships .... " (Levin, Haesun, and Riel 1990)

An illustrative example of a teleapprenticeship in Digital's corporate network is provided by Gundry:

... that conferencing networks offer the potential for learning outside formal educational channels was brought horne to me in the case of a young man who works in our group. This man is 21, and is a specialist in VAX system management, hypermedia, and DECwindowslMotif programming. He joined us

Some Pedagogical Techniques for Computer-Mediated Communication 37

four years ago having completed a Digital-sponsored information technology awareness course in the local town, after leaving school at 15 with almost no qualifications. Virtually everything about his specialties that he has learned since he joined Digital has come from participation in conferences. He has attended a couple of formal training courses, but he has gained most of his expertise through conferencing. When he has encountered a work-related problem that he cannot solve himself, his first reaction is to consult the network, and then to search and research for the answer or for someone who can tell him the answer. (Gundry 1992)

1.7 Debate

Seaman and Fellenz (1989, 65) say that: "A debate is a structured discussion during which two sides of an issue are presented and argued by two or more individuals within a given time period." Another explanation is offered by Knox (1987): "Similar to a lecture or panel, but two or four debaters argue two sides of an important issue to clarify differences and related reasoning. "

Clark organized a debate about war protesters and freedom of speech in February 1991 as a part of the: "What's in the News Telecomputing Project" (Clark 1992a and 1992b). The debate centered on a proposition which stated the issue to be discussed. The affirmative side supported the proposition and the negative challenged the affirmative. The object of the debate was to see which of the teams could do a better job of presenting the case. Before the debate, neither team knew which position it would be assigned, so each must learn as much as it could about both points of view.

1.8 Simulation or Games

Simulation can be explained as: "Imitation of interpersonal or other dynamics, often using materials and roles, to help participants feel as well as understand the dynamics of a complex situation." (Knox 1987)

Hiltz and Turoff state:

The major defect that most games exhibit, especially educational ones, is that the communications actually used in the face-to-face game environment usually do not reflect the real world. By putting the game into a computerized communications environment, we can program the structures for communications that the game implies. This may include which players in the game can talk to whom and in what circumstances; costs or resources that must be expended for communications; leaks of communication; rumor simulation and unanticipated breakdowns or busy signals. The computer can act as the game controller, scheduling the events to occur and providing the outcomes based on the actions the role players take. One very significant aspect of this flexible degree of control is the ability to control the clock. Because of this, the game can be played in a regulated time manner (such as every week of play representing a year) or in real time. There are many games where playing in real time rather than accelerated time would be beneficial to enhancing the realism,

38 M.F. Paulsen

including some of the disaster type games designed to educate people on how to deal with crisis situations. Since people can interact at a time of their own choosing, a computer-based game can go on over days, weeks, or months, just as for a computerized conference. (Hiltz and Turoff 1978)

An excellent example is the Management Practices Course taught via EIES by Hsu (1990) at New Jersey Institute of Technology. This course integrated a Business Simulation Game with computer conferencing in a Management Practices course. The students were divided into six groups of four students. Each group represented a company, and each student was assigned a role as CEO, Financial Officer, Operations Chief, or Marketing Executive. These companies competed against each other in a Business Simulation Game, through three phases of the companies' life cycles (start-up, growth and independence). The game simulated nine years during nine weeks of the course. Each year, the students "employed" in each company established crucial input data such as: price, advertising, purchase, production, size of sales force, etc. The data was submitted to the instructor, who compiled it and executed the game. This process resulted in a set of output data for each company, consisting of units sold, back orders, market share, operating income, income tax, net income etc. The companies were evaluated based on the final results after nine years. Each company was assigned a private conference in which the employees could discuss the simulation input and output data. In another conference, called Managers' Comer, the students could participate in management­related discussions.

1.9 Role Play

Role play is "a range of methods in which trainees put themselves in dramatic situations and act out scenes like actors in a play .... There are essentially two kinds of role play: structured and spontaneous .... Structured role play is based on a case study .... Spontaneous role plays are based on momentary experiences." (Rothwell and Kazanas 1989,415)

Hiltz and Turoff (1978) regard games and role playing as "one of the most promising exploratory subjects" for computer conferencing. They further state:

Once the group has compiled item lists (such as roles, action options, consequences, and environmental factors) and the relationships among these, we can incorporate the design of a generalized game controller that will allow the generation of an event-sequenced scenario-game in (computer conferencing form. This means that the group can play out the "world" model or Gestalt that resulted from their contribution of judgements and views. Such a result also becomes a helpful vehicle for conveying to others what the group has arrived at and discovering if others agree or disagree.

Finally, Hiltz and Turoff (1978) state: "The role playing could probably be done more realistically through the computer than in some of the face-to-face acting games used, especially if the student were not able to tell which of the other players were students, faculty, or real-life jobholders playing at their convenience from their own terminals."

Some Pedagogical Techniques for Computer-Mediated Communication 39

Goodman (1992) describes how the University of Michigan has used the Confer CMC system to involve schools around the world in two different role plays. In the Arab-Israeli conflict simulation, groups of students were assigned roles as countries in the Middle East, the United States, and the U.S.S.R. Other groups took on roles as representatives for the West Bank, the PLO, the Moslem and Christian factions in Lebanon, etc. In the United States Constitutional Convention roleplay, historical figures who have played important roles in shaping the U.S. constitutional history where revived to discuss the Constitution. Groups of students played the roles of Thomas Jefferson, Martin Luther King, Jr., etc.

In the Fall of 1988, Johannesen taught an Information Systems course via the EKKO conferencing system. In a conference, she presented a case and assigned each student a role. The case described a company planning to invest in a new computer based office automation system. The students were assigned roles as user, accounting officer, project manager, labor union representative etc. Over a period of about fourteen days, the students should elucidate the different facets of this project reflected through the different roles.

Another example of a role play assignment could be that each student assumes the role of one of the authors presented in the class readings and defends the author's views. The role play could be organized as a panel discussion in small groups.

1.10 Transcript Based Assignments

Davie (1987, 14) states that "One of the main advantages of a computer conference is that the medium provides a complete transcript of the course interactions". Building on this observation, Davie and Wells (1991) suggest the following three types of transcript-based assignments to promote student reflection:

First, students might be required to retrieve all the comments they authored during the course. The assignment could then ask the students to reflect on their contributions and provide a statement of the overall framework or perspective embodied in them.

A second possibility is to ask students to pull together all the comments related to a particular topic and to write an essay discussing which comments they agree with and why or to critique the comments from the perspective of a particular theory.

A third possibility is concerned with improving the student's analytic and writing skills. Too often, students write to please the teacher. This contribution is graded and then ignored by both parties. Instead of this dead end process, students can be asked to retrieve an earlier note or assignment and rewrite the work either to make it more effective, or to reflect the current state of learning. This kind of recursive learning can help the student to build skills in a way that is simply not feasible in the face-to-face classroom. (Davie and Wells 1991)

1.11 Brainstorming

Brainstorming is "an interaction strategy used to generate ideas or to help determine the exact nature of content to be discussed. This approach encourages group

40 M.P. Paulsen

members to think creatively and to expand upon ideas of fellow group members. The primary purpose of brainstorming is to create a pool of ideas on a topic." (Seaman and Fellenz 1989)

Hiltz and Turoff (1978) state that in brainstorming, criticism is ruled out, freewheeling association is welcomed, quantity is wanted, and combination and improvement are sought. Further, they suggest that a computer conferencing system "designed to optimize brainstorming would probably limit text items to a small size, might censor items containing negative words and phrases, and utilize stored profiles on individuals to suggest group members. It might also use automated indexing techniques to group and organize items."

Hiltz and Turoff (1978) offer that brainwriting is essentially written brainstorming: "each person writes an idea down and passes it to a neighbor, who must add to it. These pieces of paper are passed around until everyone has commented on every piece of paper. With slight changes, in most (computer conferencing) systems this would mean passing a comment to each participant in t~, to make a required addition before incorporating it into the conference."

In a description of the mM internal CMC network, Rueda (1992) present a brainstorming related activity that took place over four days in September 1991 on the C-LANG FORUM. The activity was initiated by a request for an elegant method of branching over the value of a character string in the C programming language. In the following interaction, nineteen participants from thirteen locations contributed thirty-five entries and nine distinct solutions.

1.12 Delphi Technique

The Delphi Procedure is a technique for "obtaining the most reliable consensus of opinion of a group of experts '" by a series of intensive questionnaires interspersed with controlled opinion feedback" (Dolkey & Helmer, 1963). It is used to scan the environment to identify possible changes, their effects, training needs, new work methods and approaches, and issues worth exploring." (Rothwell and Kazanas 1989).

Hiltz and Turoff (1978) distinguish between forecasting delphies, where a group of people come up with a joint forecast, and policy delphies, where the objective is to develop the strongest arguments for or against particular resolutions. They further expect the following advantages to emerge from computerized delphi:

Besides the reduction of elapsed time to carry out a Delphi via the computer, the other significant impact is the ability for the process to flow steadily and incrementally. In other words, forecasting of one variable or one policy resolution can be examined ftrSt in computerized conferencing and carried through the whole process. Alternatively, different items could be in different phases of the process according to the wishes for the group. This provides a greater ability for the group to focus its effort and should result in raising the quality of the result. (Hiltz and Turoff 1978)

Waggoner (1992) describes a computer conferencing delphi that was conducted by a consortium of eight intermediate school districts in Michigan. The consortium

Some Pedagogical Techniques for Computer-Mediated Communication 41

examined the question: "What will be the impacts of high technology on the content, delivery, and organization and administration of instruction (K-Adult) conducted by local and intermediate school districts over the next five years?" (Waggoner 1992) .

The study contained three components: "a statewide teleconference, a national delphi study, and a future scenarios workshop .... The delphi study was intended to develop forecasting and planning data about a range of questions by explicating the opinion of a nationally distributed panel of experts on technology and education." (Waggoner 1992)

An elaborate, thirteen-step delphi process was developed and implemented, employing eighteen paid experts and the Confer computer conferencing system. Each expert was initially asked to identify and elaborate on the significance of the three most significant technological trends or products that will influence education. The responses were compiled into thirty-five issues and the experts then voted and commented on each of the issues. For a further clarification of some of the issues, nine of the issues were entered into a second round of votes and comments. In conclusion, the delphi process was synthesized in a set of fourteen findings.

The group found the potential to be greater than was achieved in this particular case. This was so despite the fact that the group was quite satisfied with the other participants, that the process was clearly communicated, and that they were relatively comfortable with the medium (use of computers and terminals). (Waggoner 1992)

1.13 Nominal Group Technique

Seaman and Fellenz (1989) state that the term nominal group technique "comes from its use of participants as individuals--only nominally (in name only) as a group--forthe initial stage of idea generation. This approach calls first for the silent generation and priority ranking of ideas by each group member. This is followed by a public listing of ideas usually by asking for each person's top-ranked idea and then moving on to second- and third-ranked ideas until everyone's list is exhausted .... Once this is completed, group members are allowed to discuss the ideas .... After the discussion, a vote is taken in which group members are asked to rank the ideas that have been generated."

A more thorough discussion of the nominal group technique is presented by Korhonen (1990). Hiltz and Turoff (1978) note that the first stage of the technique could be handled by computer conferencing "without the uneasiness that sometimes accompanies sitting around a table and looking at one another without talking." They further (Hiltz and Turoff 1978) imply that computer conferencing is well suited to handle anonymity and that "the introduction of anonymity is ... one of the strongest techniques to prevent conformity to group pressures."

The University of Auckland has developed a groupware system to support synchronous group sessions. Sheffield and McQueen (1990) report on the experiences from a management course exercise using the groupware. Two groups of ten students took part in an assignment using the Nominal Group Technique. One of the groups utilized the groupware, the other group used traditional tools like

42 M.F. Paulsen

wall mounted sheets of paper and felt markers. Both groups expressed satisfaction with: "the technical and the socio-emotional aspect of the discussion" (p. 181). The students using the groupware completed the assignment in less time than the other students, and as the result of the groupware process, they had developed a written assignment report.

1.14 Project Group (committee)

For a definition of project group, Seaman and Fellenz (1989) refer to Brilhart (1982): "The committee is a small group of people given an assigned task or responsibility by a larger group (parent organization) or person with authority".

This example describes the NKI Electronic College Project Assignment Course, taught via the EKKO system by this author, in the 1991 spring semester. The course-work involved collecting information, both by doing interviews and literature search. An important goal was to make the students accustomed to project management and cooperation. A second main objective was to teach students to produce a written report of the project results. The students were encouraged to find a project related both to knowledge obtained through the NKI Program and to their job. They were asked to form project-groups of one to three students.

Each student received two Norwegian textbooks through land-mail. "Goal Directed Project Management", also available in English (Andersen et alI987), is a general purpose textbook for project management. It emphasizes that project development comprises people, systems, and organizations. The other book was written as a guide to project work for on-campus students at the NKI College of Computer Science. In addition to the textbooks, the students received a fifteen page study-guide that was developed for both correspondence students and EKKO students. The guide comprised the extra guidance for distance students, course requirements, and assignments. Since the study-guide was completed a few days after the course started, it was distributed via EKKO.

Twelve students, ten men and two woman, enrolled in the course. For most of them, this was the 10th and final course in the programme, so they were all advanced computer conferencing users, with no need for user support. Most of the students enrolled in more than one course during this semester.

The course started the first week of February 1991 and the final project report was due May 10th. During this period the students had to complete four assignments. The assignments were not paced, that is, they had no due date.

The first assignment covered the theory from the textbooks. Each student had to turn in a short essay via e-mail. In addition the students were asked to present project ideas in the class-conference. In the second assignment the students had to form a project group and present their project task, milestone plan, and responsibility chart. The third assignment asked each group to write a progress report, and the fourth assignment asked for the final project report.

The teacher's job was to comment on the assignments and help the students whenever they had questions about their projects and the project management tools. Most of this feedback was routed via e-mail to each specific group, but information of general interest, was posted on the class' bulletin board, or in the class' conference.

Some Pedagogical Techniques for Computer-Mediated Communication 43

References

Andersen, E. S., K. V. Grude, T. Haug, and J. R. Turner, (1987) Goal Directed Project Management. London: Kogan Page

Bates, T. (1991) Third generation distance education: The challenge of new technology. Research in Distance Education 3(2):10-15

Brilhart, J. K. (1982) Effective Group Discussion (4th ed.). Dubuque, IA: W.C. Brown Clark, G. C. (1992a) Debate, electronic style. Instructor (February):57-58. Clark, G. C. (1992b) Project idea: Electronic debate. Telecommunications in Education

News, 3(3):14-15 Davie, L. (1987) Facilitation of adult learning through computer conferencing. In

Proceeding of the Second Guelph Symposium on Computer Conferencing, 11-22. Guelph, Ontario: University of Guelph

Davie, L. (1989) Facilitation techniques for the on-line tutor. In Mindweave: Communications, Computers, and Distance Education, eds. R. Mason and A. Kaye, 74-85. Oxford: Pergamon Press

Davie, L. E. and R. Wells. (1991) Empowering the learner through computer-mediated communication. The American Journal of Distance Education, 5(1):15-23

Dixon, R. S. (1991) Value-added network services ... and beyond. A paper presented at the Applications of Computer Conferencing to Teacher Education and Human Resource Development Conference, June 13-15, Ohio State University, Columbus, Ohio

Etzkowitz, H. (1989) The electronic focused interview: Email as a dialogic interviewing medium. In Message Handling Systems and Distributed Applications, eds. E. Stefferud, O. 1. Jacobsen, and P. Schicker, 525-537. Amsterdam: North-Holland

Galbraith, M. W. ed. (1990) Adult Learning Methods. Malabar, FL: Krieger Goodman, F. L. (1982) Instructional gaming through computer conferencing. In

Empowering Networks: Computer Conferencing in Education, ed M. D. Waggoner, 101-126. New Jersey: Educational Technology Publications

Gundry, J. (1992) Understanding collaborative learning in networked organizations. In Collaborative Learning Through Computer Conferencing: The Najaden Papers, ed A. R. Kaye, pp. 167-178. NATO ASI Series F, Vol. 90. Berlin: Springer-Verlag

Harasim, L. (1992) Foreword. In From Bulletin Boards to Electronic Universities: Distance Education, Computer-Mediated Communication, and Online Education, M. F. Paulsen, i-iii, University Park, Pennsylvania: The American Center for the Study of Distance Education

Harasim, L. (1991) Teaching by computer conferencing. In Applications of Computer Conferencing to Teacher Education and Human Resource Development, ed. A. J. Miller, 25-33. Proeedings from an International Symposium on Computer Conferencing at the Ohio State University, June 13-15

Harasim, L. M. (1990) Bibliography on educational CMC. In Online Education: Perspectives on a New Environment, ed. L. M. Harasim, 229-264. New York: Praeger

Hiltz, S. R. (1978) The Network Nation: Human Communication via Computer. London: Addison-Wesley

Hiltz, S. R. (1988) Teaching in a Virtual Classroom. Volume 2 of A Virtual Classroom on EIES: Final Evaluation Report. Newark, NJ: New Jersey Institute of Technology

Hiltz, S. R. (1990) Evaluating the virtual classroom. In Online Education: Perspectives on a New Environment, ed. L. M. Harasim, 133-184. New York: Praeger

Howse, W. J. (1991) Internet--The discoveries of a distance educator. DEOSNEWS 1(21) Hsu, E. (1990) Running management game in a computer mediated conferencing system:

A case of collaborative learning. In Proceedings of the Third Guelph Symposium on Computer Mediated Communications, 201-208. Guelph, Ontario: University of Guelph

44 M.F. Paulsen

Kaye, A. R. (1992) Learning together apart. In Collaborative Learning Through Computer Conferencing: The Najaden Papers, ed A. R. Kaye, pp. 1-24. NATO ASI Series F, Vol. 90. Berlin: Springer-Verlag

Knox, A. B. (1987) Helping Adults Learn. San FranciscolLondon: Jossey-Bass Korhonen, L. J. (1990) Nominal group technique. In Adult Learning Methods, ed. M. W.

Galbraith, 247-259. Malabar, FL: Krieger Laird, D. (1985) Approaches to Training and Development. Reading, MA: Addison­

Wesley Levin, J., K. Haesun and M. Riel. (1990) Analyzing instructional interactions on

electronic message networks. In Online Education: Perspectives on a New Environment, ed. L. Harasim, 185-213. New York: Praeger

Marantz, B. and R. England. (1992) Closing the distance: a CMC learning contract tutorial. DEOSNEWS 2(4)

Mason, R. D. (1989) An evaluation of CoSy on an Open University course. In Mindweave: Communications, Computers, and Distance Education, eds. R. Mason and A. Kaye, 115-145. Oxford: Pergamon Press

Mason, R. (1991) Moderating educational computer conferencing. DEOSNEWS 1(19) McCreary, E. K. and J. Van Duren. (1987) Educational applications of computer

conferencing. Canadian Journal of Educational Communication 16(2):107-115 Miles, M. B. and M. A. Huberman. (1984) Qualitative Data Analysis: A Sourcebook of

New Methods. Beverly Hills: Sage O'Donnell, J. M. and R. S. Caffarella. (1990) Learning contracts. In Adult Learning

Methods, ed. M. W. Galbraith, 133-160. Malabar, FL: Krieger Paulsen, M. F. (1987) In search of a virtual school. T.H.E. Journal 15(5):71-76. Paulsen, M. F. (1991) The ICDL database for distance education. The American Journal

for Distance Education 5(2):69-72 Paulsen, M. F. (1991) Bibliography on Computer Mediated Communication in Distance

Education. DEOSNEWS 1(6) Paulsen, M. F. (1992) The 16th ICDE World Conference: Distance education for the

twenty-first century. DEOSNEWS 2(7) Paulsen, M. F. (1992) From Bulletin Boards to Electronic Universities: Distance

Education, Computer-Mediated Communication, and Online Education. University Park, Pennsylvania: The American Center for the Study of Distance Education

Paulsen, M. F. (1992) The NKI Electronic College: Five years of computer conferencing in distance education. In From Bulletin Boards to Electronic Universities: Distance Education, Computer-Mediated Communication, and Online Education, M. F. Paulsen, 2-17. University Park, Pennsylvania: The American Center for the Study of Distance Education

Quale, A. (1990) EKKO-base: An online multiuser selfstudy database. In The Electronic College: Selected Articles from the EKKO Project, M. F. Paulsen and T. Rekkedal, 115-123. Oslo: NKI Forlaget

Rapaport, M. (1991) Computer Mediated Communications: Bulletin Boards, Computer Conferencing, Electronic Mail, and Information Retrieval. New York: John Wiley&Sons

Rekkedal, T. and M. F. Paulsen. (1989) Computer conferencing in distance education: status and trends. European Journal of Education 24(1):61-72

Rekkedal, T. (1990) Distance education and computer conferencing: Some international experiences. In Elektronisk Fjrernopplrering--Om Undervisning og Kompetanseutvikling gjennom datakommunikasjon og nettverk, ed. I. B~, T. Carlsen, T. Rekkedal, and M. S~by, 1-23. Oslo: Norwegian Center for Distance Education

Rothwell, W. J. and H. C. Kazanas. (1989) Strategic Human Resource Development. New Jersey: Prentice Hall

Some Pedagogical Techniques for Computer-Mediated Communication 45

Rueda, J. (1992) Collaborative learning in a large scale computer conferencing system. In Collaborative Learning Through Computer Conferencing: The Najaden Papers, ed A. R. Kaye, pp. 87-101. NATO ASI Series F, Vol. 90. Berlin: Springer-Verlag

Seaman, D. F. and R. A. Fellenz (1989) Effective Strategies for Teaching Adults. Columbus, OH: Merrill

Sheffield, J. and R. J. McQueen. (1990) Groupware and management education: Matching communication medium to task requirements. In Proceedings of The Third Guelph Symposium on Computer Mediated Communication, 181-192. Guelph, Ontario: University of Guelph

Thomas, R. (1989) The implications of electronic communication for the Open University. In Mindweave: Communications, Computers, and Distance Education, eds. R. Mason and A. Kaye, 115-145. Oxford: Pergamon Press

Waggoner, M. D. (1992) Explicating expert opinion: A case study of a computer conferencing delphi. In Empowering Networks: Computer Conferencing in Education, ed. M. D. Waggoner, 157-194. New Jersey: Educational Technology Publications

Wells, R. (1992) Computer-Mediated Communication for Distance Education: An International Review of Design, Teaching, and Institutional Issues. University Park, Pennsylvania: The American Center for the Study of Distance Education

4 Learning in Groups: Some Experiences of Online Work

David McConnell

Centre for the Study of Management Learning, The Management School, Lancaster University, Lancaster LAI 4YX, England, UK Email: d.mcconnell@lancaster.ac. uk

Abstract. Recent trends in the use of technology for mediating group learning and group work pose questions about our understanding of how groups work in these so called online environments. Our understandings of how groups work in face-to-face environments may not relate directly to how groups work in online environments. In particular, the way in which "time" is perceived and used online, and the dynamics of group work online (especially the dynamics of mixed sex work) may differ to that experienced in conventional groups.

Keywords. Learning groups, online work, time, participants' experiences, gender differences in groups, group dynamics, language.

1 Introduction

The major focus of this paper is the ways in which people in online, collaborative learning groups work together. Working in groups is a social process, as well as one having a product. We know a great deal about working in face-to-face groups, but we know very little about what is behind the experience of working in groups online.

In this paper I would like to indicate some differences in the way people work in collaborative online groups, and focus specifically on some issues to do with the effects on time and group dynamics. The paper is based on research carried out over the past few years into the perceptions and experiences of staff and students on the computer mediated MA in Management Learning at Lancaster University, and on desk research into the educational potential of CMC.

2 Online Groups

The groups I am referring to here are small ones made up of three / four / five participants and a tutor. The members are all participants on the computer mediated MA in Management Learning, offered by Lancaster University Management School Members are all professional management trainers and developers. The

Learning in Groups: Some Experiences of Online Work 47

programme is part-time and is designed around five residential workshops over a two year period, with substantial periods of "online" group work in between lasting from between three months to seven months. We use the Caucus computer conferencing system. Participants logon using a modem and communications software from their home or place of work. Staff logon from the University or from their home. A major concern on the programme is with the social aspects of learning via computer mediated communications (CMC). Our starting point is not the technology: we see that as a tool, a means to en end. Rather, we are concerned to establish and maintain a supportive learning environment where participants and staff are involved in actively constructing meaningful learning relationships that are acceptable to them all alike. We use the term "learning community" to emphasise the social and learning relationships that are encouraged online and in the workshops. The programme is concerned as much with participants' self and professional development as with their academic learning. For more details see (McConnel 1992, Hodgson 1992).

3 Group Work Online

What kinds of activities are the members of these online learning groups involved in ?

1. Individuals are carrying out their own work and are being supported by each other in achieving personal aims. This might include members working on issues to do with each others personal development, or discussing their interests in learning and the different meanings and models that can be attributed to the term, and so on.

2. People are working at developing the social aspects of the group. By the social aspects I mean all those activities and issues to do with building and developing a sense of group online. We consider it vital that we address the social contracts that are necessary for any group to work and survive successfully, just as one would in a face-to-face setting. These are what Joseph McGrath calls the "member support and group well being functions" which are so important in successful technology mediated group work, yet which are often neglected, or worse still, never considered. Like us on our computer mediated MA, he believes technology mediated group work requires:

" ... the deliberate creation of the very kind of social norms that apparently arise spontaneously in natural face-to-face groups, and that are very powerful and effective devices for regulating face-to-face communication in these groups." (McGrath 1991).

3. Problem solving - people are initiating discussion about "problems" in their work and life, and inviting others to share in that and offer feedback and ideas which will help take them forward.

48 D. McConnell

4. People are working on their current self-chosen piece of work for assessment for the MA. This involves:

personal choice in what to work on sharing this with others engaging in dialogue about the piece of work testing ideas against members' experience and knowledge offering semi-drafts of the work for comment

5. Group members are also involved in receiving each others completed piece of work and in offering feedback on it and engaging in dialogue about the substantive issues of each paper

6. Collaborative assessment of each member's piece of work - by self, peers and tutor.

There is very little consensus reaching activity in these groups; there is no exact problem or issue to be solved. But, some consensus is needed over:

a) How the group is going to work together online, e.g. how often they will use the online system each week; how they will deal with each others responses; how they will "look after" the group. Sometimes discussion of these issues is formalised at the beginning of the online period; sometimes an issue emerges and is then discussed amongst the members of a group.

b) How to develop criteria for assessment. The process for determining criteria for the assessment of each participant's current piece of work is a major aspect of the groups' agenda. Sometimes the group develops criteria which can be applied across everyone's piece of work; sometimes criteria are specific to each individual's piece of work.

c) Pacing of the group work, e.g. deadlines for deciding when to stop discussions; when to submit members assessed pieces of work to each other; when to start giving feedback; when to assess; when to close the group, and so on.

The work of the groups is characterised by the importance of the social interactions that take them forward. The climate of each group is non-competitive. The forms of collaboration in these groups require a high order of involvement, a high willingness to share, and a belief that individual development is enhanced by working with others around issues of mutual interest. The collaborative work of the groups is unusual given the formal educational environment in which we find ourselves. It is not the accepted tradition for students to collaborate in these ways on an MA programme. What happens on the programme is perhaps more akin to the sort of teamwork that is sometimes found in public and private sector organisations.

Learning in Groups: Some Experiences of Online Work 49

4 The Experience of Group Work Online

We tend to take for granted most of the experiences of working with others in face­to-face groups. Most of us have worked in face-to-face groups so often that, even though the experience is complex and at times unpredictable, we tend to approach it with a certain degree of familiarity. Working with others in online environments is so unusual that we may have to approach it as if it is a completely new experience. From our experiences at Lancaster University of working online, relying on well worn strategies and working from common assumptions about how groups work in face-to-face environments is not always the best orientation to take.

Over the past six or seven years I have been working with groups in educational settings using computer conferences. I have used this medium for seminar work, problem solving, discussions and indeed most activities that any of us involved in the educational process would normally engage in. Recently, at Lancaster University we designed the MA in Management Learning that is run through face­to-face workshops and electronic computer conferences. This programme runs alongside a "conventional" part-time MA which does not employ computer conferencing. Working on both programmes has given me the opportunity to compare the experiences, and to consider what some of the differences are in working with groups in these two environments.

I would like to focus here on two differences that I find to be important to me at the moment. These are the issues of "time", and gender differences in group dynamics. "Time" is important to me because it is a central issue in any of our professional lives. Working online poses questions about the nature of time as we have come to experience it in face-to-face meetings, and what that might mean for us working in a more open, and arguably "timeless" environment such as computer conferencing. Group dynamics, and gender differences in these, is possibly the central issue on our MA programme, governing the way we look at the learning process as it occurs online.

5 Time

In conventional face-to-face meetings there is a strong sense of when the group meets. All those involved attend at the same time. Each meeting has a time limit. Time is important and is a limiter - people need to leave for other meetings or carry out other activities at certain times during the meeting or after it. So we can talk of the meeting beginning and ending according to hours in the day. The meeting is controllable. At a different level, the work of the group is punctuated by periods of time when members work together, and periods of time when they are apart. There is a loss of relationship between the group members in the periods in between meetings, and some effort has to be put into re-building those relation­ships when members next meet. During the meetings time can be used for concen­trated effort, to focus in depth on issues and problems. Quick banter between participants, with immediate feedback and solution of problems, is possible. The sense of immediacy is perhaps one of the most important aspects of face-to-face meetings.

50 D. McConnell

Groupme~ Goupme~ Groupm~

break break

TIME

Fig.I. Face-to-face groups

In online groups, meetings are continuous; there are no breaks between one meeting and the next. The concept "to meet" is different because there are no scheduled dates and times and locations. There is no sense of leaving one location to travel to another for the meeting. Time is less important and doesn't necessarily limit the group. At least, the span of limited time is greater. In a sense, you have less control over the time given to the meeting since the meeting is "there" all the time and constantly beckons you to participate. There is little if any sense of loss of relationship between participants since online groups are constantly relating to one another. Discussions are often prolonged and participants can use the time to dig deep into issues, exploring many different avenues of thought at the same time. This can lead to diverse results and outcomes. Participants can re-shape conversations on the basis of their ongoing understandings and reflections. They can re-visit "old" conversations and re-start them. The possibility to reflect on conversations is perhaps one of the most important aspects of online meetings.

TIME----------------------------~.

Fig. 2. Online groups

Time therefore has different meanings and different consequences for learning in face-to-face and in online groups. Equal time may be allocated to each type of meeting, but the possibility for different locations, times of day and week in online groups can be advantageous. For example, on our MA programmes we may allocate six hours to a one day face-to-face meeting on the "conventional" part time programme. In the online programme, this same period of time may be scattered over several weeks. There is the same duration of time, but different possibilities for working together. The productivity value of the same period of time is not equivalent.

Learning in Groups: Some Experiences of Online Work 51

What effect does all of this have on members of online groups ? It can have positive and negative impact on group members. Participants in our part time, computer mediated MA programme experience time aspects of online group work in different ways. For some, the continuos nature of the online group work means that the programme is constantly part of their life. It never goes away. It is absorbed into their lives:

"I don't feel I've been on a part time programme because the programme is what I've been doing all the time I've been on it over the past two years. My fantasy is that the other programme (the conventional part time one) would have been something that would have featured in my diary as 'lumps', and those lumps of time would have had more of a sense of something to fit into my life rather than DOING my life ... which from time to time would have erupted into a paper which I'd have had to send to someone to read. And I wouldn't have wanted this (computer mediated programme) to be different". (participant quotation, MA programme review).l

That sense of "living" the programme comes largely from the way in which the groups work online. The medium allows constant contact with group members. Each group spends time in developing a social contract for working together to which the members are personally committed. The sense of ownership concerning the purposes, processes and outcomes of the group is high. Members view their relationships with each other in a professional way. There is little sense of a "university tutor" structuring the way in which the group should work. The tutor's role is in many ways no different to that of any other participant. He or she is a "tutor participant", bringing those skills and insights to the learning process that any professional tutor could bring, but always with an eye to being aware of the possibility that they might be using their authority inappropriately.

The various activities of each group are not discrete in the way they often are in a face-to-face group. Members are working on multiple issues at the same time. And because the issues are introduced by the participants themselves and are about important concerns and interests they have in relation to the work of the group, there is a different quality to the group members relationship with the issues than might be the case if they were introduced as "tasks" or topics by say a tutor. So, members perceptions of 'time' can be governed by their relationship to the issues they are working on, and who introduced them.

The permanent nature of the group and its work over lengthy periods of time, and the way in which members are able to "re-join" at any time, means that participants can seem to carry the work of the group with them in their everyday lives:

lQuotations in this section are taken from the MA programme reviews. As a part of the face-to-face workshops, we are involved with participants in reviewing the programme to date, especially the online work. These reviews are open-ended discussions which often lead to re-designing certain aspects of the programme, such as our online group work.

52 D. McConnell

"For me its not just the being online; the reading, the writing - but I have these constant interventions in my life. I go online; I get some information. I read it there and then or I print it out. I sit down and read it. I think about it; I think about it a bit more. I think about what I am going to say to that. I think, well I could ignore it. I think" Oh god, I've got to say something good about that because that was something pretty good she put there". It's become my whole life. I'm driving along thinking "what can I say to that ?" or "I don't know what to say to that". I'm feeling I can't say anything to that! You know, its just sort of eaten into my whole life". (participant quotation, MA programme review).

Although this participant experienced her online work as a great source of enjoyment, it did also have serious implications for the amount of time needed for her to participate in a way that she felt she wanted to in order to support others in the group and benefit from them working together in general:

"In terms of time when I started the programme, I had no idea ... I thought I understood it ... but I didn't have any idea of the amount of time this programme has actually taken out of my life. I thought I knew, but I wouldn't like to go back now and add up the amount of hours. And I'm sure the other MAML programme (the "conventional" part time one ), if I were to equate the two .... there is this huge discrepancy". (participant quotation, MA programme review).

This is not to suggest that she was unhappy with this situation, because when asked the question "Is it worth it ?", she replied " Oh yes. It has to be worth it. I chose it for the cm (computer mediated) part. It does attract me and it does suit my life style, but takes a lot of time." This accords with other research which suggests that not only is individual effort increased in online group work, but the total effort of groups is increased when they work in these environments (Pinsonneault 1990).

These extended periods of time working online produce very positive outcomes for the groups. On the whole, the interpersonal processes are a major concern of the groups and are worked at assiduously. This creates a climate in which it is enjoyable and challenging to work on the "tasks" of the group, and the conversations and outcomes are usually of a very high quality.

Joseph McGrath (1991) points to this as an important aspect of the work of groups. Groups given short, tight time periods in which to accomplish tasks work at a fast pace but with low quality, and their interaction patterns are highly task focused. Those given extended periods of time, often free of a deadline, usually work slower, produce high quality outcomes and are concerned with the interpersonal relationships of the group. But the important aspect of this is that both kinds of groups continue to work to the same patterns when the time periods are changed. When more time is given to the first group, they still work in the sarpe way. When time is restricted in the second group, they also continue to work in the same way. This suggests that setting an appropriate learning culture in an online group is important to the success and well being of the group.

Learning in Groups: Some Experiences of Online Work 53

The use of time to create a climate in which the online group looks after itself and its members, and as a consequence produces good quality outcomes, may seem like a "taken for granted". But groups working via technology can often make the mistake of thinking that the technology itself will "look after the group": after all, the groupware has been designed to do just that. Our experience is that the social processes and relationships of the groups has to be worked at by the group members, just as they would do if meeting face-to-face. This takes time, but it is time well invested.

The fluidity of time in electronic environments often also allows groups to view their work and tasks in innovative ways. Conventionally, each group on our computer mediated MA forms for a period of two to three months in order to carry out the work of that period, and then all members on the programme re-form into new groups for the next period and so on. This methodology gives a sense of pacing within each group over the period of time and helps the group structure its time in order to carry out the various tasks within the given period. Within the electronic environment, it is possible for groups to continue indefinitely while at the same time individuals can re-form into new groups. This allows participants to extend work in "old" groups for as long as they find it useful. For example, this can support the continuation of discussion around feedback and assessment of members individual pieces of work. Instead of having to bring such discussions to a close often with members feeling they have not had sufficient time to explore each others' pieces of work, such discussions continue into new periods of time. Many members value the extension of their group work in this way and like being in multiple groups; it feels like a quality experience for them. For staff though, there are drawbacks. The psychological energy needed to maintain contact with discussions around each members piece of work and engage in work in new groups at the same time can be very demanding. Discussions become too spread out, and the depth of analysis that members engage in sometimes means that tutors can't always fully remember previous discussions. But the educational benefit to group members is perceived by them to be enormous.

6 Dynamics of Group Work

The dynamics of groups - the ways in which people in groups relate and work together - is an important area of study generally, no less so in educational settings where the role of group work is seen as important in achieving many educational goals.

Collaborative learning technologies are built round the concept that learning in groups is worthwhile. It is often cited that collaborative learning environments like computer conferences are also more democratic media for educational exchanges than conventional ones (such as face-to-face meetings). If this is the case, then we could imagine that the dynamics of these electronic meetings should be different from face-to-face meetings.

At Lancaster University we are becoming increasingly interested in the dynamics of mixed sex groups in online environments, and in the experiences of males and females in these environments. One example of this is the ways in which males

54 D. McConnell

and females participate in groups online. There appear to be some differences in the ways in which males and females participate in group work. Our work is still exploratory, but our experiences are sufficient to suggest that there is a gender difference in some aspects at least.

6.1 Turn Taking

An example of this is turn taking in computer conferences. Turn taking in groups refers to the phenomenon of one person talking and then giving way to another person. It is often used as a measure for looking at the participation of male and female members of learning groups in formal educational settings. The general phenomenon is described by David Graddol:

"The dynamics of face-to-face talk allow for many inequalities, however. The turn-taking mechanism is such that successful participation in conversation is often a competitive business requiring speed and confidence: a maxim of "first in gets the floor" seems to operate; one person can interrupt another and prevent them from taking or completing a turn. The need for contributions to be topically tied to the current topic means that those who cannot or do not wish to participate in that topic have limited powers to change the topic to another. There is instead, a continuous thread in which current participants have more control than those listening and a current speaker has first rights to select the next speaker" (Graddol 1989).

It is often suggested that computer conferences allow for a more participative dynamic than face-to-face meetings. There is no need to wait for your turn to talk since the architecture of a computer conference permits multiple, simultaneous conversations. There is little sense of interrupting others, and topics need not be tied to any current topic since new items can be set up for new topics. The power relations of the members of the group, in theory, can be shared more evenly. Given all of this, do computer conferences additionally support equality of participation among males and females ? Or are the gender differences in participation and talk that are often cited for mixed sex face-to-face groups still in operation?

Computer conferences lend themselves to this kind of simple turn taking analysis since it is easy to establish how often anyone "talks", or takes a turn. In four mixed sex conferences, we found that there is a strong trend towards females taking more turns than males in conferences generally, and within specific items (or topics) in a conference. The results are shown in Table 1.

In three out of four cases, females took on average more turns than males. Put simply, females tend to be more "chatty".

Who says most? Another measure of the dynamics of male and female participation in groups

relates to who says most. One way of measuring who talks most in a computer conference is to look at the words entered by each individual. This gives a measure of the length of each entry (Table 2).

Learning in Groups: Some Experiences of Online Work 55

Table 1. Tum taking in mixed sex computer conferences

Conf. I Conf. 2 Conf.3

Average Turns

Males 72 138 39

Females 97 195 35

Table 2. Male/female talk in mixed sex computer conferences

Number of items with highest number of words by gender

Males

Females

Conf. I (22 items)

12

10

Conf.2 (11 items)

2

9

Conf. 3 (17 items)

12

5

Conf. 4

134

145

Conf.4 (33 items)

19

14

(Interpretation: e.g. in conference I, males entered most words/said most in 12 out of the 22 items.)

In three out of four conferences analysed, the trend was towards males consistently entering more words when measured across all items, than females. Put simply, males tend to say more, or talk longest.

The perceptions of some members of these conferences about gender issues add to these findings. For example, one female member said that she didn't always know HOW to respond in her group; the dialogue (of the male members) was "too much in the head" and she felt she couldn't be herself. Her preferred style was "just to babble on and chat", rather than having to intellectualise about everything. A male member said that he was conscious of what he was putting into some conferences: issues that were "wholly about thinking and not about feeling". He was aware of feelings not being expressed by himself online. The female participant's perception of herself as "just chatting" seems to me to beg the question about the educational function of "chat". She may feel that chatting is intellectually less acceptable or appropriate than the more formal, structured conversations that some male members indulged in. I don't know.

In our conferences we encourage much "chat" and informal talk. We believe that much learning takes place when people "just chat", and that there is probably some underlying logic or structure to such informal talk. Researchers in Britain have looked at the educational functions of chat and informal talk in classrooms. It is

56 D. McConnell

argued that learning takes place when students are allowed to chat freely. David Graddol explains the rationale behind this thinking:

"The rationale was twofold: first, the act of planning and uttering was regarded as psychologically beneficial to thinking and problem solving. Second, it allowed the joint negotiation of knowledge, and co-operative learning. Even talk which appears off the topic may provide a vital forum for adventitious learning" (GraddoI1989).

Freeflowing, informal talk has a learning value in itself, and one which is just as important and educationally acceptable as structured, "logical", formal talk.

These experiences of these female and male participants indicate some wider issues about the ways in which males and females work in groups, what they might be looking for in terms of how to work and what kind of language might be appropriate in carrying out that work. It seems that there are differences in the ways males and females perceive the use of language:

"I can remember so clearly the first stuff that one of the men put on. I don't understand what this guy is talking about. I went over it again and I thought I don't even understand the words that he's using." "I thought, what the hell is this man saying ... he's using language that I can't identify with." (quotation from interviews with participants).2

Another female member commented:

"Stephen online and Paul online certainly, I mean as individuals regardless of sex, they have quite different ways of saying things to Helen and Jenny."

and she continues:

".... it's perfectly OK in our set conference to talk about how one churns up inside about something, or just life in general ... If I put that on to Stephen, Paul, Jed and Alan - I'd tend to think they'd drag it into the intellectual and bring it straight back into their head; and I'm not in my head, I'm in my heart and guts." (quotation from interviews with participants).

What does this mean ?

2 These quotations are taken from a series of in-depth inteviews with participants on the computer mediated MA programme. The interviews were free-flowing, focussing largely on participants' and tutors' experiences of working in online groups. This research is reported in Harder, G. Hodgson, V., McConnell, D., and Reynolds, M. (1991 Computer Mediated Communication for Management Training and Development: A Research Report. Centre for the study of Management Learning, Lancaster University, U.K., and in (Hodgson and McConnell 1992).

Learning in Groups: Some Experiences of Online Work 57

These results appear to confinn other findings into the dynamics of mixed sex groups in face-to-face situations (Jones 1980, Haas 1979). Women are often cited as being more "chatty" than males, and in many mixed sex situations they take on average more turns than males at talking . It has been suggested that this has the function of maintaining the unity, morals and values of social groups (Jones 1980). Males are cited as talking MORE in mixed sex groups, and are more loquacious. Whether these interpretations stand for the situation of male-female talk in online environments requires more research. Certainly, the above findings suggest that females are more chatty online in the sense that they take more turns at talking; but whether what they talk about could be tenned "chat", and whether male members do not chat in the same way is not clear.

There does seem to be some evidence though that female members of the conferences we looked at see themselves as wishing to express themselves in ways that to them appear to be different to male members. There is a clear wish to engage in dialogue that is not "heady" and overly intellectual (a wish to talk from the heart and guts as one of them put it). And they view themselves as more willing to participate in disclosure issues online, and talk more about themselves as people rather than just talk about intellectual issues. All of this also accords with existing research into gender in groups: females are more likely to disclose; they talk more about people than things. Males talk more about things and business issues and are factuallknowledge based in their dialogue (they tend to debate and argue rather than discuss) (Haas 1979).

It seems to me that although we have to acknowledge some major differences between sexes in tenns of the dynamics of group work and talk, it is also too simplistic to imagine that these rather stereotypical views of males and females apply to everyone. For example, in one mixed sex online group, we found that people often read an entry and if they had no immediate response they entered nothing. This had the effect of disconfinning previous speakers' entries, and the participants found this very difficult to live with. The suggestion was then made that we all enter something, anything, to confinn a speaker's entry (e.g. we could enter " I've read this but have no comment at the moment"). The point is, this strategy - which was suggested as a way of involving all participants and encouraging everyone to feel wanted and heard in the conference - came from a male member. Yet in stereotypical tenns, we might have expected a female member to be concerned about such issues and to take the initiative to do something about them.

How do we begin to understand what is going on in computer conferences from a gendered perspective? Two of my colleagues at Lancaster are looking at the issue of gender and knowledge and they quote an author who suggests that we need to find a way of allowing males and females to use language in ways that suit them and yet still maintains their gendered positions:

"Another precondition for satisfactory dialogue is an appropriate language of exchange. Traditional women's talk centres on experience and feelings, men's on analysis and logic. The challenge is to find means of bridging these differences, without compromising either register or forcing either party to speak wholly in a foreign tongue." (Judy Marshall, as quoted in Hardy 1991).

58 D. McConnell

I can agree with the notion of finding ways of allowing people to speak: in their own register. To suggest that all males talk in a certain way, which is different to the generalised ways in which all females talk is surely rather too simplistic though. My experience suggests that we can talk in different ways at different times in online meetings, depending on the context and people involved in the conversations. We can control our talk, and when necessary match it to the circumstances we find ourselves in. We can cross the sex barrier.

In general I would agree with Judy Marshall's precondition of an appropriate language of exchange for satisfactory dialogue. This must be particularly true for our use of language in a medium such as computer conferencing where words and language are the only means of communicating. Everything else is stripped away, and all we have is language.

7 Conclusions

In this paper I have relied on field research interviews, group reviews with tutors and participants on our computer mediated MA programme, analysis of online group dynamics and my own personal reflections and experiences of working with groups in online environments, to say something about the nature of group work in educational computer conferences. Specifically, I have considered the issue of "time" as it affects online groups, and have looked, albeit rather sketchily, at some aspects of gender differences in the dynamics of these mixed sex groups.

From my experiences of working in online learning groups, what conclusions might be made? - Collaborative group work of the kind carried out in our computer mediated MA puts great demands on learners and tutors to constantly "be" online. - Many participants experience this as beneficial and see the online environment as offering quality experiences. - The medium itself is constantly there (cf. face-to-face meetings where you eventually physically leave). It "calls" for participation. The presence of the computer on your desk constantly reminds you that you should be participating in the groups' work. - The consequences of not participating are feelings of guilt; of letting colleagues down; of losing track of the groups' work. - Participation in online group work can be viewed from a gendered perspective. Males and females often participate in different ways, use different language/words, look for different forms of relating online. - Because all we have in online learning environments is wordsllanguage, we should be aware of the ways in which males and females use language, and of what lies behind their gendered use of language.

Learning in Groups: Some Experiences of Online Work 59

References

Graddol, D. (1989) Some CMC Discourse Properties and their Educational Significance. In Mason, R. and Kaye, A. Mindweave: Communication, Computers and Distance Education. Pergamon Press. Oxford

Haas, Adelaide (1979) Male and Female Spoken Language Differences: Stereotypes and Evidence. Psychological Bulletin. 86(3).616-626

Hardy, Ginny and Hodgson ,Vivien (1991) Gender and Knowledge: An Exploration. Paper presented at the Women in Management Learning Conference. Lancaster University. 1991

Hodgson, V. and McConnell, D. (1992) IT-based Open Learning: A Case Study in Management Learning. Journal of Computer Assisted Learning. 8(3). pp. 136-150

Jones, Deborah (1980) Gossip: Notes on Women's Oral Culture. Women's Studies Int. Quart. 3. 193-198

McConnell, David. (1992) Computer-Mediated Communication for Management Learning. In Kaye. A.R. (Ed.) Collaborative Learning Through Computer Conferencing: The Najaden Papers. NATO ASI Series F: Computer and Systems Sciences. Vol. 90. Springer-Verlag. Berlin

McGrath, Joseph E. (1991) Time Matters in Groups. In Galegher, J.; Kraut, R. E; and Edigo, C. (Eds.) Intellectual Teamwork: Social and Technological Foundations of Cooperative Work. Lawrence Erlbaum Associates. Hillside, New Jersey.

Pinsonneault, A. and Kraemer, K.L. (1990) The Effects of Electronic Meetings on Group Processes and Outcomes: An Assessment of the Empirical Research. European Journal of Operational Research. 46. pp 143-161

5 Learning Experiences with Collaborative Working Technologies: Some Critical Factors

Cristina Simon

Educational Technology Offi~e (GATE), Madrid Technical University Ramiro de Maeztu 7, E-2804, Madrid, Spain

Abstract. This paper summarizes the discussions and conclusions of a Working Group Meeting on "Learning Experiences" held in the context of this Advanced Research Workshop. The purpose of the group was to list and analyze a set of critical factors to be taken into account when starting and continuing with educational applications of Collaborative Working technologies. Factors are outlined here in relation to the corresponding stages of the whole process where they are to be considered, that is, analysis, design, development and implementation. Finally, some issues concerning the usability of the resulting technology-based educational system are pointed out.

Keywords. Implementation, experiences, critical factors, collaborative working, analysis, design, implementation, evaluation.

1 Introduction

Any technological application designed for creating communication and interaction between humans has to accomplish the test of working in real users' context before gaining full recognition as a proper development. There are many different factors (behavioural, social, economic) that take part in the process and can influence users' final decisions and attitudes regarding the application. This undoubtedly determines the smooth running of the experience and, consequently, the possibility of continuing the project.

Within the specific field of Distance Learning, the success of the technology depends not only on learners and tutors just using the technology, but also on showing an improvement in the teachingflearning process, either in pedagogical terms (better knowledge transmission, new instructional models, etc), or in organizational ones (increase in administrative efficiency, better access of learners to the institution).

With so many different dimensions, the task of setting up a learning experience using collaborative dialogue technologies becomes a rather complex task. Its main objective is to combine features and potentialities of a given technology with users' needs and capabilities in a specific context. Then the utilization process is monitored and recommendations for use and experimentation on a larger scale are

Learning Experiences with Collaborative Working Technologies 61

extracted; the recommendation can be transmitted to the corresponding organization for further strategical planning.

As a result of a Working group on these issues, made up of some of the contributors to this volume, a set of critical factors for the setting-up of learning experiences will be briefly described here. These are aspects that have been found in our day-to-day implementations in different educational contexts, and, from the point of view of the group, their relevance has placed them in the core of the research process.

The implementation of a learning experience can be viewed as a three-stage process, consisting of: analysis of the organizational context, design, and development. Evaluation, a central point, cannot be considered as a separate phase, but rather as a continuous feedback to the process. It deserves its own discussion, but it was not possible to hold one at the group workshop; therefore there is no separate section devoted here to this issue.

2 Analysis of the Experience

Underlying educational philosophy - Distance Learning centres have their own ideas on how to articulate learners' participation in courses. In this sense, some institutions may have more participative methodologies, or any other features (sense of ownership with the course, degree of personal involvement, etc) that can create a more favourable environment for the introduction of telematic media.

Motivation - selection of key actors - A very desirable outcome of this initial analysis stage would be to identify the key people in the institution who would be willing to take part in the experience. These persons can play a central role in motivating both the rest of the organization and, of course, learners, from the very beginning.

Communication styles - A group that has been trained to interact frequently as a natural part of the course will tend to be keen on using a telematic system. To foster a sociable, communicative style is a perceived determinant factor for improving the use of the technology

Learners' maturity and motivation - The group outlined a profile of "mature learners", who enrol in a course driven by a set of needs, and adopt a problem­solving approach for following the course. They will then try to make the most of the tools available, thus constituting an ideal target group for these kind of experiences. Initial analysis performed should aim at identifying these characteristics for the design of the training action.

3 Design of the Experience

The design of the experience has to cover, for its part, a set of factors derived from complementary actions that may affect successful final performance. Some of them are:

62 C. Sim6n

Group size - This is a key variable in designing any kind of teaching. The size of a "manageable" group depends on many factors. In any case, if the size of the group becomes too large, strategies such as "cascade-making" or division into sub-groups can be implemented with the corresponding network of tutors. Therefore, the final group arrangement has to be clear well before the experience starts, in order for resources to be planned.

Design built around the media - The design of the educational activity should be integral to the computer-mediated communication patterns and their use. If the educational approach is not specifically collaborative, the communication system is likely to be used very little and much potential educational benefit will be lost.

Clear objectives and instructions - Learners have to cope with a medium that represents a change from their usual practices. In order to avoid possible confusion, it is important to provide clear instructions, both on the rhythm and schedule of the course as a whole, and on specific activities that they have to carry out.

Complementary actions - Introductory induction sessions with the aim of allowing the student to gain "hands-on" experience with the telematic system and getting some tips on good practice (working off-line, navigating through conferences, etc.) are also regarded as essential, and should be included in the initial design. The corresponding allocation of time and resources for its setting-up should not be underestimated. A good user's guide is also essential as students will be working in isolation.

Technical support - Even when the induction sessions are effective, especially during the first connections, learners need a permanent "human" contact to answer questions and to solve problems. This contact, maybe through some sort of "hot line", has also to be included in the initial design.

Stimulating participation - How can participation be encouraged? Should learners be "forced" to contribute or should they be "pulled"? The answers to these questions are not easy to find, but to a greater extent they depend on the kind of group, contents, activities and most of the factors listed here. Probably a combination of facilitation strategies (some more directive and pressing, others more relaxed) at different times while the course is running has to be considered for each situation.

4 Development and Implementation

The development team - An interdisciplinary team made up of computer scientists, telecomms experts and educationalists would be well equipped to produce a tool that serves the users' expectations and needs. Expertise in designing the interface to the system should be seen as a major skill in the development team.

Learning Experiences with Collaborative Working Technologies 63

Desirable features of a product - An ideal telematic tool should be, in the first instance, reliable; that is, it must show a consistent working pattern throughout time and use. It would also be "open" enough to include additional extensions and functionalities (e.g., fonnula editors), in order to be able to improve by incor­porating users' suggestions and the results of continous evaluation.

Prototyping - The development cycle would benefit greatly if inputs from users are being incorporated as development progresses. Methodologies such as the "rapid prototyping" approach (see Twidale et al., in this volume), in which real users are successively observed applying the tool for performing real tasks, have been extremely useful in helping to produce a successful product very close to users' needs and learning purposes.

5 System Usability

As a summary of the considerations made above, it could be argued that the "usability" of a system, which is the key to the success of the experience, is a function of the following factors:

a. Functionality offered by the system, placing the emphasis on the User Interface, which would ideally be structured in different optional layers, in order to create "sub-worlds" of functions depending on users' previous experience and skills.

b. The type of course and activities will detennine whether a sub-group of functionalities will be more used than others. It is well known that, only a small proportion of users ever use all the functionality of word processors or spread­sheets, most users use only the ones that help them with their work. In this respect, the coordinator of the experience can playa role in "customizing" the tool for learners, by providing quick reference guides, and shortcuts and tips during the introductory sessions. This effort can result into users' feeling more comfortable with the technology so that they can then concentrate on the running of the course itself.

c. Users' motivation regarding technology is a major determinant of further use. This variable is strongly linked with the expectations they may have about the functionality of the tool and their relation to the task they will have to carry out.

d. Permanent support, both for technical issues and course-related tasks, can generate a climate of fluid communication between tutors-teachers-learners. It can also generate clear infonnation and directions for following course contents and discussing and interacting with the other participants.

64 C. Sim6n

Acknowledgements

The Working Group that gave rise to the discussion summarized here was made up of the following persons: T. Abad (Catalunya Technical Univ., Spain), G. Alexander (Open Univ., UK), G. Davies (Open Univ., UK), H. Ellerman (Open Univ., The Netherlands), D. McConnell (Lancaster Univ., UK), E. Pastor (Madrid Technical Univ., Spain), G. Sanchez (Software de Base, Spain), R. Seitz (Univ. of ErHingen-Niirnberg, Germany), C. Sim6n (Madrid Technical Univ., Spain), A. Sousa (INESC, Portugal), M. Twidale (Lancaster Univ., UK).

6 Towards Collaborative Learning at a Distance

Gary Alexander1, Paul Lefrere 1 , and Steve Matheson2

1 Open University, Milton Keynes MK7 6AA, UK 2 University of New South Wales, Australia

Abstract. This paper describes the design of an experimental course in Renewable Energy Technology which is part of a general programme of research towards an 'electronic open university', in which computer mediated communication is used to provide collaborative learning techniques without requiring face-to-face contact between learners. We will describe the learning approach, and the design of the software to support it, an 'Interactive Learning Support Environment (ILSE)'. ILSE combines support for CD-ROM based libraries with a novel computer conference system with many features designed to support collaboration.

Keywords. Collaborative learning, distance learning, CD-ROM, computer­mediated communication, computer conference, renewable energy, telepresence

1 Introduction

We are developing a prototype course which aims to advance collaborative learning techniques at a distance. This is part of a general programme of research, outlined in Section 4. The project is funded jointly by the Learning Technologies Unit of the UK Employment Department and the Open University. The principal and most direct objective of the project is to produce a course on Renewable Energy Technology for teachers, planners, environmental scientists and others who may be interested. It will be an intensive eight week course, with a carefully considered educational approach, based on an understanding of what motivates people to learn effectively. It is a distance learning course, with students working in geographical isolation, but linked to others through computer-mediated communications (CMC) for discussion and collaborative activities. Our students will be provided with a set of resources, including a CD-ROM containing a library of background papers, computer models, and audio-visual materials, plus conventional texts and video. By combining resource-based learning with CMC for support and guidance, it reduces the limitations of both.

The course is a prototype, and not a final product. In the form being developed in this project, it will be run only once, with a group of around 30 students, as an educational experiment whose purpose is as much to try out the educational strategy and computer-based support tools as it is to teach renewable energy. For

66 G. Alexander, P. Lefrere, S. Matheson

this reason, and also, because of the very limited time and resources available for such an ambitious experiment, some of the materials provided will be in the form of drafts and prototypes rather than fully polished material, and short cuts will be taken with some of the tools.

The wider objectives of the project are: • To develop the methodology of collaborative learning at a distance. Several

strategies will be used and evaluated in the various course activities. We expect the use of peer review, peer support, joint document creation, role playing, etc. to be highly motivating. In addition, we hope that by making efficient use of the resources of the student group for mutual support it will reduce the need for reliance on academic staff for support.

• To create a state-of-the-art 'Interactive Learning Support Environment (ILSE), for distance learning and group collaboration. Its innovations include an advanced user environment for the communications with 'telepresence' (features to emphasize the sense of connection with the group), and integration of the CD­ROM library materials with student's project work and information received through CMC.

• To provide 'templates' for producing other courses using this educational approach. The ILSE and the learning strategies used in the course are not specific to the learning of renewable energy technology. They will be suitable for use by other courses at the Open University, other academic institutions, and for training courses by commercial and other organisations. The pUblicity and dissemination needed to get full value from these templates are an important part of the project.

2 The Course: XTOOI Renewable Energy Technology

XTOOI will be an intensive eight-week course giving students an introduction to renewable energy technology. Although it is experimental, we were determined from the start that it should also be first rate academically. Thus we have put considerable effort into the design of the course activities and teaching strategies as well as into the supporting software.

The starting point for the course design had to be academic. We had to determine the principal academic objectives and then design activities to enable students to achieve them. Our principal objectives are that students should be able to answer three key questions: • What are renewable energy technologies? To answer this, they need a working

knowledge of the technical principles of the main technologies, which are wind, wave, hydro, biofuels, tidal, solar thermal and solar electric, geothermal. They will also need to know the principal types of device used for each, and their relative advantages and disadvantages.

• How can renewable energy technologies be integrated into a nation's energy supply system? Here, they need to know about energy storage; conservation; how

Towards Collaborative Learning at a Distance 67

the time variations of renewable technologies affect the quantity of energy available; and whether mixing different renewables helps.

• What are the non-technical considerations affecting the deployment of renewable energy technologies? To answer this, they need to learn about legal, social, planning, economic, environmental impact, and other social aspects. To that end, we have devised three supporting teaching and assessment

activities, each of which requires collaborative learning at a distance: • "Renewable energy technology in Europe". Here, students work as a group,

whose task is to construct a joint document. Each student contributes a chapter describing one renewable technology in a European country for which that technology is appropriate. This supports our first course objective, namely that students can characterise the various renewable energy technologies.

• "Integrated energy in Ecotopia". We provide an interactive model which students explore. It is concerned with how the various technologies fit together, which is our second course objective.

• "A Wind Farm for Ambridge". This is a role playing exercise in which students take on various roles in a simulated public planning enquiry into the siting of a wind farm. It supports our third and final course objective, which is concerned with the social, legal, planning and environmental implications of using these technologies. Each activity is supported by materials on the CD-ROM. For the first, we have

commissioned a library of 70 recent papers covering various aspects of renewable energy technology. The papers are available both as scanned facsimiles, including all illustrations (as in a Document Image Processing system), and as formatted, styled type, although not in the same typestyle as the original articles. There is also an audio-visual lecture giving an overview of renewable energy.

For the second activity, we have designed an interactive model of the energy supply and demand for an imaginary country called Ecotopia. Options include several renewable sources: wind power, photovoltaics, tidal power, as well as storage and several energy conservation measures. Students have to devise an energy strategy for this country. Model outputs provide graphs showing the interaction of the various energy sources and give associated costs and energy savings. Students are given a sequence of four problems to solve with the model. When they have solved each problem they send their solution to their group and compare results and interpretations.

For the third activity, the simulated public enquiry, the CD-ROM will provide a set of role briefs, plus another small library of background papers relevant to wind farms and public enquiries to approve them, and a spreadsheet model used to cost wind farms. There is also a video (not linked to the computer) on wind turbines and wind farms.

For each activity, students will alternate between periods in which they do private work using the resources provided and periods in which they participate in discussions with other members of their group using the communication facilities provided. The private work will include: finding relevant information in the

68 G. Alexander, P. Lefrere, S. Matheson

libraries, performing experiments with the model, writing up the results of these searches and experiments. The discussions will include commenting on the results of other students and responding to points made in the public enquiry.

Variants of those activities have been tested in face-to-face teaching, so we know something about their social dimensions. This will help us to assess the effective­ness of our approach to telepresence and collaborative distance learning.

Table 1. Timetable for student activities

Weeks RET in Europe Integrated energy in A wind farm for Ambridge Ecotopia

1 Research & write 1st How much renewable Announcement of intentions draft energy for what savings? and public enquiry

2 Prepare position papers

3 Effects of stora~e? 4 Exchange Effects of 5 comments conservation? Public enquiry

6 Write 2nd draft Best overall 7 combination? Judges consider evidence

8 Edit & collate whole Presentations Jud~es announce results

It is important to structure activities to pace home-based students. Accordingly, each activity has clearly defined stages, with deadlines for each stage. As shown in Table 1, at most stages students will be expected to complete some deliverable piece of work: a draft chapter, comments on other students work, results of exercises with the model, etc. The ILSE must support the completion of each stage.

3 The Educational Approach

The three course activities as described each use a different collaborative strategy, respectively: joint document creation, group use of an interactive model, and a role playing simulation. Each of these is very general and could be used in courses in many different subject areas. They thus provide a starting point for a repertoire of techniques for course designers wishing to use a collaborative approach.

We have identified three principal strengths to a learning strategy which takes collaboration as its starting point (Alexander 1992): • The learners have the benefits of other perspectives on the material they ate

learning. The group inherently brings with it a wider range of experience than does any individual member. This enables learners to obtain help from their group, and to tackle larger projects than they could individually.

Towards Collaborative Learning at a Distance 69

• The learning experience can be structured so that students will find themselves presenting and explaining parts of the material to other students. Discussion with peers can be more relaxed and free than with a teacher. By communicating what they have learned to others, the material will become more integrated into their general understanding (As in the old adage, the best way to learn something is to teach it!).

• Working in a group is highly motivating to [most] people. It provides a pace for its members. People want to be seen to be doing their best. The support and sense of identity provided by the group allays fears and builds confidence.

4 Overview of the ILSE Design

Our design for an ILSE for a third generation Open University has been influenced both by studies elsewhere of desirable features of communication tools and by research studies at the OU, in our group and others, concerning the current use and non-use of CMC by UK OU students.

(Derycke 1992) lists a number of policy-relevant findings from the literature, particularly a) the need to go beyond the textual orientation of CMC (e.g. the need to enable students to exchange graphs and diagrams, or even multimedia material), b) the need to help users to manage information (the navigation issue) and c) the need to make it easy for users to form groups or to join or leave a group (the group structuring issue). Those findings are consistent with our observations of UK OU students, a significant proportion of whom are put off current CMC systems for those reasons and, also, because d) they find it hard to budget for CMC access; and e) they feel isolated by the asynchronous nature of most interactions. (Some students try to overcome the latter problem by keeping on-line for hours, and engaging in a quasi-"talk" mode of working, even though this is expensive.)

In our prototype, we have taken account of OU students' revealed wants, plus work on group formation in CMC, as discussed by (Kaye 1992); Waern's concept of "communication knowledge" (Waen 1992); theories abo.ut group decision making and synchronisation, such as (Johnson-Lenz and Johnson-Lenz 1991), (Chen et al 1991) and (Clarke and Smyth 1993); and Rogers' core-conditions for communication that permits creative co-operation, as set out by (Zimmer 1992).

We have made an attempt to address those needs in our prototype ILSE. The metaphor for the user environment is a suite of three "rooms", each with a

different function, but presenting a common user interface: • A multimedia "library" on CD-ROM, containing an introduction to the course

and its software, plus documents, simulations and lectures. These can be browsed, searched and copied.

• A "study", with tools for creating and editing structured documents, including a hypertext-based outlinerlbrowser, a text editor and a graphics editor.

• A "meeting room", linked transparently via communications, where groups can have discussions, comment on documents and coordinate their activities.

70 G. Alexander, P. Lefrere, S. Matheson

Qi c: co

Q..

(5 o ~

Menu Bars

Outline Pane

Sastt

Document Pane

Sill

Fig. 1. The consistent display presented to users

:;:'

A detailed description of the ILSE design is given by (Matheson 1992). The structure of the common interface is shown in Fig. 1. It has a main contents area, divided into two panes, an Outline Pane and a Document Pane. There is also a file exchange area, the 'sill' to allow students to copy material across from one room to another. Thus, they can browse through the library, taking extracts to use in their own work; or exchange work with other students; or comment on their work. The 'sash' which separates the Outline Pane and the Document Pane is moveable to vary the proportion of the screen devoted to either pane.

Thus the outline pane displays the structure of both the content and of the context of each document.

The Outline Pane is a window on a tree-structured description of the material within the room. Documents are contained within folders, which themselves may be contained in other folders, and so on. The Outline Pane thus acts like an organiser/filing system for the room, allowing material to be categorised and allowing fast browsing and searching through the contents at a high level.. Figure 2 shows a sample of the outline pane for the library. The triangular symbols are controls which expand or contract the tree below them, either displaying or hiding its contents.

The outline pane enables documents in the Library and Study to be displayed in the document pane. These documents in general will themselves have structure, and this structure is also shown in the outline pane. It thus permits browsing and

Towards Collaborative Learning at a Distance 71

... ~ Library

... ~ Introduction to XTOOI

• Overview of the course and activities

~ Using the course software (general)

~ Using the Study

~ Using the Library

~ Using the Meeting Room

... ~ Activity 1 - Renewable Energy in Europe

~ About this activity

~ Overview of Renewable Energy: Mike Flood

... ~ The Mike Flood Reference Library

... ~ Energy Policy

~ Energy Policy in the Greenhouse

(etc)

Fig. 2. The outline display

searching at the level of the document section. Documents are displayed in styled and formatted text, with graphics included where appropriate.

In the study, the outline pane can be used to create and edit the structure of the student's own document, just like the outline facility in better word processors.

In the Meeting Room, as will be explained in more detail below, the 'documents' are conversational threads, made up of linked messages. Other than this difference in interpretation of the contents (and different icons to mark structure in the outline pane) the overall appearance and operation of the controls in the Meeting Room is the same as in the other two rooms.

5 The ILSE Meeting Room

In the Meeting Room, students exchange messages with other students and course staff. The higher levels of the structure displayed in the outline pane (analogous to that of folders in the other two rooms) consists of groups. Groups may in turn include sub-groups, and so on.

At any time, any user can form a new group with a membership of their choice. This is one of several ways in which the Meeting Room is designed to bring people to the foreground, to provide 'telepresence'. Our intention is to create a sense that the individual is in active contact with friends, members of a support group. Thus their comings and goings are clearly visible and are not something the user has to request.

72 G. Alexander, P. Lefrere, S. Matheson

Messages are always sent to a group and are structured into threads. A thread may consist of anything from a single message from one person to a rich conversation amongst a group, with comments on an original message, further comments on the comments, and so on. Whatever this structure, it is displayed in the outline pane (it is analogous to the internal structure of a document in the other two rooms). With this approach there is no need to distinguish between electronic mail and computer conferencing. Mail is simply a message sent to a small group.

The list of members of each group is displayed in the document pane at the top of each thread in the form of a scrolling list, as shown in Fig. 3. Each person has an electronic face and a nickname. The face need not be a reasonable likeness. It is meant as an iconic representation to promote quick recognition and may be a caricature or even something completely fanciful. The borders indicate the status of the user as follows: 'active' (heavy line) means (roughly) that the user has contributed a recent message to this thread, 'in touch' (single line) means that the user has been in contact recently but hasn't contributed (a 'lurker' in CMC jargon), and 'inactive' (greyed line) means that they haven't been in contact recently.

i MarkyC Glennis Gary II .JudyR Madeline Nicky

Fig. 3. The member list as displayed at the top of a thread

The interface is designed not just to put people in contact with each other, but specifically to support collaborative activities and to aid coming to agreement. Various structures are designed to support this, including some special types of thread and special types of message.

The most general and informal type of thread is the 'conversation'. It starts with an initial message which may come to have various other messages as comments on it. The most general comments are referred to as 'annotation' messages. They appear in the document pane following the initial message, in order of arrival, all presented as a single scrolling document. There is no limit to their length, and they may in tum have comments on them.

There are also some special message types which are presented in the document pane as small pop-up windows, instead of in a scrolling list. They are indicated by special icons which appear in the margin of an annotation message. These icons are buttons which are used to display or hide the pop-up message. The types of pop-up message are 'post-it', 'agree' and 'disagree' messages. A post-it message is used for short, immediate comments and permits space for a few lines of text only. 'Agree' and 'disagree' messages are used for informal polling or voting. The

Towards Collabomtive Learning at a Distance 73

number of people agreeing/disagreeing is shown next to the icon. The pop-up window lists the names. (This is not a secret vote!) The marginal icons are shown in Fig. 4.

(JohnD ]

23 ch = annotation,

= agree,

I JohnD ~

16~ Fig. 4. The marginal icons used to indicate message types

= post-it,

= disagree.

A second, more formal type of thread is the 'commented document'. It is designed as direct support for collaborative learning. Its principal use is for circulating documents created in the study for reference and/or for comment by the rest of the group. If a structured document from the study or the library is inserted in the initial message of this type of thread, its structure is displayed in the upper pane. Comments then appear linked to the relevant sections to which they apply. The comments may be any of the message types described above. Only a single level of comments is allowed.

The third type of thread is called the 'virtual circle' following the usage of the 10hnson-Lenz's (Johnson-Lenz 1991). It has a very simple structure, and is the most formal. It is intended to be similar in function to the procedure in face-to-face meetings where everyone present is asked to give their views, often replying sequentially round a table. A virtual circle thread contains an initial message (called a 'virtual circle request') which contains a reply-by date. A single reply is accepted from all recipients. That reply must be a message of type annotation. This type of thread is used for several purposes, often to do with synchronising the activities of members of the group. Virtual circles with agreed deadlines are used for reports On the completion of various stages of the XTOOI course activities. One final type of structure is provided to promote "communication that permits creative co-operation" (Zimmer 1992). We call it an 'empathy template', and suggest that people use it to check for understanding, especially when they are about to disagree with someone else. If an empathy template is requested from the menu, all that happens is that the following appears in the editor where the user is creating a new message (Fig. 5):

Your view:

My view:

Fig. 5. An 'empathy template'

74 G. Alexander, P. Lefrere, S. Matheson

This is a reminder to the sender to indicate to the receiver that their intention has been understood so that the discussion is not at cross purposes.

6 The Communication System of ILSE

ILSE's Meeting Room, as just described, can be considered as a 'front end' or 'off­line reader' to a conventional conference or mail system. It stores messages received, maintains their linking structure and provides facilities for creating new messages. However it doesn't include functionality for operating modems or actually transmitting and receiving messages, and it doesn't include a central host. For those functions we are using a separate system, UUCP mail, using a SUN Sparcstation at the Open University as the host.

This approach has greatly reduced the effort we needed to put into the communication system, since the underlying system, UUCP mail, is a well­established standard. Also, it becomes very simple to allow students to participate in the course from anywhere in the world, so long as they have access to the Internet, as do most academic institutions. In fact, we have had requests from interested academic groups in Finland, Israel and Australia to include students of theirs in the XTOO 1 trial.

When a user gives the ILSE command 'communicate', any messages which have been created and not yet sent to the host are prepared for the UUCP program. A header is created in UUCP format which includes the list of recipients, then the message is put into a file in an external directory. Then ILSE activates a separate UUCP program which is based on a public domain Free Software Foundation application. That program handles the connection, uploads the messages from the directory, downloads any waiting messages into another directory, and then disconnects.

7 ILSE in Action

At the time of writing (February 1993) the design of XTOOI and ILSE and much of its implementation is complete, but much still remains to be done. As the course has not yet been presented, we are unable to present the results of student experience or evaluation. Instead, to help readers to understand the ways in which the ILSE Meeting Room is designed to support the collaborative learning strategies of the XTOOI activities, we have concocted a short fictional episode, showing how a student might use ILSE:

Our hypothetical student, Glennis, works in the computer industry in Wales, and has a strong personal interest in environmental issues and renewable energy technology. It is now the fourth week of XTOOI. A few days ago, she completed her draft chapter on Solar Energy in Spain for activity 1, and sent it to her group for comment. Today she has just finished writing the results of the second set of experiments on the Ecotopian model, on the effects of energy conservation. Her 2

Towards Collaborative Learning at a Distance 75

page essay includes a couple of graphs of the model output, illustrating the points she is making.

She is now ready to use the Meeting Room. She uses the outline pane to select her group, and selects 'new commented document thread' from the menu. She then pastes the essay she has just finished into the document window. Its first line, "Conservation in Ecotopia" appears as the title of a new entry in the outline pane.

~m1 EOU Meeting Room ., ml Glennis, Ga:r;yA

01

@> mJ XT001 Renewable Energy Course Group ~ MarkvC 6/6/93: Chat

~~conrad 26/5/93: Ambridge Residents

®~BObW 1/6/93:'Ecotopia - Conservation"

~ Glennis 6/9/93: Solar Energy in Spain

~~ Ga:r;yA 1/6/93: Error in your calculations? ~1~Madeline 1/6/93: I suggest you re-word the

I 6' GarvA Madeline MarkvC Nickv

Fig.6. ILSE display for Glennis with new messages marked

o

I I

I o IO~

She now selects, "Communicate" from the menu. A window appears superimposed over ILSE with messages indicating the progress of the automatic connection: dialling the telephone, login to 'eagle' at the Open University. Her essay is uploaded, and a number of messages are downloaded. The connection is then closed and the superimposed window disappears. Another window appears, announcing that "ILSE is linking the new messages to their appropriate places in the tree structure." After a delay, this too disappears, and she can resume work.

She selects "show new messages" from the menu, and circles appear around certain entries in the outline pane. She could use the outliner to find these and open

76 G. Alexander, P. Lefrere, S. Matheson

them, but instead she chooses to use the "show next" button, which she knows will systematically display all the entries with circles around them.

The first message appears in a conversation thread with the title "chat". She decides not to bother reading it now, and presses "show next" again. This time the thread displayed is the "Solar Energy in Spain" commented document thread she created last week. There are three post-it messages which appear attached to different sections. Two of them pop-up to point out typographical errors and one of suggests a re-phrasing of a sentence. There are also two annotations. These are displayed at the end of her essay. One comments on one of her calculations and the other makes suggestions about the way she has structured the essay.

The next few messages are part of the virtual circle thread marking the end of the position paper stage for the 'local residents' group of the wind farm enquiry. Several people indicate that they are satisfied with the paper the group has produced and make suggestions as to how their groups procedures could have been improved.

The final message is a 'virtual circle request', a message setting up a virtual circle thread. It is from Bob, who has taken on the role of coordinator for the Ecotopian activity. It reminds the group that the conservation experiment is now half complete, and that they should begin to think about their overall assessment in it for this virtual circle, due at the end of next week.

8 Conclusions

We have reported on the design of a course on Renewable Energy Technology intended to develop the methodology of collaborative learning at a distance. Its three activities are based on three forms of collaboration: joint document creation, group use of an interactive model, and a role playing simulation, all of which would be as suitable for other courses as for Renewable Energy. ILSE, the computer environment we have designed to support the course integrates the use of a CD-ROM library with document creation facilities and a communication system. ILSE is designed to make it easy to move materials between all of these areas. Moreover, the meeting room has a set of facilities to support collaboration, including structures for joint authoring, simple voting, and checking for mutual understanding.

It has never been our intention that the approach and designs of this project should be proprietary and restricted to use by the Open University. Rather, we mean for them to be freely available for the international academic community, where we hope they will stimulate much discussion, and help to advance the state of collaborative learning and computer communication.

Towards Collaborative Learning at a Distance 77

References

Alexander, G. (1992) Designing human interfaces to promote collaborative learning. In Collaborative Learning Through Computer Conferencing: The Najaden Papers (A.R. Kaye, ed.), pp. 201-210. NATO ASI Series F, Vol. 90, Springer-Verlag

Chen, R., Lynch, K. J., RimIer, A. K. & Goodman, S. E. (1992) International Journal of Man-Machine Studies, 36, pp.419-445

Clarke, A. A. & Smyth, M. G. G. (1993) International Journal of Man-Machine Studies, 38, pp.3-22

Derycke, A. (1992) Toward a hypermedium for collaborative learning? In Collaborative Learning Through Computer Conferencing: The Najaden Papers (A.R. Kaye, ed.), pp. 211-223. NATO ASI Series F, Vol. 90, Springer-Verlag

Johnson-Lenz, P. & Johnson-Lenz, T. (1991) Post-mechanistic groupware primitives: rhythms, boundaries and containers. International Journal of Man-Machine Studies, 34, pp.395-4l7

Kaye, A.R. (1992) Computer networking for development of distance education courses. In Computer-supported Collaborative Writing. (M. Sharples, ed.), Springer Verlag

Matheson, S. (1992) ILSE Design, Version 3. Mimeo. Department of Electronics, The Open University, Milton Keynes, UK

Waern, Y. (1992) Communication knowledge for knowledge communication. International Journal of Man-Machine Studies, 37, pp.215-239

Zimmer, R. (1992) The collaborative-learning interface: team-dynamics, team-assessment and software implications. Mimeo. Institute of Educational Technology, The Open University, Milton Keynes, UK

7 An Experimental Network-Mediated Study Support System

Henk Ellerman, Ad Schellekens, and Willibrord Huisman

The Netherlands Open University, OTIC Valkenburgerweg 167, Postbus 29600, 6401 Heerlen, The Netherlands

Abstract. In this paper a sketch is presented of the ideas that have guided the development of a network tool to be used in the practice of distance education. The long-term goal of building a prototype that fits the current ways of organizing the educational processes in distance teaching, and that can satisfactorily be used by students is defended in the first sections. In particular it is pointed out that such a prototype should be introduced in the practice of the educational process as soon as possible so as to acquire relevant knowledge about the functions telematics could fulfill in distance education. As such it may also pave the way for increased use of more advanced technology in the years to come.

In the following sections the didactic functions the prototype should subserve are described in more detail. A description of the prototype as such is not presented here. Based on earlier findings and advice to be found in the literature, the prototype will offer an integrated study environment and the network is not functionally presented as an add-on to the course.

This environment consists of the printed course materials, a computer with standard and tailormade software, a modem and a normal telephone connection. The network realizes asynchronous contact between students and tutor via a central computer that ensures the availability of all necessary information in the computers of the participants. The tutor's environment is similar to the student's, but it is extended with data analysis and management software. Tutor and students also use the telephone to discuss information that has been sent through the network.

The network-mediated support in the StudieNet prototype is focussed on monitoring the student's study progress, practising by means of tutor-corrected assignments, and practising by co-operation in small study groups.

We end with a description of the objectives on which the first evaluation of the prototype will be focussed.

Keywords. Computer-mediated communication, CMC, didactic functions, study guidance, linear algebra, evolutionary development, monitoring, telephone, assignments, formulae editor, study environment, learning environment

An Experimental Network-Mediated Study Support System 79

1 Introduction

Computer Mediated Communication (CMC) brings together the capabilities of computers and telecommunication networks. It increases the capacities for information processing, and can support interaction between, say, students and tutors over long distances. For those purposes CMC systems may include electronic mail, computer conferencing, computer bulletin boards, facsimile, teletex and videotex, voice messaging and desktop videoconferencing. It is clear that CMC addresses two of the key issues in higher distance

education: information transfer and interaction between students and tutors. Although important for education in general they are key in higher distance education because of the obvious logistic problems associated with the distance aspect. Therefore it isn't surprising that CMC has aroused considerable interest among those involved in developing and teaching distance courses.

CMC systems may remove or relax some of the temporal, physical, and social constraints on communication, while allowing for the presentation of large amounts of information in different modalities (text, audio, images, video), in many different ways. In fact, most workers in the field, when reflecting on the functions CMC systems can offer, would probably agree with Kaye (1989) that " ... CMC will ultimately emerge as a new educational paradigm, taking its place alongside both face-to-face and distance education". Although there is general agreement that CMC has large potentials for distance education, and also that the potentials can be realized, there is no detailed view about how CMC should be used. Consequently, the future role of CMC in higher distance education is hard to predict in any detail. We are still lacking sufficiently detailed didactical models to guide the realization of its potentials. Moreover, many factors influence the way CMC will function when used in practice, while there are no adequate organizational models of the role CMC can fulfil in distance education. In our project, StudieNet, an attempt is made to gather such information within the context of distance education as it is given now. The project is not attempting yet, to develop new course models that incorporate CMC-related functions (see Alexander, this volume), nor is it an attempt to use CMC as the only vehicle for education. It is an attempt to make the potentials of CMC useful within the current context of the Open unversity in the way it is presently organized. This paper presents the overall R&D strategy underlying the work that is bound

to be undertaken in StudieNet. The use of CMC in the practice of distance education has turned out to be a difficult issue, and no small number of projects have failed to live up to their expectations simply because a careful analysis of the context in which CMC should function has not been performed.

The emphasis on use in practice requires considerations about the financial aspects of CMC in distance education. If CMC will lead to a drastic increase in costs the role envisaged for the prototype will not be fulfilled.

80 H. Ellerman, A. Schellekens, W. Huisman

1.1 Costs of CMC in Higher Distance Education

The benefits derived from the implementation of new computerized information systems in organizations, whether they are commercial organizations or educational institutions, are commonly divided into the two categories of cost displacement and value added. Items falling in the category of cost displacement are financial benefits derived from the reduction, or perhaps abolition of the comparable traditional methods of providing similar services. Under value added are subsumed the extra facilities or services offered by the newly introduced technology. A number of these added values have been given above.

Exactly what are the financial consequences of CMC is, to say the least, not clear. In general it has to be remarked that there is at present no correct method of costing an educational project involving advanced technology (Rumble, 1989). Given that no clear view is as yet available about the actual use of CMC, tagging uses with explicit costs would be an exercise in vain. The problem of costing pre-developed mass distance education, on the other hand, is rather well understood. In most modem institutions for higher distance education the traditional labor-intensive approach to education is replaced, and in part abolished, by costs for the production of learning materials. If in traditional education, the cost of education depends on the number of students involved, this relation is far less pronounced in higher distance education.

CMC, being a vehicle for communication, may change the cost structure. If CMC is used mainly in the phase of learning, costs will become more and·more dependent on the number of students involved. Clearly, simply combining the methods of mass production of media and learning materials in its current form with CMC to increase interaction between students and tutors inevitably leads to increased costs. However this is an unlikely situation in any rational organizational model of distance education. First, the opportunity for increased interaction between tutors and students, may relax the demands on the production of courses. Of relevance here is the upsurge of interest in the flexible production and presentation of learning materials. Secondly, CMC is only one aspect of the new technology that will be used in education. The use of this technology in the exploitation phase may be expected to lead to a decrease of the costs involved.

In terms of value added, the extra costs of CMC are even harder to pinpoint. As a rule a value added, either in terms of study efficiency or study results, will increase costs. A more precise cost-benefit analysis can only be made when CMC systems add to the value of education. About this, nothing in particular can be said now.

1.2 Strategic Issues

As stated, the problem is to formulate a strategy to obtain knowledge about the adequate use of CMC in higher distance education.

One strategy could be to formulate a set of general didactical principles to guide the further development of CMC, ignoring the bounds that current models

An Experimental Network-Mediated Study Support System 81

of teaching place on the use of CMC. Without being bound to existing ways of organizing the educational process, and without being committed to obtaining information in that process, an educational exploration of CMC is possible that ignores many mundane problems like the cost of the prototype for the student, and the current availability of technology. Too frequently such alternatives actually presuppose the availability of the newest technology like satellites and high-speed datal inks (ISDN) to provide truly interactive distance courses (Pelton, 1990; Nipper, 1989). Perhaps the main advantage of such an approach is that a full focus on the learning processes becomes possible. In laboratory settings new courses can be developed, and research is possible into questions concerning the optimality of the media-mix, the usability of the equipment, optimal design of group-work on a distance, finding optimal didactic scenarios, and so on. These questions are important, no doubt, and may lead to significant improvements with respect to learning efficiency. Work done in this general context expands rapidly (EADTU, 1991; JANUS, 1991).

Another strategy would be to start with an analysis of the current ways of implementing distance teaching, find weak spots there that could possibly be overcome or mitigated by CMC-use, and test whether CMC can be actually used profitably. The focus then is on problems related to the gradual injection of technology in the educational process. It has often been noted that the acceptance of technology is a difficult process,

in varying contexts (Rowntree, 1988). By confronting these problems from the start one may hope that the problems

can be handled. The general idea is that by injecting a tool in the practice, it becomes a participant in the educational process. On the basis of observation and incremental change, CMC should find its way in that practice. Once it is an accepted participant in that proces, it may enforce certain changes in that process. Without disqualifying the first approach, the project StudieNet follows the

latter. It explicitly attempts to make CMC an actor in the practice of education. This inevitably implies that no full use of the promises of CMC will be made from the start, but has the distinct advantage that CMC can be evaluated in an 'ecologically valid' context. - In the following sections we describe in more detail some of the external factors that will influence the work to be undertaken in the project StudieNet. These include a description of the 'model of distance education' used at the Ou, a short description of earlier CMC projects in which the Ou participated, as well as of other projects undertaken elsewhere.

2 Distance Education in Terms of the Industrial Model

Distance education, in global terms follows the 'industrial model' of distance education (Peters, 1983). Education following this model is characterized by a dominant usage of 'one-way' media like print, broadcasting, video, and audio­cassettes. Two-way communication does occur, of course, but is mostly provided

82 H. Ellennan, A. Schellekens, W. Huisman

by occasional face-to-face meetings, telephone conversations and correspon­dence.

The model has met with considerable success over the years, not at least for the quality of the courses that have been produced by many institutions for distance education. Bates (1991) for instance notes that perhaps the most significant contribution of this model to distance education is that it has led to the development of a "method that leads to highly coherent, comprehensive, accurate, integrated and highly effective learning materials especially designed for independent study". The model does have its weaknesses, however. Especially, the lack of

interaction between learners, learning material and tutors it incorporates has often been noted in this respect (Bates, 1991). The project StudieNet can be seen as an attempt to address certain weaknesses of the industrial model directly. The development of the prototype to subserve that purpose is such that in close

co-operation with tutors of a course, a delineation of the didactical functions it can support will be given first, and after building the prototype, it will be put to test in co-operation with the very same tutors. The evaluation of the prototype should lead to an assessment of how well the didactical functions are supported by the prototype when seen as a component of the overall educational process, which at least resembles the 'industrial model' alluded to above. An assessment of the degree in which the primary didactic functions are affected by the overall educational model may in fact form the specific contribution of StudieNet to the theory and practice of educational CMC. Because experience with CMC should not stop at the end of this project, the

prototype should easily evolve into a system that can be used on a regular basis. Then it will become possible to assess the system more fully and obtain insight into the ramifications of the use of that system for the practice of education.

Having described the context in which the prototype should function it would seem appropriate to start with a definition of the functionality of the prototype. We will however start with a reflection on earlier work that has had a similar purpose. Both at the Dutch Ou and elsewhere experience has been obtained on how CMC can be used. These experiences have in part shaped the functionalities we think the prototype should have. The description it will be taken up in Sect. 4.

3 Related Work and Findings

In this section we summarize very briefly some related work. After reviewing two earlier projects in which the Dutch Ou has participated, an attempt is made to sketch the main findings reported in the literature. It should be noted at the outset that the literature on educational CMC is very scattered. It would be no mean feat to organize that literature. Especially with respect to empirical studies it is very difficult to find original source material. Consequently, no claim is made that the sketch accurately reflects all that is known.

An Experimental Network-Mediated Study Support System 83

3.1 Work in Which the Dutch Ou Participated

This section shortly describes earlier attempts to use CMC in which the Dutch Ou participated. The main lessons to be learned are summarized at the end. The projects are relevant for StudieNet because in both an attempt was made to build a system to supplement or supplant traditional ways of teaching. They were meant for use in practice.

Tele-Educatie Project

In 1986 the Ou joined forces with industrial partners to start a project with the goal to develop a system to be placed at the homes that would function as an educational terminal linked to a large-scale network for tele-education. Sixty students received such a terminal (50 MSX computers, 10 IBM XT's) and

could study an Ou course. Through the network they were connected to a tutor and to their fellow students. Beside the network there was also a computer­assisted learning (CAL) program which formed an integral part of the course. The experiment was subjected to an external evaluation (Bouwhuijs & Wijnen,

1987). It was found that, in terms of grade-points, the test students performed about as well as the students who took the course in a traditional manner, and that subjects thought positively about the possibility of CMC in distance education. With respect to the technical side, it was found that the students were able to

operate their equipment satisfactorily. Major problems with the user-interface did not occur. Most students did complain however about the speed of the data transfers. It did happen that for the transmission of necessary files four hours were needed. This was a source of irritation. Bouwhuijs & Wijnen (1987) however remarked that the results were negatively

influenced by the fact that there was little or no explicit attention to the educational and organizational aspects of the use of CMC. The potential uses of CMC have for that reason not been explored to the extent that might have been possible.

TeleCOO

In 1990, a project was started by the Hogeschool Utrecht to explore the potentials of CMC for distance education on a course in economical statistics. The Ou was asked to join this project while it was already running. In this project students were given terminals at home (PC-XT's) with which students could follow a CAL course and, through a modem, contact the tutor or their fellow students. The goal was to replace the normal practicals on economical statistics. A fuller description of the project can be found in Bakker (1991). In an evaluation of this experiment Schellekens (1992) found that the students

made only minimal use of the communication facilities. Communication proved to be time-consuming because of the need to formulate messages precisely. It was a reason for many students to refrain from an active participation in communication via their personal computer. Students did spend considerably

84 H. Ellerman, A. Schellekens, W. Huisman

more time on their statistics courses than would otherwise have been the case. Schellekens (1992) concluded that in this project there should have been more attention to the educational and organizational aspects of the experiment. It was also found that the students themselves thought highly about the possibility to use CMC, but insisted also on a combination of CMC with a more traditional approach.

3.2 General Remarks

Both projects are similar in the sense that the use of CMC supplemented existing courses. In the case of the TeleCOO project practicals were replaced by working at home with the PC, in both projects CMC was a value added to a normal (CAL) program.

What both projects have shown quite clearly is that the use of CMC does not automatically imply its usefulness. In both cases there was less than sufficient attention for the educational objectives CMC should serve, and for the organizational and practical aspects that could motivate students to start and/or continue working with their equipment.

A simple consequence of all this is that it is not yet clear what some of the possible 'values added' of CMC could have been.

3.3 Other Findings

Mason (1989) presents eight general findings one may expect to obtain when using CMC in distance education. The list seems to be an accurate reflection of the research findings. These findings have been corroborated in what is probably the strongest attempt to use CMC for a similar purpose as StudieNet sets out to do. In this experiment large numbers of students of the British Open University could follow a course on information technology, of which CMC was an integral part, were given the COSY system (Alexander & Lincoln (1989); Mason, 1989), and were given the means to contact tutors and students with the system. CMC was also used here as a supplement to an existing course.

A list of the corroborated expectations presented by Mason is now given.

• CMC proves to be 'handier' than physically meeting other persons. Tutors and students will be easier to reach.

• Because communication is mediated by a computer, the social inequality between people tends to disappear. Moreover, people with sub-optimal social skills will contact others more easily, and more effectively.

• Students will show themselves to be enthusiastic about the possibilities of computer mediated communication. This should enhance the motivation to study.

• Because many distance learners have to study in isolation, CMC can increase communication and offer a social context in which to study.

An Experimental Network-Mediated Study Support System 85

• Because most students lack a conceptual mo~el, or metaphor, for conferencing, they will find it difficult to get used to this format of communication. It may take several weeks to get accustomed to it. Sending personal notes is easier, because the metaphor of traditional mail helps in this process.

• Technical difficulties with the equipment are to be expected. Any experiment should have provisions to help students in case of calamities.

• In a conference it is important how the moderator fulfills his or her role. Sadly enough, no clear guidelines for moderating behavior exist, but it is clear that a moderator should play an active role, even to the degree that people are urged to participate. If that is not the case, the motivation to participate diminishes.

3.4 Main Lessons

What appears to be clear is that for CMC to be useful in distance education it should serve well-described didactical functions. In particular, it is not sufficient simply to offer facilities. Moreover, the students should receive reinforcement to start and continue the use of the system, especially in the first phases. The way most often chosen to reach this aim is by making CMC a subject of the course itself, as was done by Mason (1989), but also by Veasey D'Souza (1991). Under these conditions a successful use of CMC is possible.

Another lesson to be learned is not to underestimate the conceptual difficulties that need to be overcome to use CMC. Especially with respect to computer­mediated communication, the students should be provided with metaphors, or something similar, in order to understand what is going on.

We hope to meet the expectations stated above by Mason and the problems encountered in the two earlier projects in which the Dutch Ou took part, by presenting the student with a maximally integrated study environment rather than a computer and a network added to a course. This environment integrates the study activities, where the integration scheme is set by the didactics of the course. Some of these activities involve network communication, others involve pre-programmed guidance by programs on the student computer, and many involve general-purpose tools for text processing, formula editing, and calculations. The presentation of an integrated study environment should give the student a clear perspective (a metaphor if one prefers) on the functionality offered.

However the main study activity, reading and handling the textual course materials, cannot be performed on the student's computer, given the- current ergonomic state of computers. Nevertheless the amount of integration (in combination with good user-interfacing and error-free performance) should bring the student to accept network-based study support as a normal and useful part of the study.

But other guidelines are necessary for a more positive specification of the prototype. Following the general plan outline above, we start with the choice of a content domain, and analyze what role CMC can have there. In this paper the

86 H. Ellerman, A. Schellekens, W. Huisman

analyses will not be presented in large detail. Those can be found in StudieNet (1992).

3.5 The Content Domain

It was decided to focus StudieNet on a single course in mathematics (Linear Algebra). This choice was based on a mixture of practical considerations. These are spelled out in more detail in StudieNet (1992). A major consideration was that subject matter is such that a relatively heavy study support is needed by most students.

This course has two characteristics that distinguish it from many other Ou­courses. First, the students will only master the subject if they practise it by making assignments; traditionally these have the form of homework assignments of which the results are corrected by the tutor. Secondly, the learning materials cannot be presented on the computer screen in simple textmode; at least a sensible way of editing of mathematical formulae and graphs is necessary, but a more appropriate tool for handling this subject area is an integrated mathematical program for editing and calculation.

The course consists of units of approximately 4 to 5 hours study time. The total study time for the course is around 100 hours. Each unit contains practice materials and each unit ends with one or more homework assigmnents of which the results are not in the printed materials.

In accordance with the overall plan for StudieNet, tutors were asked to indicate support problems that might be tackled by using CMC in their course. As important problems related to study support were mentioned:

• The homework assignments are appreciated by most students; however the tutors do not have means to react properly to the individual problems the student have when working out the assignements. The delay between sending the working-out and receiving the correction is long (at least several days).

• Students do not learn to tackle problems systematically. • Some students do not use support facilities (hesitate to contact the tutor) and/or

do not finish the course for unknown reasons; the tutors have too little insight in the student's progress and in the factors that influence it.

• Students are confused by errors in printed materials. • One of the course units is too difficult for many students.

It is obvious that not all of these problems can be overcome by the use of CMC. That one of the course units was too difficult is an obvious example in this respect. For StudieNet the assignments will form the basic thread that will bring students and tutor in contact with each other at intervals very regularly spread over the total study time. In this way, the study as well as the network-mediated support is logically connected to the course units, and CMC is given a clear and important didactical function.

An Experimental Network-Mediated Study Support System 87

4 A Description of the Prototype

Having stated in general terms what the prototype should be able to do, we now focus on a description of the prototype itself. The prototype is described in large detail in StudieNet (1992).

Here we focus on the more salient characteristics.

4.1 The Study Environment

An important notion is that network mediation almost automatically implies the computerization of the student's desk. Seen from the standpoint of the student, the new study environment will be more prominent than the network connection itself. The network communication should be well embedded in the course, so the network should be in the student's study environment.

4.2 The Student's Computer

The student's computer is a PC based on an Intel 386 processor connected to a modem. The computer will be equipped with a set of programs (all build around a kernel adapted from IBM's Lotus Notes) for making homework assignments and for the transmission of data to a central computer. The set of programs is currently not sharply delineated. The programs have to be chosen such that the following functions can be performed.

- Editing of text - Editing of formulae and graphs - Mathematical calculation - Network communication - Monitoring of study progress - Presenting content- and context-specific information - Presenting echoes (network conferences)

These programs will be integrated on the level of the user-interface (a graphic windowing enviromnent, possibly DOS with Windows 3.0 or OS/2 2 .. x with Presentation Manager) but, more importantly, also on the level of the tasks that the student is to perform with them. This is obvious for the first three of the list above. The integration of the lower four will involve the development of a special program, which will probably be presented to the student as the 'monitor'. The monitor combines a number of service and support functions as if it were a

combination of technician, secretary and study assistant. It automatically performs all network communication with the central computer; in fact the student has no need to bother about what information has to be down- or uploaded. The monitor will support the student to map, plan and monitor the pace and progress of the study. Because of that, it 'knows' in what course unit the student is currently working and it makes sure that the most recent version of the appropriate information is downloaded from the central computer. It also

88 H. Ellerman, A. Schellekens, W. Huisman

brings the information to the attention of the student only where and when it is relevant: information about text errors at the beginning of the studying of a unit, hints about the homework assignment only after the student has made a fIrst try, etcetera. It will send messages or assignment results to the central computer for further routing to the tutor or to fellow students and it will negotiate with the central computer about an appointment for a telephone conversation with the tutor. It will also upload information about the student's study progress for further evaluation by the tutor.

4.3 The Central Computer

The central computer is hardly visible to the student. It contains a multi-user server system that allows for fully automatic data exchange with the student's computers and the tutor's computer. It contains and structures all available data. Also it runs a program with which the student's computer can negotiate about an appointment for a telephone conversation with the tutor (see below).

4.4 The Tutor's Environment

The tutor principally has the same task as he has in the traditional situation. However with the network the tutor has many more means to follow the students' study and to offer help. A special support plan will give guidelines on when and how actions are expected from the tutor. The tutor may also use the study information for evaluation of the educational

process. In StudieNet, this will not only mean the quality and use of the course, but also the more didactical aspects of the network facility. The tutor's computer contains the same environment as the student, but with a number of facilities added, especially for management and analysis of data concerning the student's work. The tutor's computer is connected to the central computer in such a way that the tutor always has a telephone line free for conversation with students.

4.5 Communication

The network facility makes contact between students possible. It is generally believed that contact between students has a positive effect on their motivation and even on their study results. However with respect to network communication there are two diffIculties: network contact is different from face-to-face contact, so it is not sure whether it will work in the same way; and even if so, how will the students contact each other via the network? Presently it seems that where network contact is something new for the student, it will function only if

- it is introduced well enough - it functions well enough, and - the students expect to benefIt enough from it.

An Experimental Network-Mediated Study Support System 89

In StudieNet, group activities are planned both for the sake of the student (possibly better learning results) and for the sake of the project (gain experience in network-mediated inter-student communication). In the field test, the students will have one introductory (face-to-face) meeting; in this meeting the students will get to know each other to some extent, and the group activities and network communication will be introduced by the tutor.

4.6 The Telephone

The student's telephone is used for two purposes: compressed, automatized datacommunication with the central computer and traditional conversation. In the non-network courses, telephone contact with the tutor is a normal provision, that is however not systematically used by the students; many do not contact the tutor at all, often for unknown reasons.

In StudieNet, telephone conversation is seen as an important form of contact between student and tutor. The network facility contains special means to organize, regulate and improve this contact. The expectation is that

- students who need personal support will more easily contact the tutor, - students who do not need personal support will less easily contact the tutor, - telephone contact will be more efficient. It is unknown whether this will lead

to more telephone contact.

In StudieNet, it is always the student who makes the telephone contact. This involves making an appointment for a call; the monitor program on the student's computer negotiates with the central computer about an appointment. The central computer controls the appointment agenda of the tutor and makes sure that the tutor is informed when and about what the student is going to call. In many cases the conversation will concern work that has prior been sent through the network and that is therefore visible at both the student's and the tutor's computer. Indeed in many cases such a conversation may be the final result of a procedure like:

- the student sends a message to the tutor, - the tutor sends an answering message containing the suggestion to call, - the student makes the appointment.

Instead of 'messages' one may read here also 'the working-out of an assignment' and its 'correction'. If the monitor program itself performs simple tutoring tasks (pre-programmed guidance), one of its possible pieces of advice may be to contact the tutor, for example because the situation is too difficult to be handled automatically. If in rare cases, the tutor will not call the students, but the telephone contact seems necessary, the tutor will send the student a message with an invitation to make a call about the matter.

90 H. Ellennan, A. Schellekens, W. Huisman

4.7 Printed Materials

Since 'Linear Algebra' is currently in use without any network, the printed materials for the test group will have to be adapted. It is possible that a number of in-text study support could be taken out and replaced by much more flexible network-mediated support given by the tutor. However such a complete revision is not practicable in the context of StudieNet; the main part of the printed materials will remain in their current state. The tutor (as well as, of course) the student will have the freedom and the means to work at least as freely with it as originally intended.

4.8 Network-Mediated Study Support

It is expected that the introduction of CMC will lead to

- an intensified, partly automatized, study support, - a more effective study support, - more complex tutor activities, and therefore - a more efficient use of the tutor's capabilities.

These expected effects are interrelated and all have to do with two network characteristics: first, there is more exchange of information between tutor and students, and second, the more straightforward support activities can be automated by software on the student's computer and data-management on the central computer. Again, these two characteristics are related. Automation of support becomes possible because it doesn't have to cover all needs; the presence of the network-mediated tutor guarantees more correct support in case the automated support fails. (An example of this is worked out below, in the section on 'monitoring progress').

4.9 Tutor Support

The student obtains study support from the tutor in the form of homework assignment corrections, answers to problems and questions, moderation of the group assignment, and various sorts of additional information about the study. In StudieNet, the tutor also follows the student's progress and may help the student if problems occur. The student and tutor communicate either over the network or over the network in combination with telephone contact.

4.10 Monitoring Progress

Monitoring progress is a basic function of the network-mediated support in StudieNet. It has a technical ground: in order to have the appropriate data at the appropriate time available at the student's computer, the 'monitor' has to know what part of the course is being studied. However it is also based on a number of educational considerations such as

An Experimental Network-Mediated Study Support System 91

- the student obtains an overview of the progress in relation to the planning, - the student learns to plan and keep track of progress, - the tutor gains insight in the student's progress, - the tutor may help if progress is insufficient in relation to the student's

objectives, - the tutor may draw conclusions about the course.

In StudieNet, the student computer contains the means to 'map' progress. In fact, various ways to study the course are possible other than the order of the printed materials. The map supports these other ways by indicating possible links between study units. For each unit, information is stored like time spent; results of the assignment, and remarks made by the student; possibly the student will have to answer a short questionnaire. Also, information is available about the author's idea of the progress that can be made, as well as information about where other students are currently working. The tutor obtains continuously updates of the progress maps of the individual students. Students can use this information to discuss study problems with the tutor. Also,students who have expressed a need for more structural guidance by the tutor can be contacted by the tutor about their progress. This progress map is to be seen as a first step towards a more elaborate construction.

There, the initiative is given more to a guidance program that forms part of the monitor, in combination to the progress map. This program asks the student questions about the study progress as well as about the student's study aim, planning, and opinions. Based on this and other available information, it produces a rough study advice. This advice has a number of purposes. Firstly, it should motivate the student to answer the questions; secondly, together with the questions it should encourage the student to make explicit decisions about how to study; third it may ease a possible tutor's intervention, both for the student and for the tutor. The latter purpose is accomplished by the possibility that the advice consists of a suggestion to contact the tutor.

4.11 Tutor-Corrected Assignments

After finishing the study of a learning unit, the student starts working on an assignment. Before the final solution is sent to the tutor, the student can obtain two hints, one in the form of a suggestion to handle the problem, the other the outcome of the problem. Once the solution has been sent away, the monitor gives the student access to the results of other students and the comments of the tutor on those. The tutor returns the standard solution of the assignment as well as a comment on the student's solution. In addition, the latter may contain a suggestion to contact the tutor by telephone.

4.12 Group Assignments

In the network-mediated version 'Linear Algebra', group assignments are used to develop the student's ability to tackle a problem strategically. The idea is that

92 H. Ellerman, A. Schellekens, W. Huisman

students will develop this ability by discussing the way in which a (complicated) problem is to be solved. For this purpose, a number of assignments are being made in which the subject maner has to be applied to a more or less real-life problem, for which more than one solution is possible.

Currently it is unclear in what way the co-operation of students in groups will be realized. Probably a conference system will be used. It is also not decided whether the tutor will just follow or take part in the process.

4.13 Problems, Questions, Messages, Errata

The student can bring problems and questions to the attention of the tutor primarily in the form of a message (which can have any format that th~ text/formulae/graph editor allows for). In many cases these problems will be related to one of the three support activities mentioned above, but they can also stand alone. The shortest form is just making an appointment for a telephone call (or, if the network facility fails, a direct call). Generally however, a message with a proper description of the problem will be more efficient. The tutor can answer the problem again in the form of a message (possibly with the suggestion for a telephone call). The tutor can make these files of problems plus answers (elaborated to a

suitable degree) accessible to other students having similar problems, or to all students. Similarly, the tutor may prepare and update files containing repairs of errata and other shortcomings of the course materials. The monitor program in the student's computer makes sure that these files come to the right students at the right moment. With the same facility, students can make, send and receive messages to and

from each other.

4.14 Additional Services

Having introduced a network, there are a number of services that can more easily be realized with it than without it. Some of these are course-related, such as the sending of old exams on request, or messages concerning organizational aspects of the course (study center meetings, if any; exam dates). Others concern the Open university study as a whole, such as the information that is currently distributed in the student's newspaper. It is possible that the students are even provided with a free-to-use bulletin board system and an open conference system.

5 The Evaluation of the Prototype

Part of the strategy underlying StudieNet is the development of a prototype that can be used as a tool to gather information about how is it used in practice. The prototype is in a sense an evaluation tool by itself. For that reason it is not opportune to evaluate the first prototype in the light of the general objectives

An Experimental Network-Mediated Study Support System 93

that guide the development. Some of these goals in fact require many years of work. The fIrst evaluation of the prototype will take place on a middle ground. The evaluation will be directed at objectives that are less far ranging than the ones guiding the development, but are general enough to influence further development of the prototype. Clearly, prerequisites for the assessment of the educational objectives are

- that the prototype works properly (bug-free). - that the students and tutors can work satisfactorily with the proton'pe.

When the prototype has a badly designed user interface, it will be impossible to judge the prototype in educational terms only. When it fails to reach one of the objectives mentioned earlier, it is not clear whether it is due to an ineffIcient operation, of the didactic functions, or whether it is due to that user-interface.

With respect to the evaluation of the user-interface the following points should be noted.

- The student has to make assignments using the personal computer. This involves the entering of mathematical formulae. As has become clear in, for instance, the TeleCOO project, this is a major problem. As part of the development cycle the many possible ways io enter mathematical equations will be systematically explored. This work should prove to be of wider interest, and is not even restricted to CMC.

- With respect to conferencing it has been shown that students in general need to be provided with good interfaces that show the overall structure of the information. The interface should make the structure and function of the facilities easy to use. Worked-out examples of such interfaces have been designed already (StudieNet, 1992) waiting to be tested.

- The prototype will be developed within a modem windowing environment. These environments have been developed for maximal usability, and are based on a rich set of experiences. These guidelines have been laid down in great detail, and will be followed as much as possible. If possible, existing software products will be used for certain sub-functions of the system. Conformation to these guidelines will play an important role. A preliminary sketch of the user­interface has already been given (StudieNet, 1992).

The intermediate goals set for the evaluation of the prototype are the following:

- An increased effIcacy of support provisions in general, combined with an increased effIciency of the support activities of the tutor.

- At least equal exam results. - Better study results in the sense that students tackle mathematical problems

more strategically. - A positive judgement of the students concerning the facilities for contact and

communication as have been realized in StudieNet.

94 H. Ellerman, A. Schellekens, W. Huisman_

- A detailed insight in the study time spent by the students in relation to the factors that determine this study time and the opinions of the students on this subject.

In StudieNet (1992) the evaluation methodology is presented in more detail. The methodology include a global design of a field experiment in which approximately 20 students will use the prototype while taken the course on linear algebra. Proposals for the measurement of the intermediate evaluation goals are presented there.

6 Project Goals

To direct the project work, a number of explicit goals have been formulated. These should be understood in the longer perspective of introducing CMC at the Dutch Open University; within the limits of the project StudieNet, these goals will at best be partially reached. Thus, the project aims at its completion:

- To have realized a prototype of network-mediated study support that has functioned technically and didactically sufficiently well to allow for conclusions on the possibilities of network-mediation in the education by the Ou.

- To have gained experience in designing, realizing and using network-mediated education; this experience is sufficient to make a realistic proposal for a regular provision in the Ou.

- To have gathered data on the didactical functioning of the network-mediated support; these data can help in deciding on the educational desirability of a regular network facility for the Ou.

- To have gathered data on the technical functioning and the usage of the network facility; these data are useful for an estimation of the cost of a large­scale network provision.

- To have gained insight in the extent to which introduction of network­mediated support leads to a revision of a course design and of its usage in practice.

- To have gained experience in implementing network mediation of study support, which may serve as a basis for network mediation of other didactical functions.

- To have developed ideas about the possibilities of using an educational network facility also for organizational purposes.

7 Epilogue

The project StudieNet is aimed at building a prototype that can be used on a large scale. Intermediate evaluation criteria have been formulated that can guide to further development of the prototype in a tool that can be used on a regular

An Experimental Network-Mediated Study Support System 95

basis. We have hinted at the methods to be used to assess those criteria. These methods involve the running of a full scale experiment. It is estimated that the phase of development of the prototype will take about two years. This includes the development of the program as well as running the first experiment. After this period a version will be available that can be a model for a viable component in the overall educational process. In that role the general outlook of the system will very likely change. The change will be based on experiences obtained in practice. The process of obtaining further experience with the system should be monitored and, where needed, supplemented by more formalized procedures of information gathering. Together with the gradual introduction of networks, two other developments may become increasingly important in the near future. The first is the introduction of a more or less standardized computer system. This allows for an integrated use of computer-based materials throughout a course. The (distant) presence of the tutor 'in' the same computer makes computer-based materials more attractive. The second development is the rapid improvement of the ergonomic qualities of hardware and software. On the basis of the close co­operation with educational practioners it is hoped that the injection of advanced technology into the process of education may become a gradual process, and may ease the burden of the tutors to adapt rapidly to new educational tools. Seen in this manner, it is hoped that we, more or less as a planned side-effect, pave the way for the introduction of more advanced technology based on high-speed datalinks, perhaps satellites, and who knows what may come into existence in the next couple of years.

References

Alexander, G. Lincoln, C. (1989) The thought box: A computer-based communication system for distance learning. In: Mason, R., Kaye, A.R. (Eds.) (1989) Mindweaves. Oxford: Pergamon Press.

Bakker, J.P. (1991) Het TeleCOO project. In: Veen, W., Bakker, J.P. Baak, P. (Eds.), Educatief telematica-gebruik in het Nederlands onderwijs. Amsterdam: Swets & Zeitlinger.

Bates, A.W. (1991) Third generation distance education: the challenge of new technology. Research in distance education, April, 10-15.

Bouwhuijs, Wijnen (1987) Ondenvijskundige evaluatie Tele-Educatie project Open Universiteit. Eindrapport SVO-project 6104.

EADTU (1991) EADTU Working Group Media, Methods and Technology. Report on Janus. Paper held at WG-MMT, Amsterdam, 25 February, 1991.

Hiltz, S.R. (1990) Evaluating the virtual classroom. In: Harasim, L. (Ed.), Online Education. Perspectives on a new environment. New York: Praeger.

JANUS ( 1991 ) JANUS Feasibility Study (DELTA project 7081) Final report. Prepared by: British Open University, Marconi, Jutland OU, Dida*EI.

Kaye, A.R. (1989) Computer-Mediated Communication and Distance Education. In: Mason, R., Kaye, A.R. (Eds), Mindweaves. Oxford: Pergamon Press.

96 H. Ellerman, A. Schellekens, W. Huisman

Koper, E.RJ. (1990) De invloed van leertaken op leerresultaten: theoretische uitgangspunten voor knowledge acquisition support systems (KASS) OTIC Research report, 17

Mason, R. (1989) An evaluation of Cosy on an Open University Course. In: Mason, R., Kaye, A.R. (Eds), Mindweaves. Oxford: Pergamon Press.

Nipper, S. (1989) Third Generation Distance Learning and computer-conferencing. In: Mason, R., Kaye, A.R. (Eds.) (1989) Mindweaves. Oxford: Pergamon Press.

Pelton, J.N. (1991) Technology and Education: friend or foe? Research in Distance Education, April, 2-9.

Peters, O. (1983) Distance teaching and industrial production: a comparative interpretation in outline. In: Sew art. D., Keegan, D., Holmberg, B. (Eds.), Distance Education Inrernational Perspectives. London: Croom Helm.

Rowntree, D. (1988) Educational Technology in Curriculum Development. London: Paul Chapman.

Rumble, G. (1989) On-Line costs: Interactivity at a price. In: Mason, R., Kaye, A.R. (Eds), Mindweaves. Oxford: Pergamon Press.

Schellekens, A.M.C. (1992) Het TeleCOO experiment. OTIC Research report, 50. StudieNet (1992) StudieNet. Verslag fase 1: Vooronderzoek en Ontwerp Interne notitie

OTICICOP. Turoff, M. (1990) Foreword. In: Harasim, L. (Ed.), Online Education. Perspectives on a

new environment. New York: Praeger. Valcke, M.M.A., Daal, M.M., Martens, R.L., Dochy, F.RJ.C. (1992) Opsporen en

beoordelen van ingebouwde begeleidingscomponenten. Handleiding. OTIC Research Report, 42.

Veasey D'Souxa, P. (1991) The use of electronic mail as an instructional aid: an exploratory study. Journal of Computer-Based Instruction, 18,3.

Verreck, W.A. (1990) Leren met inforrnatiesystemen. OTIC Research report, 12. Wolfe, R. (1990) Hypertextual perspectives on educational computer conferencing. In:

Harasim, L. (Ed.), Online Education. Perspectives on a new environment. New York: Praeger.

8 Innovative Support Technologies for Tele- and Team-Work at Universities

Ralph Seitz and Freimut Bodendorf

Department of Information Systems, University of Erlangen-Niimberg, Lange Gasse 20, D-90403 Nllmberg, Germany

Abstract. This paper outlines the main aspects of the pilot project "Portable PCs for Students" started in December 1991. The intention is to explore possibilities for decentralized and cooperative learning. Until now the main effort has been to establish a suitable learning environment for computer supported tele­and team-work. The infrastructure includes teachware provided with monitoring software agents, online libraries and several information and communication systems, as for example computer conferencing with GDS-facilities. Challenges in developing such systems and introducing new kinds of pedagogical techniques in education will be reviewed. Finally we present some preliminary results from an empirical study in progress.

Keywords. Cooperative learning, collaborative learning, distance learning, teachware, agents, libraries, electronic mail, bulletin board, computer conferencing, Group Decision Support Systems (GDSS), co-authoring, learning environment, empirical studies.

1 Introduction

Successful management of service-oriented enterprises will increasingly depend on flexibility and adaptability of organizing work. Autonomous teams which decide upon their self-organization and self-management according to the problem at hand seem to be especially appropriate to meet the new requirements. It is the universities' task to prepare students as soon as possible to cope with modem techniques for decentralized and cooperative work.

2 Research Project "Portable pes for Students"

The pilot project "Portable PCs for Students" started in December 1991 at the University of Erlangen-Niimberg. Its intention is to explore possibilities for decentralized and cooperative learning. The ftrst effort was to establish a suitable infrastructure for computer supported tele- and team-work that allows students to decentralize parts of their studies. In addition to adequate teachware and libraries, information and communication systems have been developed and implemented.

98 R. Seitz, F. Bodendorf

During this process the concept of education has been changed and enriched with components in order to support new kinds of learning and working at universities. A target group, including students of business science, social science, infonnation science and computer science, was provided with portable personal computers, communication equipment and software resources. By means of this technology students are able to connect to the university's LAN. Thereby they have access to services they can use to get infonnation for their studies or to participate in decentralized courses. The project, running for three years, is promoted by the Bavarian Ministry of Science in cooperation with Siemens-Nixdorf Infonnation Systems Inc. (Seitz 1992a).

Our major strategic issues are to find out - which organization of learning and education is useful and feasible employing innovative infonnation systems, - to what extent students accept new fonns of learning and working, - which functions should be at the students' disposal, - how the attitude and behaviour of students change, both, joining the experiment and not joining but watching it, - whether new kinds of organization and technology influence the success of learning.

The project is in the context of a much wider research interest where we investigate whether experiences concerning new possibilities of tele- and team­work at universities actually lead to innovative ideas for flexible and cooperative work in enterprises.

3 Research Methodology

For the experiment a research methodology that has to meet our special needs has been developed. We divide the whole program into four phases shown in Fig. 1 and these are outlined in the next sections.

This cycle is to be seen from a scientist's point of view who has to manage a pilot project. It is not accidental that the cycle is quite similar to that of a software development project. Hence it is obvious that the wider project cycle includes numerous shorter cycles. There are cycles, for example, where we had to develop systems based on new pedagogical techniques to be used in existing courses. Nevertheless all of the following details focus on the whole project.

3.1 Analysis and Organizational Design

The first task was to analyse the faculty'S organization of education. We found various courses where new technologies like decentralized CBT (computer-based training) and CMC (computer-mediated communication) would provide a lot of benefits for both, teachers and students. Another aim of this analysis was to select students from these courses for the target group of our experiment. We tried to find proper educational concepts and pedagogical techniques that fit the contents of the courses as well as the new fonns of learning and working we want to introduce. Doing this we explored several situations where tele- and team-work would be a serious and useful enrichment of education.

Innovative Support Technologies 99

Analysis and ~

Organizational Design "'"

+ .... Application

~

Development ...

+ Implementation

and Usage

• Evaluation

Fig. 1. Research methodology

In such a phase of educational change new definitions of teachers' and students' roles became necessary. Lewin mentioned three steps that describe group dynamic processes occurring while organizations are being changed (Lewin 1947): After unfreezing people's behaviour a phase of changes takes place. Later on the new acquired behaviour will be refrozen.

According to the change of educational concepts and roles, organizational structures and processes have to be adapted to completely different learning and teaching situations. A former lecture-oriented course, for example, where students and teachers come together in one classroom, requires a different organization from a course where students work at home and stay in contact with the teacher by means of asynchronous communication systems.

The above mentioned factors build the framework we have to manage. So all activities have to agree with the educational interests. That means: the project management has the task to promote the benefits of the project and to stay in good contact with the teachers. Otherwise the teachers' and students' acceptance will be lost.

3.2 Application Development

Redesigned, extended and new organized courses where teachers and students fulfil new roles and work together in new ways are the results of the analysis and organizational design.

Based upon this new reality we have to extract corresponding models of learning (Singer 1988) (Hiltz 1988), communication and conversation (Shepherd 1992) (Winograd 1986), coordination (Malone 1992) (Dittrich 1991) and cooperation (Piepenburg 1991). This is the theoretical background for the

1 00 R. Seitz, F. Bodendorf

development of teachware, infonnation, communication and workgroup systems for team-work in education.

The applications we developed are described more detailed in Section 4.

3.3 Implementation and Usage

The next step after developing and testing the systems is to introduce them. To achieve acceptance it is necessary to explain the advantages and facilities of the new software to the teachers. The possibilities for coordinating students are of major interest. Some short examples or scenarios can make this process easier. Our experiences have been that group activities run more fruitfully if teachers "pull" the usage of systems.

But students also need help. During a workshop all relevant functionalities are explained to the students in order to be successful with their first steps. We illustrate the whole spectrum beginning with how the modem is connected to the notebook up to the download of teach ware from the fileserver as well as the communication with colleagues and teachers. This is one way of "pushing" activities because the students' interest in the new technology grows and they overcome their first inhibitions.

3.4 Evaluation

A current empirical study examines the effects of how students act and how their attitudes change while using the tele- and team-work technologies provided. Research has been done for a period of six months. Treatment and control groups are compared by going through the same instructional phases. The feedback we get brings in new ideas and demands for the organizational design as well as the interfaces and functionalities of the systems. As our evaluation cycle only lasts half a year we are able to react promptly whenever we detect undesired effects. The theoretical background of the study and first experiences will be presented in Sections 5 and 6.

4 Learning Environment

The learning environment installed allows students

- to take part in several educational courses outside the university, for example at home, to collaborate with other students by using group-teachware,

- to get infonnation in a more flexible way and make the planning of their studies more efficient and comfortable (Seitz 1992a),

- to keep contact with teachers and colleagues by using easy to handle communication systems while working in a decentralized way,

- to learn more efficiently in teams and develop the ability for team-work (Aiken 1992),

- last but not least to experience the potential uses of modem infonnation and communication systems they will apply in enterprises later on.

Innovative Support Technologies 101

Explaining our spectrum of applications we use the classification of Rapaport and Paulsen (Rapaport 1991) (Paulsen 1993). Thus we pick up the four communication paradigms to classify the systems.

4.1 One-Alone Technique

The teachware used in some practical courses was provided with monitoring software agents which observe each learner's actions. The agents generate an encrypted session protocol that will eventually be sent to the teachers for grading (Seitz 1992b).

Further examples of one-alone applications are pUblication databases and black board systems containing information about the institution, educational courses, teachers, dates, etc. The black board system is called student information system (SIS). It was implemented using a hypertext system where information pages are linked by keywords (Kozuschnik 1992). Thus navigation in the MS Windows­based SIS is easy. For remote access to this system a client-server architecture was designed (Adler 1992). The remote system presents information conveniently on a graphical user interface.

4.2 One-to-One Technique

Various components have been added to the SIS: Electronic mail and file-transfer functions as well as an application which enables students to register for excursions and examinations. The systems that allow asynchronous human-to­human communication are the base for distributed flexible team-work among students and teachers. A growing number of studies show the advantages of the application of electronic mail in virtual classrooms. Hiltz for example found that students tend to be more active when teachers use electronic mail instead of face­to-face communication in conventional classes (Hiltz 1988).

Furthermore some innovative teachware systems for group learning have been developed: Students produce parts of a common document in single-work as members of special teams. A coordination module distributes tasks and helps to clarify who produces which part and when it will be completed. Using electronic mail students are able to communicate in order to solve problems they are confronted with. This is why this teachware is attached to the one-to-one technique class. In a further step they join the different parts into one document and transfer it to the file area of the coordination module. Meanwhile an agent produces an access protocol including all activities that take place when students interact with the coordination module. Thus teachers are able to check the single­work and team-work abilities of their students.

4.3 One-to-Many Technique

Multiple bulletin boards have been integrated into the SIS. Due to these functions the enhanced SIS is now called student information and communication system (SlUeS) and opens uni-directional (see Sect. 4.1), bi-directional (see Sect. 4.2)

102 R. Seitz, F. Bodendorf

and multi-directional asynchronous communication channels. Thus information exchange in groups consisting of students only, students and teachers or teachers only is possible.

4.4 Many-to-Many Technique

An integrated text-oriented electronic meeting system (ITEMS) has been developed to support synchronous and asynchronous conferences (Grebner 1993) «(Johansen 1988) (Wilson 1991) (Nunamaker 1992) (Petrovic 1991) (Hiltz 1992). The system has been enriched by group decision support functions (Bui 1987) and co-authoring facilities (Trigg 1992) (Leland 1988). A system like ITEMS enables the application of pedagogical techniques as for example: delphi technique, nominal group technique, brainstorming, debate, role play etc. (paulsen 1993).

I students' side I tsachwal8 with

morittri'lg

~ -~.~- ~ transniSllion

I ibrariEB and da~bases - -server

I u niversily' s side I

Fig. 2. Learning environment

The installation of these systems demands major technical efforts. However, spreading these new possibilities all over the university is a much higher challenge (Bullen 90). In order to reach this goal we have to change historically grown patterns of students' and teachers' behaviour.

Innovative Support Technologies 103

5 Experiences with Computer Supported Team-Work in Education

To give more details concerning the educational use of ITEMS it is useful to describe a typical workgroup session. In this case the group consists of one teacher and five students. The design of the session is as follows:

Pedagogical Design

Playing roles of leading persons in politics and business, students discuss the theme "The Current Problems in Germany". Coming up with every-day­knowledge from newspapers as well as more theoretical-oriented capabilities from their studies they represent the interests of a certain group or party. So they have to give arguments, opinions and strategies that do not match their own ones all the time. The role of the teacher is a more passive one. He plays no real part in the group but acts as a moderator and animator. His part is to give content­oriented or intellectual contribution as well as more organizational contributions as for example hints to come closer to the real problems. When students have to come to decisions or if they have to vote for anything the moderator initializes and starts the decision process.

The advantages of such a kind of role play are, on the one hand, to improve reasoning and acting especially when interests of other people are concerned. Thus they practice social competence. On the other hand, they adopt the very important ability to problemize themes, to discuss aspects in detail and to generate plans in groups.

Organizational and Technical Design

The complete session can be organized only by the help of electronic media. First of all we send e-mails to each member of the project "portable PCs for students". This message includes a short description of the role play itself as well as the roles which are involved. Furthermore, organizational details are given to the students. Interested students who want to join the session send a reply for booking. So we are able to constitute a group of volunteers for the group work. Half an hour before the actual session, members meet for the first time. The moderator then explains the role play in more details and shows the handling of the system ITEMS. Later on the session is divided into two phases. During the first hour all participants come together in one room, each person provided with one workstation. The situation is quite similar to that in an electronic decision room (EDR). So communication is possible both face-to-face and by using the system. After one hour we change the situation by switching the role play to a distributed mode. That means that only the system is allowed to communicate within the group.

The session itself, geographically distributed or not, is moderated to run in cyclic phases. The phases are "discussing the most interesting problem or question", "generating plans, actions or new questions" and "leading through votings concerning plans, actions or new questions". Later on the phase "discussing the most interesting problem or question" recurs on a more detailed or new field of interest.

104 R. Seitz, F. Bodendorf

Evaluation

The role play has been evaluated intensively by analysing the session protocol and two paper-based questionnaires. Contrary to the wider empirical study (Sect. 6) that inquires about the whole project and its effects, our focus here lies closer to special results concerning team-work among students and the evaluation of ITEMS. Furthermore we intend to find out whether both students and ourselves are able to transfer the experiences made in our university to the world of business. What we have learned until now is that students believe in sensible usage of computer conferencing systems including group decision support systems for education. They confirm that it is important to learn how to work with such systems, because they are sure that they will have to work with such or similar systems in business later on. In general they consider computer support for team-work to be helpful and suggest spreading this kind of learning over other parts of education.

6 Empirical Study

6.1 Evaluation Methodology

The evaluation of the complete project is based on foundations as for example the Fishbein theory (Fishbein 1963) (Fishbein 1967). Its purpose is to test the mental attitudes of treatment and control groups before and after the experiment. Following the Carroll theory (Carroll 1985) we check the learning success of students. Based on these and some other theories we designed a spectrum of data collection instruments like:

Computer-based Instruments

- Analysing the login-script of the communication server we extract a lot of quantitative data material. The script shows how many connections were made and how long the students stayed online. In some special cases it is possible to review what kind of service they were using. - A computer-based questionnaire was developed and installed on the students' portable PCs. Two different modi gather qualitative and quantitative information concerning the utilization of the equipment on a daily as well as weekly base. The software generates files which are transferred to the evaluation team at the university.

Paper-based Instruments

The paper-based questionnaire is more detailed than the computer-based one. While the computer-based component can be used only by the treatment group, the paper also reaches the control group. Both groups get the questionnaire three times within six months. This instrument contains the components of the Fishbein theory.

Innovative Support Technologies 105

Interview-based Instruments

Furthennore, several interviews are held with students from both, the treatment as well as the control group and from students that are not involved in the project. Afterwards teachers, experts in designing and developing infonnation systems and members of the project team are questioned.

6.2 Results

The results presented here have been acquired throughout a period of 18 months. The extensive empirical study had been started just before the third project semester. Then the project was highly frequented. We had the chance to select a treatment and a control group showing very similar demographic structures. Our sample group comprised 108 persons. 56 students belonged to the treatment group, 52 students to the control group.

The following statements give a short overview of our preliminary results (Eggert 1993):

Demographic Data

The average income of students participating in the experiment is low. In most cases they have to work for income. This is why flexible and decentralized learning can be seen as a very helpful strategy. Nearly 90% of the persons belonging to the treatment group have access to a telephone network port to use their telecommunication equipment.

Table 1. Demographic data

Over All Treatment Group Control Group

Men to Women 86: 22 43: 13 43 :9 Proportion (Rates)

Average Age 25.04 25.11 24.96 (Years)

Average Sum of 7.7 7.9 7.6 Studied Semesters (Semesters)

Attitudes

83% of all participants said that they accepted spending money for telecom­munication. Nearly 93% were very interested in using the possibilities provided by the learning environment.

Members of the treatment group in comparison to the control group see a high educational potential of the learning environment before participating in the experiment. Changes in attitude could be observed: Electronic mail, for example, was seen as more useful for education at the end of the study compared to the beginning.

106 R. Seitz, F. Bodendorf

Before participating in the experiment students of both groups had said that pes were very important for education. After the half-year-period this opinion was weakened. Both groups believed in a high learning success for people joining the project. This positive opinion was even stronger at the end of the project.

According to the students' statements, knowledge acquisition by use of computers grown during the process of the project.

The following three advantages had been expected by students before participating in the project: (a) more flexible learning, (b) practical experiences with computers that are useful for business, (c) saving time and transfers to the university.

In comparison they mentioned the reduction of social contacts which was expected to limit actions of personal exchange.

Overall the attitudes towards decentralized and cooperative work were very positive in both groups. Generally students of the treatment group judged their participation as a personal gain.

The following statements are extracted from the computer-based questionnaire answered by the treatment group only.

Behaviour

Nearly 67% stayed at home more often while participating in the project. The average number of reduced drives to the university adds up to four a week. Students were able to learn more flexibly during their project participation. This can be shown by the help of the login-script. According to this the main system usage took place between 10.00 p.m. and 11.00 p.m. The remainder of the work is equally distributed over the whole day.

In addition to time-based flexibility the students appreciated taking their notebooks with them when they left home. A sample test with 20 students shows: 295 out of 2,680 hours the notebooks were neither used at home nor at the university. The same group's communication equipment was used during 9,1% of the total usage time (245 out of 2,680 hours).

Systems

From the students' point of view there was no gap between the systems supplied and their needs. In particular, programming and computer-based courses were considered as courses where decentralized work could be done very usefully.

The average number of logins per week was more than five per person. The functionality of our offered software was considered to be reliable and good.

The following systems were most frequently used: information and communication systems, word processing systems and graphic processing applications.

Among 660 feedbacks from the students there were 41 messages concerning problems with our systems. The mentioned problems concerned the fields of hardware, software and telecommunication.

In comparison with the control group members of the treatment group showed an obvious increase of knowledge concerning the user interface Microsoft Windows, the network operating system Novell Netware and computer-mediated communication.

Innovative Support Technologies 107

References

Adler, L. (1992) Entwicklung eines externen Zugangs- und Kommunikationsmodules fiir ein zentrales Studenteninformationssystem an der WiSo-Fakultiit. Diplomarbeit. Niirnberg

Aiken, M. W. (1992) Using a Group Decision Support System as a Teaching Tool. In: Journal of Computer-Based Instruction. Vol. 19 (3), 82-85

Bui, T. X. (1987) Co-oP, A Group Decision Support System for Cooperative Multiple Criteria Group Decision Making. Berlin: Springer-Verlag

Bullen, C. V., Bennett, J. L. (1990) Learning from User Experience with Groupware. In: CSCW' 90, Proceedings of the Conference on Computer-Supported Cooperative Work; New York: ACM, 291-302

Carroll, J. B. (1985) A Model of School Learning. In: Anderson, L. W. (Editor-in Chief) Perspectives of School Learning. Selected Writings of John B. Carroll. Hillsdale, New Jersey: Lawrence Erlbaum Associates, 19-31

Dittrich, J. (1991) Koordinationsmodelle fiir Computergestiitzte Gruppenarbeit. In: Friedrich, J., ROdiger, K.-H. (Hrsg.) Computergestiitzte Gruppenarbeit (CSCW). Stuttgart: Teubner, 107-118

Eggert, V. (1993) Konzeption und praktische Durchfiihrung einer Programmevaluation des Modellvorhabens "Mobile PCs fiir Studierende". Diplomarbeit. Niirnberg

Fishbein, M. (1963) An Investigation of the Relationships between Beliefs about an Object and the Attitude toward that Object. Human Relations 16, 233-239

Fishbein, M. (1967) A Behavior Theory Approach to the Relations between Beliefs about Object and the Attitude toward the Object. In: Fishbein, M. (Editor-in-Chief) Readings in Attidude Theory and Measurement. New York, London, Sydney

Grebner, R. (1993) Konzeption und Realisierung eines computergestiitzten integrierten Konferenzsystems . zur Unterstiitzung von Gruppenprozessen in der Universitiit. Diplomarbeit. Niirnberg

Hiltz, R. H., Turoff, M. (1992) Structuring Computer-Mediated Communication Systems to Avoid Information Overload. In: Marca, D., Bock, G. (1992) Groupware: Software for Computer-Supported Cooperative Work. Los Alamitos: IEEE Computer Society Press, 384-393

Hiltz, S. R. (1988) Collaborative Learning in a Virtual Classroom: Highlights of Findings. In: CSCW 88, Proceedings of the Conference on Computer-Supported Cooperative Work, 1988. New York: ACM, 282-290

Johansen, R. (1988) Groupware: Computer Support for Business Teams. New York: The Free Press

Kozuschnik, F. (1992) Konzeption und Realisierung eines prototypischen Kommunikationssystems fiir Studenten an der WiSo-Fakultiit auf der Basis eines Inhouse-Netzes. Studienarbeit. Niirnberg

Leland, M. D. P., Fish, R. S., Kraut, R. E. (1988) Collaborative Document Production Using Quilt. In: CSCW 88, Proceedings of the Conference on Computer-Supported Cooperative Work, 1988. New York: ACM, 206-215

Lewin, K. (1947) Subjective and Objective Elements in the Social Field: The Three Step Procedure. Human Relations 1

Malone, T. W., Crowston, K. (1992) What is Coordination Theory and How Can It Help Design Cooperative Work Systems? In: CSCW '90, Proceedings of the Conference on Computer-Supported Cooperative Work, 1990. New York: ACM, 357-370

Nunamaker, J. F. et al. (1992) Electronic Meeting Systems to Support Group Work. In: Marca, D., Bock, G. (1992) Groupware: Software for Computer-Supported Cooperative Work. Los Alamitos: IEEE Computer Society Press, 350-371

108 R. Seitz, F. Bodendorf

Paulsen, M. F. (1993) Some Pedagogical Techniques for Computer-mediated Communication. In: Pre-Proceedings (Part Two) of the NATO Advanced Research Workshop in Segovia, April 24-27, 1993. Segovia, 2-13

Petrovic, O. (1991) Standardsoftware als Basis eines Integrierten Electronic Meeting System. In: Friedrich, 1., ROdiger, K.-H. (Hrsg.) Computergestutzte Gruppenarbeit (CSCW). Stuttgart: Teubner, 183-196

Piepenburg, U. (1991) Ein Konzept von Kooperation und die technische Unterstutzung kooperativer Prozesse in Biirobereichen. In: Friedrich, J., ROdiger, K.-H. (Hrsg.) Computergestiitzte Gruppenarbeit (CSCW). Stuttgart: Teubner, 79-94

Rapaport, M. (1991) Computer Mediated Communications: Bulletin Boards, Computer Conferencing, Electronic Mail, and Information Retrieval. New York: John Wiley & Sons

Saur, T. (1992) Entwicklung und Realisierung eines Konzeptes zur Dezentralisierung von Dialogpraktika im Bereich der universitiiren IV-Ausbildung. Diplomarbeit. Nurnberg

Seitz, R., Bodendorf, F. (1992b) Modellvorhaben "Mobile PCs fUr Studierende" -Nutzungsmoglichkeiten, Bericht Nr. 2. Arbeitspapier Wirtschaftsinformatik 13/1992. Niirnberg

Seitz, R., Bodendorf, F., Schmidt, W. (1992a) Modellvorhaben "Mobile PCs fUr Studierende". Arbeitspapier Wirtschaftsinformatik 3/1992. Niirnberg

Sephard, A., Mayer, N., Kuchinsky, A. (1992) Strudel - An Extensible Electronic Conversation Toolkit. In: CSCW' 90, Proceedings of the Conference on Computer­Supported Cooperative Work, 1990. New York: ACM, 93-104

Singer, J., Behrend S. D., Roschelle, J. (1988) Children's Collaborative Use of a Computer Microworld. In: CSCW 88, Proceedings of the Conference on Computer­Supported Cooperative Work, 1988. New York: ACM, 271-281

Trigg, R. H., Suchman, L. A., Halasz F. G. (1992) Supporting Collaboration in Notecards. In: Marca, D., Bock, G. (1992) Groupware: Software for Computer­Supported Cooperative Work. Los Alamitos: IEEE Computer Society Press, 394-403

Wilson, P. (1991) Computer Supported Cooperative Work. Dordrecht: Kluwer Academic Publishers

Winograd, T., Flores, F. (1986) Understanding Computers and Cognition. Norwood, New Jersey: Ablex Publishing Corporation

9 Remote Collaboration in Medicine: Some Tools and Experiences in Portugal

A. Sousa Pereira!, J. Arnaldo Martins!, J.A. Veiga Pires2

! Universidade de Aveiro I INESC, P-3800 Aveiro, Portugal

2 Centro Hospitalar de Vila Nova de Gaia, P-4400 Vila Nova de Gaia, Portugal

Abstract. This paper identifies specific scenarios in healthcare where remote collaboration tools can be used. ISDN based systems with graphical user interfaces seem to be the most appropriate environments for remote collaborative work in medicine. Several applications developed and under development in Portugal are described.

Keywords. Cooperative work, Collaborative work, ISDN, Medicine

1 Introduction

In the last years there has been a qualitative change in medical practice. The strong one-fo-one relationship that existed between the family doctor and the patient is evolving into a one-to-many relationship. This is due to the fact that medicine is becoming more specialized, being necessary to establish medical teams involving specialists from very different areas. This fact also imposes new requirements for medical training, and continuing education and training of doctors is becoming fundamental to a good quality of healthcare services.

2 Collaborative Work in Healthcare

As previously stated, healthcare is becoming very specialized, being delivered by a team of health care providers, located at various places and needing to share information. Different actors participate in the process of delivering healthcare, with different backgrounds and belonging to different organizations. This system sometimes lacks integration between the different actors, and the communication channels are very inefficient.

As an example we can have the following scenario: In the case of a pregnancy the woman usually refers to her family doctor, a General Practitioner, resident in

110 A.S. Pereira, l.A. Martins, lA. Veiga Pires

the local Healthcare Centre. There is a clinical protocol for the follow up of the pregnancy which involves, besides the General Practitioner, Obstetricians, Midwives and later on Pediatricians. This team is not physically located in the same place, but all of them generate clinical data which are important in the follow-up of the pregnancy. According to the status of the pregnant woman, decisions will need to be taken involving the know-how of the different members of this team. In this scenario asynchronous communication channels are usually used imposing severe constraints in the whole process of delivery of care.

3 Distance Learning and Training in Medicine

Information dissemination, and continuing education and training of the doctors, are vital for good quality of care delivered to the citizens. The increased mobility of the Healthcare service users has as consequence a very fast spread of diseases (e.g. new flue types). New pharmaceuticals or R&D results in the treatment of diseases or specific clinical situations, can have a strong impact in the daily practice of the doctor. Doctors are sometimes located in remote areas where specialists are lacking; situations can occur where a non-specialist doctor could solve a problem if advice could be taken from an expert.

All the above described scenarios strengthen the importance of the establishment of distance learning and continuing training mechanisms in Medicine.

4 IT Impact in Healthcare

Recent developments in the Information Technologies (ITs) are producing a big change in our society, introducing new mechanisms and tools for data archiving, display, manipulation and transmission. Healthcare is one of the areas in which ITs had a greater impact either at management and also at clinical level.

The introduction of new telecommunication infrastructures, such as ISDN, potentially enables the development of advanced telecommunication services, for instance enabling a better integration of Primary Care and Secondary Care Health services by means of Telematics, or the remote support to the elderly or disabled, between other possible applications. By means of Telematics remote areas are no longer isolated, and General Practitioners located in remote Health Centres or Hospitals can look for advice from experts located in Central Hospitals or specialized Health Centres.

4.1 Telemedicine

Telemedicine, the integration of medicine and telecommunications, is thus becoming a subject of capital importance, its functional definition being the following (Van de Velde 1992):

Remote Collaboration in Medicine 111

The investigation, monitoring and management of patients, and the education of patients and health care staff using systems which allow ready access to expert advice and relevant patient information, no matter where the patient is located.

Remote collaboration is one of the aspects of Telemedicine and can be considered one of the newest improvements in Healthcare.

5 Remote Collaboration Scenarios in Medicine

In the last years significant effort has been undertaken in Europe to identify and characterize different scenarios in Telemedicine. Several projects sponsored by the European Research and Development programs AIM and RACE were set up and significant results were achieved. Specifically dealing with remote collaboration the following scenarios were identified and will be described: - Teleconferencing I Teleconsultation - Professional Training

5.1 Teleconferencing I TeleconsuItation

By Teleconsultation/Teleconferencing we mean an integrated working environment where the different actors can speak and interact as if they were on the same place.

Teleconferencing can be used to support multidisciplinary conferences between specialists, or to access Specialty Centres from remote locations, increasing the quality of services in places where specialists are not present.

In this integrated environment the actors usually talk and interact over clinical data, for instance results from laboratory examinations, biological signals (e.g. ECG, EEG, Cardiotocographic data) or medical images (e.g. XRay, Ultrasound, Computed Tomographies).

Multimedia terminal equipment will be used, involving voice, text, data and image transmission, generating a significant amount of traffic in the communication infrastructure.

ISDN can be used in these applications being most of the time enough of a basic access interface, also known as the 2B+D interface, including two bi­directional 64 Kbitls channels (the 2B), and a 16 Kbitls bi-directional channel (the D) used for signalling purposes between the terminal equipment and the network, and also, for packet oriented communications. With this interface one B channel can be used as a voice channel while the other can be used for data transfer. If we have large amounts of data to be transferred and we want very low response times 2 Mbitls channels can be used with primary access ISDN.

5.2 Professional Training

There are two scenarios regarding professional training and remote education in the medical field, which will be described in the following paragraphs.

112 A.S. Pereira, J.A. Martins, I.A. Veiga Pires

Actualization courses: Hospitals can train their young doctors and technical personnel, at their sites, giving them actualization courses through connections to remote training centres (hospitals, healthcare units, etc.) or to specific professional training organizations.

Access to remote expertise: A less qualified employee could have a real time connection with a specialist located in a remote site to discuss and obtain the expert opinion about a specific problem. For example, in the medical field, we can have a young doctor in a remote hospital that needs the advice and the diagnosis of a specialist located in a central hospital, so it makes a connection to the central hospital, transmits a medical image of the patient which is displayed at both workstations, in a public window. The two doctors can make a conference about that image, using voice and some tools to manipulate the image (e.g. definition of a region of interest, marks on the image, zooms, etc.). The use of ISDN will enable the easy exchange of different kinds of data (text, graphics, images and data).

These forms of remote collaboration will save time and money, but for countries like Portugal, they also will be important for patients that are far away from the big cities, enabling them to have access to better healthcare. They will have better diagnostic tools and specialists and, on the other side, these tools will help young doctors to move to the countryside.

6. Remote Medical Applications and ISDN

The response time is an important parameter to the application success since it must be less than the response time of the human user so that his concentration is not hindered to some extent. With the use of ISDN and some compression techniques it is possible to have transmission of several types of medical images (x-ray, computed tomography, magnetic resonance, nuclear medicine, ultrasound, etc.) within acceptable response times and, since in almost all these scenarios, it is only needed to transmit one image at a time, narrow band ISDN (2B+D) will be suitable for a great number of remote collaboration medical applications.

Through the analysis of table 1 we can see that applications dealing with high mediumlhigh resolution graphic and image data take several tens of seconds for transmission. But, with compression techniques like vector quantization (Real and Alves 1990) and JPEG (Baran 1990), these times can also be reduced by factors of 15 to 30 depending on the type of image and on the amount of loss wanted for the decompressed images. Using a ratio of 15 to 20 a decompressed JPEG image shows minimal quality loss when displayed in the screen. This means, for example, that an image of 1024xl024x12 pixels (one of the most demanding images in medicine) would require about 10 seconds for transmission which is a good response time (Watanabe 1987).

Remote Collaboration in Medicine 113

Table 1. Transmission time for several kinds of applications using a B channel

Application Size Tx Time (sec)

ASCII text screen SOx24xS b/c .24 Page ASCII text 60x60xS b/c .45 Scanned Page (b/w) 6in xllin x3002dpi x S 743 Voice SOOOxS b/S 1 Graphics/Images 256x256xS b/pixel S.2 (120) = 0.41 Medium resolution 512x512x 16 b/pixel 66 (120) = 3.3 High resolution 1024x1024x12 b/pixel 197 (120) = 9.9 V. High resolution 204Sx204Sx 12 b/pixel 7S7 (120) = 40 Low Resolution Video 144xlS0xS b/pixel x S 26 (/26) = 1

With the compression algorithm developed in (Real and Alves 1990), a ratio of 26 can be achieved, depending on the scenario, making it possible to have about 8 images of 144 x 180 pixels per second, which can be useful for scenarios where the scenes do not change rapidly (surveillance of wards, elderly people, video­conferencing, etc.).

After the above descriptions we can state some of the facilities that remote medical stations must have:

• an ISDN basic access interface. • a module for data compression/decompression. • a window based display system with a large display. • devices for image and sound acquisition and reproduction.

7 Applications

In this section we will describe some of the applications that were developed or are under development by INESC in the framework of Portuguese and European research and development projects and deal with the previously described scenarios.

7.1 Teleradiology

A Teleradiology system (Figueiredo et al 1992) was developed to provide an integrated radiology environment, including the support of interactive work between medical staff, typically located in a Central Hospital and in a remote Health Care Centre. Each Teleradiology workstation is based on a PC compatible computer, including an ISDN basic access interface card, a SVGA (l024x768 pixel) videoboard and a 21" color monitor. The system can be operated in one of three basic operating modes: - local mode: providing access and processing of locally stored images;

114 A.S. Pereira, J.A. Martins, J.A. Veiga Pires

- remote access mode: providing access to images stored in a remote system for local processing, without the need of remote operation; interactive mode: providing simultaneous display and processing facilities on the two remote workplaces, and a voice channel for user communication.

A window based iconic interface was developed for user interaction. The system also provides a compression/decompression facility, with a user selectable compression rate according to the admissible error rate.

In the interactive mode doctors establish a cooperative work environment where different tools are available. A telepointer controlled by the mouse can be moved over the displayed images (e.g. x-rays, computed tomography slices) to show specific structures or lesions appearing in the images. Some basic image processing tools, such contrast enhancement or a magnifying glass, can also be used by either of the system users', affecting in a complete transparent and symmetric way both display working areas. A voice channel is also present allowing simultaneous discussion while the image are being displayed and manipulated. A text window can also be used to exchange alphanumeric messages between both workstations.

It is expected that during 1994 this system will be installed in different Medical Imaging Departments of several Portuguese hospitals, to be used as remote medical expert advice tool.

7.2 EPIC - Antenatal Demonstrator

In the framework of the A2007 EPIC - European Prototype for Integrated Care an Antenatal Demonstrator is being developed (Sousa et aI1993). The purpose of this demonstrator is to integrate Primary and Secondary Health Care Centres by means of Telematics. The pregnant woman data (demographic and clinical) as well as the new born data will be available in both places, the Maternity Hospital and the Health Centre. Cardiotocographic data will be included in the system by means of a specific acquisition module, and remote collaboration procedures to interpret examinations will be established. The system is currently being modelled using an Object Oriented methodology and a first prototype of the user interface was developed using a graphical interface in a window based environment.

7.3 TeleCommunity

The European TeleCommunity project, sponsored by the RACE program, explores the telecommunications infrastructures to assist remote people with special needs (elderly and disabled).

The Portuguese pilot site is by INESC / Universidade de Aveiro, Faculdade de Motricidade Humana and TLP (the Portuguese PIT operators of Lisbon and OPorto), and has as its objectives the implementation of several remote support services, using the ISDN pub lic network, which are classified in the following classes: Remote Delivery of Expertise, Learning and Training, Interpersonal

Remote Collaboration in Medicine 115

Communication and Collaborative Decision Making. INESC is developing a PC based multimedia terminal with specially designed graphic interfaces, taking into account the limitations of the final users, with ISDN support and the possibility to transmit and display real time images (Pacheco et aI1993).

7.4 CO-LEARN Portuguese Pilot Site

INESC / Universidade de Aveiro is engaged in the set-up of the Portuguese pilot site, in the medical area, for the evaluation of the tools developed in the CO­LEARN European project (DELTA 2005), which are: Asynchronous Conferencing System, Real-time Conferencing System, Real-time Teleteaching and Real-time Teleassistance. They will be used for teaching and training young doctors in medical imaging. Two units of the HUC-Hospital of University of Coimbra will be connected, using ISDN, for the purpose of this trial. By using a multi-media editor lessons in the specific domain of Medical Imaging will be prepared, involving text, images and graphical information. Diffusion mechanisms will be established allowing for young doctors remote access to lessons stored in the information server.

The pilot site is expected to start its operation during the first semester of 1994.

8 Conclusions

The described scenarios for remote medical collaboration are all point to point, for which is good to use ISDN. For remote collaboration among several groups ISDN has some limitations that must be overcome by additional tools, but from our experience, collaboration among more than 2 groups of doctors is not frequent.

Another conclusion is that narrow band ISDN (2B+D) is adequate for a great number of remote collaboration applications in medicine, since we deal, basically with still images.

A user friendly graphical interface under a window environment seems most appropriate for these remote collaboration applications in medicine.

References

Baran, N.: Putting the Squeeze on Graphics. BYTE, pp. 289-294, December 1990. COLEARN-Cooperative Learning: Technical Annex, Project DELTA 2005. Filipe Figueiredo, L., Sousa Pereira, A., Ramos, F.: A low cost ISDN based

teleradiology system. Proceedings of the 14th. Annual International Conference of the IEEE Engineering in Medicine and Biology Society. 1206. Paris, 1992.

Pacheco da Rocha, N., Moniz Pereira, L., Cidade, C.: Remote Learning and Training Services for Elderly and Disabled People. Teleteaching 93. Trondheim: Norway, August 1993.

116 A.S. Pereira, 1.A. Martins, I.A. Veiga Pires

Real, L.C., Alves, A.P.: Vector Quantization of Image Sequences Using Variable Size and Variable Shape Blocks. Electronic Letters, Vol. 26, pp. 1483-1484, August 30, 1990.

Sousa Pereira, A., Trigureiros, P., lustiya, A., Sarmento, P., Batista, I.: Antenatal Demonstrator - Present Situation. Epic Internal Report. Aveiro: February 1993.

Van de Velde, R.: Hospital Information Systems - The Next Generation. Berlin: Springer-Verlag 1992.

Watanabe, H.: Integrated Office Systems: 1995 and Beyond. IEEE Communications Magazine, Vol. 25, No. 12, pp. 74-80, December 1987.

Part Two

Models and Systems

10 Models and Systems for Collaborative Dialogues in Distance Learning

Stefano A. Cerri

Dipartimento di Scienze dell'Infonnazione, Universita di Milano, Via Comelico 39, 1-20135 Milano, Italy E-mail: cerri@hermes.mc.dsi.unimLit

Abstract. The paper reports about the writer's interpretation of the Models and Systems stream of work from a methodological viewpoint more than from a technical one. Inputs were provided by the discussions held during the workshop among the participants that contributed to the Models and Systems working group.

Keywords. Theory construction, experiments in learning, applications to learning, CSCW, CSCL.

1 Introduction

In order to better focus the discussion among the workshop participants, three texts each oriented to specific aspects of the meeting were circulated before the beginning of the workshop. In Section 2 of this paper we will introduce the Models and Systems area. In Section 3 we will report about the discussions that emerged during the workshop, particularly the conclusions that will be tentatively linked to the introductory text and the workshop papers.

2 Models and Systems Within CDTDLl: the Questions

2.1 Purpose of this Dote

Models and systems represent the formal - technical component of the workshop. Issues that deserve discussion are reflected by the texts available2 and will probably be even more focused by the other ones not yet published in the pre­proceedings. In the following, a first short reaction to the papers I have read and some suggested topics for group discussion at the workshop.

lCollaborative Dialogue Technologies in Distance Learning 2 in the fonn of pre-proceedings, for the author, coordinator of the working group at the moment this introductory text was written, i.e. at the beginning of the workshop.

120 S.A. Cerri

2.2 Understanding the state of the debate about models and systems

In the workshop title we have specifically indicated "dialogue" in order to contrast our discussions from the ones - perhaps even more popular at the moment -concentrating on the simple availability for the user of complex documents to be accessed through the network. Those issues, mainly related to multimedia and networking technologies, are considered to be perhaps necessary but probably not sufficient conditions for Computer Supported Collaborative Learning. It is not certain that for the simple reason that one enters a library one becomes literate, even if good books help literacy. Similarly, the learning scenario will be richer when networks and multimedia materials will be available, but nevertheless the human teacher/expert - human learner dialogue will remain a necessary component of most of our learning experiences, providing motivation, guidance, expertise, support and stimulating creativity. Access to sophisticated sources of information is better defined as a monologue, where the information flow is mainly unidirectional.

One question is concerne with if and how a learner may engage in such a dialogue with a system that plays the role of a teacher or expert. Attempts to answer this question are done by people that design tutoring systems, programs that engage in tutoring I training dialogues with students. Even if these systems may be part of a distance learning set-up, usually one does not identify technologies in distance learning with software sent by mail or net and used by the student once loaded into a local computer. Making computer programs better able to manage dialogues with learners is usually referred to as research in flexible learning; while research in distance learning refers to methods and tools for managing significant dialogues and conversations among remotely located human actors (learners, teachers, experts, tutors, ... ) through a telecommunication network.

Dialogues between two partners, or conversations among many (human) partners are interactions that usually have a purpose identified along the dialogue by mutual consensus. The identified common purpose becomes the parameter for evaluation: when the purpose has been achieved, the dialogue is successful. Until that moment is reached, a "distance" from the purpose is continuously evaluated by each partner. At each move, considering the "distance", each partner applies the best strategy (s)he knows in order to reach the goal by reducing the distance. When we explicitly refer to the goal of teaching (and learning by the other partner), the dialogue may be called a teaching dialogue. But learning may occur as a side effect of informative conversations (querying a data or knowledge base, navigating in a hypertext, ... ), or even conversations that include imperative phases (such as phases of design / tailoring of tools / playing with simulated artefacts / ordering an assistant to perform an action, ... ). Therefore in principle any conversation may have learning as a direct effect of a teaching phase or indirect effect of an informative or imperative phase. One should, however, distinguish (and verify) when the control of the conversation has the explicit purpose of teachingllearning and when, even if it does not, learning occurs as an uncontrolled consequence of a conversation that has another explicit objective. Now, the real issue for a distance learning scenario is to establish how

Models and Systems for Collaborative Dialogues 121

conversations may be activated that, maintaining the global purpose of teaching / learning, use phases of explicit teaching but also informative and imperative phases.

Inevitably, models of these conversations are required. It appears to me that a few generic tools are available now, asking for an application, and, conversely, a few modelling experiences are available asking for new specific tools. Why is it not easy to combine the two components of the puzzle? I have thought of five paradoxes; each to be easily interpreted as a question expecting an answer from the audience attending the workshop.

2.3 The first paradox: Need for specific (teachingllearning) conversation models

Until specific models are available, specific systems can hardly be build. Conversely,until systems are available, it is difficult to develop models as it is difficult to collect significant concrete experience unless one substitutes experience with wishful imagination. In order to start the process anyway, most authors have accepted an evolutionary view of system construction by successive refinements. Who is going to refine / maintain those systems? The programmer? The user? This question leads to the following one.

2.4 The second paradox: Need for (automatic) knowledge acquisition

One would need automatic methods of knowledge acquisition in order to extract and store significant (formal) data and knowledge. However, in order to build automatic knowledge acquisition tools one should analyse large amounts of dialogue protocols that are currently unavailable. Is it possible to conceive semi­automatic knowledge acquisition methodologies as a link between the two extremes: the "programmed" and the "radically tailorable" systems of the future (as, for instance, spreadsheets or hypertexts)? Or, instead, would it be preferable to include a programmer / a knowledge engineer in the overall design of the con­versation cycle aiming .at system refinement including knowledge formalisation and acquisition?

2.S The third paradox: Need for evolving classification models

One may assume to ask the users to classify documents formally (messages of any kind transferred in the network, including software, electronic books etc.) in a more and more precise, controlled and sophisticated way, in order to facilitate the subsequent formal manipulation (search, access, transfer, reuse, etc.). The process may go to the extreme and be counter effective: where is the trade off? How "structured" are/should be semi-structured objects? In order to control large libraries of different types of documents, one usually decides on a standard classification and requires everybody to accept it. It is well known, however, that for any imposed standard there is a natural inclination to think of a better standard

122 S.A. Cerri

incompatible with the previous. Should classification criteria for documents in such a delicate domain as learning be also considered an objective to be approximated by natural selection and evolution? Are our existing "tailorable" tools - in principle - able to control the evolution of classification criteria, at least in a limited way?

2.6 The fourth paradox: Need for models of diagnosis

Assuming that users are collaborative, the application of typical CSCW tools tailored to learning according to CSCL models seem to allow the management of complex interactions. Typical office interactions, however, assume that the event that a human agent performs incorrectly is an exception that should be avoided or, at least, inhibited. In learning. on the contrary, mistakes are normal and correct performance is the exception: once reached, learning has been accomplished and the dialogue may be concluded. Mistakes, in addition, can be a fruitful source of knowledge for the learner, provided there is a skilled trainer/teacher able to diagnose the mistake and help the learner to remedy the underlying misconception. Therefore, the process of diagnosis and remediation is not to be rejected as an enemy but instead to be used for the purposes of the conversation. That process is such that the knowledge available to any of the two partners, by definition, is not shared and cannot be made explicit to the other partner until the diagnosis and remediation have been successfully performed. Most models and systems available today focus on managing conversations for tasks of an informative or planning/scheduling or imperative nature. What about conversations for diagnosis?

2.7 The fifth paradox: Simple but real obstacles to computer supported collaborative learning

Finally, thanks to the evolution of various technologies it seems that virtually any software can run on any computer, and conversely, virtually any machine can access and run software on virtually any other machine connected in any network. But, the easier the access the more difficult it is to convince people to profit from it. These "negative human factors" constitute an obstacle to the most promising consequence suggested by CSCL, i.e. a better collaboration among humans? If this is the case, are there concrete arguments today - from the models and systems viewpoint - that may help to overcome this obstacle? .

3 Models and Systems Within CDTDL: the Answers

The report will be divided into two aspects: the overall scenario that identifies two complementary views within CDTDL, i.e. the pedagogical-learning view and the models and systems view;

- reflections on the second aspect.

Models and Systems for Collaborative Dialogues 123

3.1 Overall scenario

3.1.1 Content centered view

CDTDL may be viewed as applications of theories of learning and pedagogical techniques to concrete settings with the use of telecommunication technologies. Alternatively, they may also be seen as an application domain for new information and telecommunication technologies. In both cases the content of learning and teaching dominates the applications. For each specific learning and teaching issue or, alternatively, for each specific technology one may envision and realize an experiment that verifies if the application makes sense, i.e. it is effective and efficient. The purpose of these activities is specific to the learning/teaching issue or to the technology to be applied. For instance: one may show that "in order to teach Economics we may use electronic mail and business games tailored to the specific domain" or, alternatively, "in order to find an interesting application domain for telecommunication, one may use a business game and link the players in network so that they learn economics".

Results of these activities are intrinsically specific, often qualitative observations that, by definition, do not aim at enlarging significantly the knowledge about the psychology of learning (assumed to be embedded into the assumptions of the activity performed) or about the potential of telecommunication technologies (assumed to be available before the applicative experiment is performed). Further to validating the existing psychopedagogical theories or existing telecommunication systems, one may refine them at a level where they may be usefully integrated with each other and used for the purpose identified.

Several papers at the conference are reporting about useful and successful applications performed in various settings or suggested for different learning situations. The point here is often to relate the purpose of the initiative and the associated choices with the results in terms of learning.

3.1.2 Content "free" view

On the other side we have a complementary, though very different attitude: the one that aims at an the experimental construction of theories. Within this view, the goal is to build scientific models that are falsifiable (but, to date verified) of the human learning and teaching processes. In order to build these models, one uses the formal languages and the suitable "laboratory" tools; among which are the Information and Telecommunication technologies. These are good candidates for modelling and experimentation. In this context, R&D has a more fundamental objective: new quantitative models and new systems should broaden our knowledge on how learning and teaching occurs and how we may build generic systems that embody functionalities that support learning and teaching. New generic formal models and systems will eventually enlarge the archive available for applications with some guarantee that instances of these models and systems will perform usefully in a variety of concrete situations.

Neither of the two views excludes the other one, nor occurs necessarily before. Neither of the two views is necessarily associated to a single action. The view is

124 S.A. Cerri

chosen by the actors, and, by consequence, goals and plans for action. The (mainly) specific qualitative observations generated by the applicative researchers feed the experimental ones stimulating hypotheses and, conversely, new general theories from the experimental theorists influence the emergence of new specific applications. The level of abstraction and the objectives define the view. Pedagogists, computer scientists, engineers, psychologists may at any time engage in either of the two endeavours. What is interesting for each of them is a neat comprehension of the role played each time in the cycle.

It is inadequate to say that the Pedagogy and Learning stream of the book (of the workshop) may be purely associated to the Applicative view, while the Models and Systems to the Theory Construction view. A more balanced evaluation would consider that each paper addresses themes and presents contributions in both directions. However, it seemed useful for the workshop participants to have a general reference where to classify (at least tentatively) each progress.

3.2 The models and systems

The discussion by the Models and Systems group reached a few points of conver­gence that may be outlined briefly as follows.

3.2.1 Learning as a side effect

The issue central to CDLCD is the analysis and synthesis of the pragmatic aspects of learning dialogues. Pragmatics should be considered in linguistic terms, while (human) learning in psychological terms. Learning may be considered mainly as a side effect of activities among which explicit teaching is an example. The notion of dialogue is considered different from the notion of communication, which is broader in scope. The behavioural level related to learning as a side effect of dialogues was distinguished from the action level related to each communicative act. The granularity of descriptions at the action level was considered to be one of the most important to define: for instance; is "pointing" and "clicking" an act in the sequence of communicative acts?

3.2.2 Where to start from: the definition of a common lexicon

Two directions of investigation were recognized to be good candidates: one that starts from the analysis of teaching-learning dialogues (protocol analysis, interviews, etc.); the other that starts from the epistemological questions associated with the concepts we are talking about. For instance, what is a multimedia presentation, what are the roles allowed I encouraged in dialogues, etc. The idea is to study classifications of concepts on the domain in order to build progressively an agreement on a few conceptual, epistemological "standards" necessary for unambiguous descriptions of experimental results.

Models and Systems for Collaborative Dialogues 125

3.2.3 Defining what we do not mean

It is also interesting to start from an agreement about the negation of those concepts, for instance while CDTDL do profit from multimedia and networking technologies, they are not (mainly) concerned with either, as the first may be considered as presentation modalities in communicative acts, while the second as enabling factors for remote communication. Once the goal is to verify the behavioural level of distance learning dialogues, single acts as well as enabling factors are not the main concern but parameters of the experiments.

3.2.4 The industrial standards slowly emerging

In the models and systems area there was a common agreement that the lack of standards for the pragmatic level of human-computer communication was hindering any consequent theory construction. An example of a slow, but necessary process of agreement in the Information and Telecommunication technologies may be considered the ISO-OSI tower of standards. Similarly, standards emerge in programming languages more and more as soon as there is a need for industrial production of large products and their reuse.

3.2.5 Identifying a promising evolution

This encourages our hope that the process of standardization in our CDTDL domain will be accelerated in the next years because now basic technologies are ripe for concrete, monitored experimentations and therefore an increasing interest for industrialization and reuse will push on agreements about standards. Technological standards will enable epistemological standards and the coherent performance of real experiments that will feed new technical developments and new applications, further to a more systematic construction of validated theories of learning.

Acknowledgements

The participants to the working group made its 3 sessions, 'one each day, quite alive, perhaps sometimes also challenging. All the participants contributed several times, both during the sessions and in discussions outside the classroom theatres. I wish to acknowledge and thank all of them: Nicholas Balacheff, Andre Boder, Jose Cuena, Ana Garda-Serrano, Thore Danielsen, Alain Derycke, Dominique Guin, Christian Lemaitre, Carla Simone, Antonio Souza, Maria Felisa Verdejo, Peter Zorkoczy.

11 Cooperative Open Systems Architecture

Christian Lemaitre, Victor German Sanchez, Cristina Loyo

L.A.N.LA Laboratorio Nacional de Infonmitica Avanzada, A.C. A.P. 696, Xalapa, Veracruz, 91000, Mexico {cll, victor, elv}@xalapa. lania.mx

Abstract. The concept of Cooperative Open Systems is introduced as the confluent area of the work done in open systems, networking, interoperability, computer supported cooperative work and distributed artificial intelligence. A cooperative network architecture based on a network of knowledge-based systems called Facilitators is presented. The facilitators are specialists in cooperation, and have the global information about the domain of cooperation, as well as a model of the agents they serve as "executive secretaries". A case study is presented of a cooperative network allowing expert systems to cooperate solving more complex problems that those they solve alone. A discussion follows of how the basic architecture of the facilitators network can be used as an intelligent computer supported cooperative work system. In such application the expert systems are substituted by human agents, and the facilitators act as they personal secretaries.

Keywords. Cooperation network, cooperative open systems, CSCW, DAI, dialogue, facilitator, open systems, speech acts.

1 Introduction

The environment of modern computer science is characterized by its tendency toward globalization through the multiplication of interconnected local, extended and world-wide computer networks. This new telematic infrastructure which is being put in place around the world vastly increases the potential of communication and information processing. The development of this potential opens up a huge field of application requiring appropriate technology. Two immediate problems demand to be solved in order for new services to be offered. They are Heterogeneity and Cooperation.

The spread of different types of computer networks employing heterogeneous systems and equipment, which are further characterized by a set of applications requiring information interchange, has occasioned the inception of a new field of research within the discipline of computer science. This field is known as interoperability.

The effort on the part of the International Standards Organization (ISO) to define standards for open systems represents one of the greatest achievements of the 80's.

Cooperative Open Systems Architecture 127

This effort led to the laying of foundations which could assure interoperability within networks at the level of equipment and systems. At the application level different types of interfaces (called application programmable interfaces, or API) have been developed in recent years and can be programmed for interoperability. These interfaces are intended to solve the problem of interoperability at the syntactic level. They address syntactic interoperability by translating the syntax of the different languages used by the application systems. There is however, yet another level above the syntactic level which involves the meaning of the messages. We are referring to "semantic interoperability". The semantic level is an open problem crucial to obtaining user-friendly distributed cooperation systems.

The institution of a means of communication and processing as powerful as computer networks would naturally have a dramatic impact on human labour. The novelty of telematic networks consists of their capacity to connect human agents and computer agents on an equal basis. Human cooperative labour here acquires a new dimension because telematic networks can, in principle, not only facilitate communication among agents (as in the case of electronic mail), but they can also provide assistance for the coordination of the tasks of different agents.

In the not too distant future it will be possible to have heterogeneous networks for cooperative work which would be capable of connecting human/computer agents with equal ease. Naturally, these -which we will call cooperative networks­will be networks of applications based on high-speed, extra wide-band computer networks. This article proposes an architecture for a cooperative network in a very specific environment involving the connection of existing expert systems. It also discusses the possibility of extending this type of architecture to the control of networks for computer supported cooperative work (CSCW). Such a cooperative network is formed by a special type of open systems which configure, in their totality, a new concept which we shall call cooperative open systems.

A particularly interesting application area in cooperative open systems is distance education. In the case of distance education, CSCW aids teaching/learning tasks by the automatic transfer of enormous amounts of messages, filtering these messages to provide the instructors and students with only truly relevant messages, and allowing the students to access both information and agents capable of explaining points which are not clear.

2 Cooperative Open Systems

The concept of Cooperative Open Systems is introduced in this paper together with a set of heterogeneous computer systems which interact through a computer network, which is equally heterogeneous. The objective of this interaction is doing cooperative work.

The architecture of these systems has three basic levels: A Computer Network. It provides the basic communication services of the

hardware and software systems.

128 C. Lemaitre, V.G. Sanchez, c. Loyo

An Application Network. Examples of its components are data base controllers, servers, etc., which allow the accessing and operation of global information.

An Agent Network. This network coordinates the cooperative work of the human/computer agents.

The distinctive elements of Cooperative Open Systems are the following: A collection of autonomous, heterogeneous, loosely coupled agents which are capable of inter-communication using a model based on the capability of each of the component agents.

- Each agent in the system has the knowledge of possible cooperation strategies, applicable to different types of problems. These strategies involve the capacity of agents to control dialogue between one agent and another.

- The dialogues are essential units of the communication protocols among the agents; the dialogues are formed by a structured set of messages which have clearly defined goals. This architecture is inspired on the layer methodology of OS!.

- The dialogues occur between peers, following the OSI layers philosophy, but also between adjacent layers of a single agent. The cooperation which is structured through a series of layers, use as basic communication services those offered by the computer network.

The idea of dialogue between cooperative agents with which we are dealing in this proposal is a generalization of the idea of dialogue as it is dealt with in open systems. It incorporates the results of the work done on the topic of dialogues between cooperative agents in the field of Distributed Artificial Intelligence (DAI), using the speech act theory of Austin and Searle (Searle 1969).

This concept of the open cooperative system is the result of the confluence of concepts taken from various fields of computer science:

1) the development in computer networks, and in particular in the drawing up of a set of norms for the OSI by ISO. The networks are defined according to a seven­layered hierarchy. Layer n of a machine dialogues with layer n of another machine. Meanwhile the relation between the two adjacent layers of a single machine is defined in terms of a service interface for what layer n-l offers to layer n;

2) the work which comes out of the field of distributed systems and the development of heterogeneous interoperative systems. The work developed in the field of interoperability of heterogeneous data bases, in particular, has enabled us to distinguish more clearly between the problem of syntactic interoperability and the problem of semantic interoperability. The former can be solved using algorithmic methods of translation which are independent of the content of the registers. The latter, on the other hand, implies an analysis of the meaning of the contents. Cooperative open systems are situated precisely at this level of communication and semantic analysis.

3) The research which forms the body of work on CSCW consists mainly in the creation of computing tools based on computer networks and which facilitate human cooperative work. The CSCW systems are based on the convergence of telecommunications and computation, and they privilege tasks involving

Cooperative Open Systems Architecture 129

coordination, communication and problem solving between human agents. 4) In the field of DAI the solution of problems is considered as a result of the

interaction of multiple agents based on loosely coupled knowledge. One of the areas of DAI, Cooperative Distributed Problem Solving (CDPS), seeks to solve problems which are similar to those of the CSCW systems, but involving computer agents instead of human agents. CDPS systems are looking for a way to organize the cooperation between independent agents, designing agent's architecture in terms of the type of knowledge and functionalities necessary to assure cooperation. In particular, CDPS is interested in topics such as what models each agent must have of the other agents, the type of cooperation strategy which is most appropriate for each problem, and the mechanisms for generation and control of the dialogues between the different agents which are needed to coordinate the global solution of a problem.

3 A Metaphor for a Cooperation Network

In order to explain the general problem of the networks of open cooperative systems, let us make an analogy with the wayan office works. Let's suppose that the office is a sort of cooperative network formed by a group of characters which includes bosses and secretaries. Each boss has, as her associate, an executive secretary to whom she is hierarchically related. The available communication services for these two characters are: the telephone, the fax service and the messenger service. In the cooperation work between bosses their secretaries play a double role: 1) allowing the use of different communication services, and 2) helping to organize the work through the dialogues between all the secretaries. This metaphor is a reasonably good abstraction of reality as long as the services we have mentioned are in good running order. Inspired by the philosophy for the model for the interconnection of open systems, proposed by (ISO 1979), the model presented here is able to discriminate between at least three hierarchical levels of communication: level (n) bosses, level (n-l) secretaries, and level (n-2) communication services.

In addition, it is important to establish, as well, three different types of dialogues: {boss.n-boss.n}, {boss. n-secretary. n-l }, {secretary.n-l-secretary.n-l}.

These dialogues are carried out using the various communication services existing in the office. They in turn, can be described in terms of layers of dialogue which are controlled by protocols of levels below n-l, and services of lower levels.

Let's suppose communication with the outside is always established through one of the secretaries (there are no direct lines into the boss's offices, so to speak). The outside environment is identified with the client of the cooperation network. The client "tells" his/her problem to the network through 'secretary .n-l' who is associated with'boss.n'.

130 C. Lemaitre, V.G. Sanchez, c. Loyo

Exec-Sec

Dialogue

Executives Dialogue

Secretaries Dialogue

PBX Dialogue

:§~;9:

~elephone Networl<

~.il Network

Fig. 1. The office metaphor

Exec-5ec

Dialogue Secretaries Dialogue .......... ~ -

\----'-.-.., PBX Dialogue

7"-----.;:---' .¢~~.CS

If elephone Network

~ail Network

Fig. 2. The restricted office model

Exec-Sec

Dialogue

Exec-Sec

Dialogue

Cooperative Open Systems Architecture 131

This metaphor is very general so far, and lends itself to different computational models, the type depending upon the restrictions imposed, whether on the characters or on the dialogues which they can maintain with each other. For example, if we restrict the dialogue between the secretaries to simply establishing communication between their respective bosses, and the dialogue between the bosses is restricted to cooperation in the environment of the business in which they are situated, the system would be functioning like a CSCW for the bosses.

Let's suppose that the restrictions on our analogy are the following: 1. There is no direct dialogue between boss and boss. 2. Each boss can communicate only with his/her secretary, by means of orders,

questions and affirmations. 3. Dialogues between secretaries establish communication through the trans­

mission of requests, questions, assertions and transmission of documents. In view of these restrictions, each boss is conceived as an isolated specialist

whose only connection with the rest of the office is through his/her secretary, who bears the burden of cooperation. The secretary is familiar with the global environment, the different indirect interlocutors of the boss, has a mastery of the appropriate protocols for different matters, and knows both how to interpret the orders of his/her boss and how to break them down into tasks involving one or more interlocutors.

4 A Computational Model for a Cooperation Network

The restricted metaphor which we have presented in the previous section will serve as a basis for the definition of an architecture for a cooperative network of existing expert systems. Let us associate the role played by the bosses in the general metaphor to the expert systems distributed in an office. The problem we propose to solve consists in the design and construction of a system's network that played the role of executive secretary network. These systems are called facilitators because their purpose is to facilitate cooperation between expert systems which were constructed to work independently. The term facilitator has been employed in slightly different contexts: as communication facilitators in the case of Finin (Finin 1992), and as CSCW facilitators in the case of Cuena (Cuena 1992).

Before presenting an architecture for the model proposed, let us summarize our initial design premises:

1. The expert systems which are to cooperate are pre-existing. Therefore, in the majority of cases it is not possible to access source programs. Their inacces­sibility implies that the only modification which can be carried out will be at the level of the knowledge base, and even these must be minimal.

2. The incorporation of an expert system into a cooperation network must be user-friendly. In particular, the language of the human expert system interaction must be invariable. The objective of this invariability is to maintain the stability of use dynamics, avoiding the need for additional user-training.

132 C. Lemaitre, V.G. Sanchez, C. Loyo

3. The communication between the components of the cooperation network should be layered according to the same philosophy as the OSI protocol layers, and they should correspond to the upper application layers of this schema (ISO 1979).

4. Document flow should be minimized as far as possible, by taking maximum advantage of the facilities of the client-server systems. In particular, this point refers to the access to data bases. It is presupposed that all expert systems have access to all the data bases in the network, either directly or through a facilitator. This is a plausible hypothesis, given the variety of interoperability tools already on the market.

Data

Bases

=O~i:~it~ E,S,-FaFi~tor Dial~u~

acilitato '. -....... .

0=: E, S, -~a9i1itator

DialQglJe \/

.' acilitato

Facilitator - Facilitator Dialogue

Open Systems Network

Fig. 3. Cooperative network for expert systems

Data

Bases

The computational model of the expert system cooperative network is organized in three layers: The expert system's network, the facilitator's network, and the communication network.

The expert system's network is composed by a set of expert systems, that can solve together more complex problems than the ones they can solve on their own. We suppose that each expert system has an access to a data base management system. This cooperative problem solution is made possible with the aid of the facilitator's network.

The facilitator's network embraces a set of computer systems able to define and coordinate a cooperative solution of users' problems. The problem solution is obtained through dialogue interchanges, on the one hand, between the facilitators,

Cooperative Open Systems Architecture 133

the users, and the expert systems, and on the other hand, between the facilitators and, both the users and the expert systems.

The communication network corresponds to the computer's network and the physical communication layer, whose main tasks are, a) to assure the logical interconnection of the network of facilitators, b) to assure the message transmission service.

In terms of the layer schema in open systems, the users of a facilitator system are situated in the last layer of a cooperative open system, the layer of n applications, the facilitators are situated in the penultimate layer, n-l, and therefore constitute the communications service layer of the upper layer.

Each facilitator is associated locally by particular interfaces with a user, a data base management system and an expert system. Externally each facilitator is associated with the facilitator's network.

5 The Facilitator Architecture

Two of the three layers mentioned above, the expert systems network and the communication networks are the premises of the problem and the aim of our work is the middle layer, that is the facilitator's network. This section contains the basic ideas of the facilitator's architecture.

The main objective of the facilitator is to assure the coordination of the expert system attached to it with the other expert systems in the network, in order to assure the global solution of the user's problems.

The main tasks of the facilitator are the selection of the cooperation strategy to follow for the solution of a user problem, and the execution of that strategy which is meant to control the different types of dialogues involved in it. The modules in charge of these tasks are the Cooperation Control and the Dialogue Control modules.

The cooperation control module is the central module of the facilitator. Its first task, with a new user's problem, is to create the work group of agents available in the network, and after that to choose the cooperative problem solving strategy the group will follow, as well as the types of dialogues involved. This module must have knowledge of the overall domain of the expert system's network, metaknowledge about the different types of cooperative strategies, as well as a model of its expert system and of the other expert systems in the network.

The dialogue control module is responsible for the application of the selected strategy and the management of the corresponding dialogues. There are three different types of dialogue, the dialogue between a facilitator and its user, the dialogue between the facilitator and its expert system, and the dialogue with other facilitators in the network.

The main functions of the facilitator are the following: - The facilitator controls the user interface, without altering the syntax of the

language of communication of the original expert system. The objective of transferring the interface with the final user to the facilitator, is to allow it to

134 C. Lemaitre, V.G. Sanchez, C. Loyo

User Interface

(cooperation Control

E. s. Interface

( Dialogue Control

Communication Interface

Fig. 4. Facilitator's Architecture

)

) Inf.

System

Interface

assume full control of the dialogue between the user and the expert system. The design decision to include in the same facilitator the user interface and the expert system interface was taken in order to simplify the global design, based on the fact that in this first model the user do not playa crucial role in the cooperation work of the expert systems.

- The facilitator controls the interface with the expert system, discriminating between user-bound messages and network-bound messages.

- The facilitator controls the interface with the accessible institutional information systems for the cooperative network. This access is considered important to assure the interoperability of the entire system, regardless of the fact that each expert system has access facilities for its "personal" data bases.

- The facilitator controls the network interface with all the rest of the facilitators. This interface is, in fact, the facilitator's communication module, similar to that proposed by Huhns (Huhns 1990) and his collaborators.

- The facilitator contains enough knowledge about the network's global application domain to make correct interpretations of expert system orders and requirements in terms of dialogue plans to be carried out with other facilitators. In the case of the general model, a "boss" order which says "Get me the statements for our accounts in SIL VERBANK" is translated into a whole series of actions and secretary phone calls until he or she can finally obtain all the necessary information and present it in an appropriate form to hislher boss. The secretary, as we can see, has global knowledge of his/her environment and knows where and how to find the needed information, put it together, and give it to the boss.

- The facilitator has the metaknowledge which guarantees network-wide communication, as well as communication with the expert system and the user. Here we are dealing with the management of high-level protocols, and cooperation

Cooperative Open Systems Architecture 135

between different network components. The cooperation strategies, in this case, are provided by the organization into which the network expert systems are inserted by their initial definition.

- The facilitator has an expert system model in the same way as the rest of the network agents (facilitator + expert system). These models allow it to choose its interlocutors depending on the problem. They also enable it to answer questions about the type of problems its expert system can solve.

- The messages which one facilitator exchanges with another, as well as those received by its expert system, are handled in terms of speech acts. We consider three low level basic speech acts: questions, orders and assertions. These constitute the basic services provided by communication network layer n-2.

6 Modifications in the Expert System

One of the design premises is minimal modification of the expert systems. This is motivated by the need to take advantage of the work crystallized by the systems themselves. Besides, in most cases the source programs of the systems cannot be accessed.

The modifications that we carry out are twofold. The first consists in redirecting the input and output of the expert system toward the facilitator. The second modification involves locating all the expert's knowledge base rules which act to produce an output message for the user. The messages selected are those which, in the new context of the cooperation network, will be controlled and interpreted by the facilitator; they are then marked so that the interface with the facilitator expert system can distinguish them.

7 Example

The case study we are working on is, as previously mentioned, focussed on the granting of bank credits. For this purpose we consider three expert systems. An expert system for granting credits (SEl), an expert system which evaluates projects (SE2), and a calculating expert system (SE3). Each of these expert systems has a data base and reviews the users one at a time. In the particular case under study, we are only concerned with the user of the credit-granting expert system.

The principal characteristics of the three expert systems are the following: SEl: This system is a banking expert which grants different types of bank

credits. It has a number of bank policies which allow it to extend credit on the basis of criteria such as solvency, earnings for the bank if the credit is given, directive capacity of the individuals soliciting the credit for their company and technical and financial viability of the project. This system also has access to a data base which contains the application files of those who are requesting the loan, information about bank policies, current interest rates, etc.

136 C. Lemaitre, V.G. Sanchez, C. Loyo

SE2: This is a project-evaluating expert system. Its general purpose is to analyze and evaluate investment projects. It is in charge of rating the projects it evaluates according to the size of the project, the relevant costs, whether empirical or quantitative models are used, whether it is safe or risky, if all the information is available or not, if background information on the project is favorable, etc. It is interesting that, although this system has a user familiar with its functioning, in this example it will be used only as a server of the cooperation network.

SE3: This is an expert calculating system which can apply different types of models and carry out calculations of varying complexity. It also knows market behaviour and can access a data base containing all the information necessary to carry out these calculations. This system will be used as a server, just as in the case ofSE2.

There is a data base which contains general information on the company, its financial status and data referring to the project for which the credit application has been made. This is a data base which is not assigned to a single expert system, but rather is accessible to either one.

Although it is true that the minimal cooperation network is formed by two expert systems with their respective facilitators, the introduction of a third agent into the network--in this case the calculating system and its facilitator--enhances the capacity for dialogue significantly without notably increasing the complexity of the global problem, since the calculating system is relatively simple. In the following drawing we can see a schematic representation of a cooperation network developed for our case study.

Credit Assistant

ES1

ES = Expert System FAC = Facilitator

~ User

FAC3 ES3

Fig. 5. Credit assistant cooperative network

Project Evaluation Assistant

Computing Assistant

Cooperative Open Systems Architecture 137

8 Extensions of the Model to CSCW

The computational model for cooperative open systems restricted to the control of a cooperation network of existing expert systems can serve as a basis for the development of two additional types of cooperative networks. The first of these is one in which the expert systems are substituted by human agents in order to solve specific cooperation problems; the second might be described as a network which integrates those networks mentioned previously, the cooperation network formed by heterogeneous agents, whether humans or systems.

Before addressing the problem of the cooperation network of heterogeneous agents, let us examine that of the cooperation network formed by human agents. In the latter case, the facilitator cooperation network must assure that the network of human agents (i.e. the upper level network) acts as a cooperative network. This gives us an intelligent CSCW network, capable of facilitating cooperation between human agents by making coordination dialogues and minor problem­solution transparent for the users.

User-1

Fig. 6. CSCW extension

Just as in the case of the cooperative network of expert systems, the facilitators associated with human agents must have knowledge of the domain of work in which they are involved, and of the type of dialogues which can be established

138 C. Lemaitre, V.G. Sanchez, C. Loyo

with the other facilitators, as well as with their respective human agents. They must have a detailed model of the human agent they are attached to, as well as a less detailed model of the other network agents. These agent model must allow the facilitator to coordinate the working group and liberate the human agent of the bureaucratic overhead of the working session. Each agent model must deal with different roles as human agent can play, in order to facilitate the dialogue control within specific working group the human agent is participating in.

A necessary extension of the environment of human cooperative work is the need of the human agents to establish a direct dialogue between one another. The facilitator network must provide the necessary services to assure this type of communication. Galegher and Kraut (Galegher 1990) discuss two approaches to the control of dialogues between CSCW network users; the "prescriptive" and the "permissive" approaches. The first of these tends to impose a rigid group discussion model, while the second pretends neither to restrict nor to direct the behaviour of the human users.

Exec-Fac Dialogue

Permissive -;;;:;.::...;..,;;..~ ..........

Facilitator

Executives Dialogue

c=::>~ ~ c==:>

Permissive

FaCilitators Dialogue

Exec-Fac Dialogue

........... +-~ .... ... c==:> --.. ---+- c=.? .' Facilitator

Prescriptive

Open System Network

Fig. 7. Type of dialogues in the CSCW extension

Seen from this viewpoint, the proposal for a facilitator cooperative network use of CSCW allows us to divide different types of dialogue among the various components of the network. The more prescriptive dialogues are thus established between facilitators themselves and between the facilitators and their human agents; on the other hand, dialogues between human agents will be more permissive.

Cooperative Open Systems Architecture 139

9 Related Work

In the specialized literature of the past few years, there have been important contributions on different aspects of DAI and CSCW. Below we discuss some of the research related in significant ways to our own.

In relation to the connection of knowledge based systems, we hold to the general proposal of Huhns and his collaborators (Huhns 1990), as was mentioned in the previous section, as well as in the problem of the transmission of knowledge and protocols by layers of KQML following the philosophy of OSI reported by Finin et al (Finin 1992).

Regarding the functionality of the cooperative agents, we retained some of the proposals Gasser (Gasser 1990) and Shoham (Shoham 1990). These include the definition of the Gasser agent, the management of interlocutory models and the agent itself as dealt with by both authors, as well as the idea of mental state and control of the behavior of the Shoham agent. At the architectural level there exist important differences of opinion between our proposal and those of these authors. Insofar as our proposal implies a clear distinction between the specialized problem solving aspect (the expert system), and the cooperation and communication aspect (cooperation module).

Regarding the actual communication of the agents, we consider salient the proposals of communication acts in general and speech acts in particular. Many authors (Bouron 1991, Campbell 1990, Cohen 1985, Cuena 1992) have made interesting proposals on this point; however, we hold most closely to the work of Cohen and Perrault which puts forth the idea of speech act planning as a means of behavioral control of agents, and that of Bouron, which proposes a hierarchy composed of message, dialogue and discourse.

The general proposal of cooperative open systems, as a result of the confluence of work in the areas of DAI and CSCW, coincides to a considerable degree with work done within on the ESPRIT 5362 project IMAGINE (Lux 1993, Steiner 1990). On this project, Steiner and his collaborators propose the construction of heterogeneous cooperative networks where human agents as well as computer agents are associated in communication, cooperation problem solving and task execution, thus extending the conventional paradigm of CSCW to a new paradigm of human/computer cooperative work (HCCW). The architecture which we propose is somewhat similar to the model proposed by these authors. Its agent is composed of three modules: mouth, head and body. The mouth is in control of communication with the other agents; the head controls cooperation; and the body enables the agent to solve problems by itself. The main difference, regarding its architecture, with our proposal consists in the fact that our agent is divided in two: the facilitator which corresponds essentially to the head and the mouth of the IMAGINE agent, and the user agent, since this is an expert system or a human being, which would correspond to the body of the IMAGINE agent. The IMAGINE project attacks the general problem of the heterogeneous agent cooperation network. For us, this is an objective left for solution at a future time.

140 C. Lemaitre, V.G. Sanchez, C. Loyo

10 Conclusions

The computer model we have presented of the cooperative open system archi­tecture based on the concept of the facilitator, involves a promising methodology for the modelling of cooperative work with heterogeneous agents as part of the general problem of distributed cooperative systems.

The problem of linking expert systems for cooperative tasks is a fertile field for experimentation, one in which clear-cut problems demand effective solutions if in the future we expect to achieve cooperative human-AI networks which are truly heterogeneous.

The applicability of the concepts of the cooperative open system, and the facilitator goes beyond the communication of existing expert systems, and may be extended to auxiliary systems, as well as to general interoperability problems which require solutions at the semantic level.

Acknowledgements

Research support was provided by grant 0711-A9111 from Mexico's National Council of Science and Technology, CONACYT. Special Thanks to Cora B. Excelente for her comments on earlier drafts of this paper.

References

Bouron, T. (1991) What architecture for communications among computational agents? In Proceedings of the Workshop on Cooperation Among Heterogeneous Agents, AAAI

Campbell. J.A .• D'Inverno. M.P.(1990) Knowledge Interchange Protocols. In Descentralized A.I. Demazeau. Y .• Muller. J.P .• (Eds.) Elsevier Science Publishers

Cohen. P.R., Perrault, C.R.(1985) Elements of Plan-Based Theory of Speech Acts. In Readings in Distributed Artificial Intelligence. Bond, A.H.. Gasser, L. (Eds). Morgan Kaufmann

Cuena. J., Garcia Serrano. A.(l992) Sistemas Basados en el Conocimiento de Trabajo Cooperativo. In 3er. Congreso Iberamericano de Inteligencia Artifical. La Habana. Cuba

Finin. T., McKay. D .• Fritzson, R.(l992) An Overview of KQLM: A Knowledge Query and Manipulation Language, DRAFT

Galegher, J .• Kraut, R.E. (1990) Technology for Intellectual Teamwork; Perspectives on Research and Design In Galegher. J .• Kraut. R.E .• Egido. C. (eds): Intellectual teamwork. Social and Technological Foundations of Cooperative Work. LEA Publishers

Gasser. Le Braganza, C.; Herman. N. (1990) MACE: A flexible testbed for distributed AI research. In: Distributed Artificial Intelligence. Huhns, Michael N. (Ed.) Morgan Kaufmann Publishers

Huhns. M.N .• Bridgeland. D.M .• Ami, N.V. (1990) A DAI Communications Aide. In: Distributed Artificial Intelligence, Huhns. Michael N. (Ed.) Morgan Kaufmann

ISO International Organization for Standarization: "Reference Model of Open Systems Interconnection ISOrrC97/SC16" (1979

Cooperati ve Open Systems Architecture 141

Lemaitre, C. (1989) SISCO: Sistema Semantico Conceptual. In Memorias de la VI Reuni6n de Inteligencia Artificial, Queretaro, Mexico

Lux, A., de Greef P., Bomarius, F., Steiner, D. (1993) A generic Framework for Human Computer Cooperation. In First Conference on Intelligent Cooperative Information Systems, North-Holland

Searle, lR. (1969) Speech Acts: An Essay in the Philosophy of Language. Cambridge University Press

Shoham, Y: (1990) Agent-Oriented Programming, Technical Report, ST AN-CS-1335-90, Robotics Laboratory, Computer Science Department, Stanford University

Steiner, D., Mahling., D., Haugenerder, H. (1990) Human Computer Cooperative Work. In: Proceedings of the 10th International Workshop on Distributed Artificial Intelligence, MCC Technical Report ACT-AI-355-90, AustinffX

12 The Role of Knowledge Based Systems for Automatic Coordination in Distance Learning*

Jose Cuenai , Ana GarcIa-Serrano i and M. Felisa Verdej02

i Departamento de Inteligencia Artificial, Universidad Politecnica de Madrid, Campus de Montegancedo sin, E-28660 Boadilla del Monte, Madrid

2 Departamento de Ingenieria Electrica, Electr6nica y de Control, UNED, Escuela Tecnica Superior de Ingenieros Industriales, Apdo, E-60149, Madrid

Abstract. In this paper we will focus on one aspect, the organizational model of distance education. Our attempt is on the one hand to improve the interaction between the participants - teachers, tutors and students - and on the other hand to support actively some group tasks in distance learning settings. This is the main motivation, but we aim also at exploring and characterizing the underlying generic tasks involved (either individual or group), the degree of automatic support that can be provided in terms of the knowledge required, and the viability of a con­ceptual model and its scope. A proposal for coordination modeling based on a knowledge-based approach is presented. This approach is particularly appropriate for a distance learning environment because contents are highly revisable with time, allowing the system to be adapted to the experience and criteria modification of the coordinated group and because explanation is a key feature both in learning and collaboration activities. The long-term goal is to create a framework where cognitive architectures for active group support could be defined.

Keywords. Coordination in distance learning, computer supported cooperative work, knowledge based methodology, knowledge based systems

1 Introduction

A distance learning environment requires the provision of high quality services to students in two aspects: scientific level and minimal delay in reponses. Frequently a large number of students use these type of services and the available number of tutors is comparatively scarce.

Information technology may help to improve the operation of tutor teams by simplifying their coordination problems. Current cooperative work and groupware products are, in principle, useful for distance learning institutions. Thus, from the point of view of organization they share common problems with organizations in other fields of activity.

* Research supported by the CICYT (Comisi6n Interministerial de Ciencia y Tecnologfa)

The Role of Knowledge Based Systems for Automatic Coordination 143

Existing technology for cooperative work (e-mail communication, distributed databases) can be classified as passive support, in the sense that the system allows data sharing or message transfer, providing control of the operation at a low level, as for example verification of the access or communication permission.

Active support is an innovative mode of group support where the system has some knowledge about the global task carried out by the group in such a way that the system itself participates in the task, for instance evaluating the current state, proposing new steps to be performed in order to reach the goal or verifying some operation features. Systems for active support are beginning to appear (Greif 88), (Greenberg 91), (Cuena Garda-Serrano 93).

In this paper we will focus on one aspect, the organizational model of distance education. Our attempt is on the one hand to improve the interaction between the participants: teachers, tutors and students and on the other hand to support actively some group tasks in distance learning settings. This is the main motivation, but we aim also at exploring and characterizing the underlying generic tasks involved - either individual or group -, the degree of automatic support that can be provided in terms of the knowledge required, and the viability of a conceptual model and its scope. The long-term goal is to create a framework where cognitive architectures for active group support could be defined.

We follow a bottom-up approach, starting from the concrete to proceed later to a more abstract level. In this way, our first step was to identify relevant activities in a distance education setting and to discuss the kind of facilities that a computer-mediated environment would offer.

As a result two activities were selected: academic consulting and structured document production. They are described in Section 2. A knowledge-based architecture for active support is proposed in Section 3 where models for the two case studies are also given.

2 Distance Learning for Higher Education

Distance learning for university studies is a facility currently offered by a number of institutions in different countries. Although the staff and resource organization differ one from another, most of them started from the principle of stand-alone study and therefore a main activity of the institution is to create and deliver course material to support the individual learning.

The demand of distance learning for university studies has increased in a dramatic way. For instance the U.N.E.D with 130.000 students is the second largest university in Spain. Human resources, particularly professors and tutors have not followed the same rate of growth. As an example, for some undergraduate courses the ratio is 11500, but apparently the one-to-many traditional technology is robust enough to support this evolution.

The development of computer mediated communication, offering the possibility of either group or individual interactive distance teaching and learning is widely recognized as a significant change for distance-learning settings.

However one has to be aware that conventional computer telecommunication technology will not necessarily succeed in situations as described above. On the contrary, will probably produce an unmanageable overload of information.

144 J. euena, A. Garda-Serrano, M.F. Verdejo

Nevertheless, there are possibilities to promote in a realistic way interactive communication and group activity in such distance-learning settings.

Different works related to cooperation and/or collaboration in learning, tutoring and authoring activities are reported in the literature. As pointed out (Falcone 90) the technologies identified for group work are:

- Editing systems as for cooperative courseware production, - Electronic mail, - Conferencing systems as Thought Box or the Andrew System, and - Distributed databases. Our approach is to focus on some activities and, relying on a work

reorganization, design, implement and evaluate a software system providing support, at different levels, for task coordination and cooperation. In the following we propose a computer scenario and next we describe two case studies.

2.1 A Proposed Computer-Supported Scenario

Let us consider a distance-learning scenario including two or three teachers, a small group of tutors and a large number of learners. Teachers are centrally located and they are responsible for the course design and the final assessment of students. Additionally, they perform mainly two kind of activities: those related directly to support the student such as course delivery, assessment of the students work, individual consulting etc., and those aiming at coordinating and organizing tutors and teachers work. Tutors are geographically distributed in study centres, and they provide lectures or problem-solving activities on an individual or small­group basis. There is no communication between tutors and currently it is the case that most of the students never consult either their tutor or their teachers.

Let us assume that there exists an electronic network connecting all of them, providing the means to carry out one-to-one, one-to-many and group communication. With the current organization, the communication would be many-to-one overloading the teachers instead of facilitating their task. The alternative of increasing human resources is excluded, so the approach should be based on two criteria: sharing tasks and responsibilities - i.e. collaboration - and, appropriate software for active group support.

2.2 Description of Two Case Studies

The two initial cases concern academic consulting and co-editing of short papers such as exams or information notes. A description of each of them is given below. - Individual consulting through a communication system able to manage dia­

logues. A dialogue consists of sequences of structured messages involving two or more participants. The system, using criteria previously defined and known by the participants, will classify, automatically, messages and assign resources (human or software) for treating them. For instance it should be possible to answer automatically certain kind of requests such as delivery of course material or corrected assignments, solving typical questions, and others. This automatic treatment not only can eliminate a considerable amount of routine work,

The Role of Knowledge Based Systems for Automatic Coordination 145

improving tutor and teacher availability but it also enhances the time response to student requests.

- Co-authoring of structured material, linked with procedures to support discus­sion and negotiated decisions. An example is a group of teachers preparing the text of an examination for a particular subject-matter. A typical procedure would be first to fix requirements, such as number, type, importance, or focus of topics to be asked and, once an agreement is reached, to distribute among teachers the work of writing proposals fot the questions, exercises or problems, then comments on partial or complete proposals would be distributed and finally the final exam version would be produced. The exam case study has been selected for its simplicity to illustrate the knowledge based approach to software support design on a co-authoring scenario. A similar approach may be applied in a more relevant case from the point of view of education.

The system should offer facilities for supporting group activities at two levels: • The object-level where operations such as define structure and requirements

for the object to be built, create a part of the object, integrate parts, verify requirements, compare alternatives for a subpart, ... should be provided.

• The negotiation level where participants make proposals, arguments, justi­fications, decisions, ... about elements or tasks related to the object-level.

3 The Design of an Intelligent Coordination System

An organization can be described in terms of a set of human agents with some common goals. Agents have their own problem-solving abilities and, as part of an organization, established ways to interact - operation protocols - in order to attain the organization goals. These operation protocols are known by the agents involved in a particular task.

An intelligent coordination system for an organization should be able to monitor the group when performing a previously established activity plan, taking care about the current task state either to validate it or to propose modifications to the group in their operating procedure.

This kind of coordination can be carried out if the system: 1. Has a model of the organization in terms of goals, agents, and their possible

group plans, 2. Is able to understand the operation protocols carried out by a group of agents

(while solving an organizational goal), i.e. there is a predefined language for structured messages to communicate with the system or through the system to other agent, and

3. Is able to act according to its knowledge and the current situation. This idea of operation protocols allows a global vision and justification of the

different operations carried out by a group when solving a task. This is a new feature in computer support tools for organizations.

146 J. euena, A. Garcia-Serrano, M.P. Verdejo

3.1 A knowledge Based Architecture

Computer support in our proposal implies a system monitoring the transfer of information between agents engaged in a common task. The degree of manage­ment performed by the system will depend on the amount of knowledge available: knowledge about the group goal, the agents performance and the agreed protocol of cooperation among agents to develop an activity.

The design should take into account the following requirements: • The system operation must be explanatory, i.e. whenever the systems leads a

discussion, detects inconsistencies or proposes an action it should offer justifications for its proposals.

• The system must be easily revisable because its knowledge base represents criteria for coordination previously agreed by the coordinated agents. So it must be easy to modify these coordination criteria according to the group experience.

Thus a knowledge based architecture seems an adequate approach. In early knowledge-based systems a uniform representation was applied based either on rules or frames. But a coordination model may involve knowledge of different types cooperating to produce the specified output. The proposed approach to design such a type of model requires a first step of knowledge level definition (Newell 82) of the general characteristics of an intelligent agent:

(a) What are its relations with the environment? (b) Which knowledge elements are to be used? (c) How the knowledge elements are used to meet the agent goals.

Once this specification is clear (text may be used for its characterization) there is a knowledge modularization step. An adequate concept for this modularization was proposed by Chandrasekaran: the generic task concept (Chandrasekaran 83,86,87). He introduced the concept of generic task as an alternative to the uniform paradigms of reasoning proposed in the early eighties. The proposal of Chandrasekaran suggested the design of knowledge units to encapsulate typical reasoning tasks named generic tasks. Such knowledge entities may be defined at the knowledge level in terms of:

• The premises for the task to be performed. • The goals of the task: the set of concepts that may be obtained as result of

the task reasoning. • The knowledge elements of the task (i.e. other components knowledge

entities ), several may be used to attain the different possible task goals. • The inference methods that are capable of using the different knowledge

components to obtain the possible values of the goals. So the knowledge level specification may be organized as a collection of

generic task specifications where the knowledge level concept (a) specifies the task premises and goals, (b) specifies the task knowledge bases that may include other generic tasks and (c) specifies the set of inference procedures to provide the different answers using the knowledge components.

A next step is required to define the knowledge level in terms of computable knowledge representations (such as rules, frames or constraints). This layer may be named symbolic level definition. Several approaches may be considered for the symbolic level, some may be more understandable by the user but less efficient

The Role of Knowledge Based Systems for Automatic Coordination 147

from the computational point of view. However the approach to symbolic modelling must take into account the differences among the knowledge elements identified at the knowledge level so that knowledge modules may be defined embodying different knowledge formulations.

The symbolic level model specification, (Cuena 93) may be expressed based on a knowledge unit structure, which has the format of Fig. I, where knowledge components and inference procedures are displayed. This format filled with a natural language description for both components of every knowledge unit, knowledge and procedures,and expanded at different levels of refinement can be also used as an expression of the result of the knowledge level analysis.

At the symbolic level, every inference procedure, "inf.proc.i", may be defined by several inference steps, by calling inference procedures defined for the components KBj. Every KBi may be an instance of a predefined knowledge unit or may be a basic knowledge unit formulated in term of rules, frames or constraints. Every basic knowledge component in general will be described by a conceptual framework (Le. the concepts, attribute, values or predicates used to describe the knowledge) and the base formulation. It could be considered as an alternative to the definition of a general conceptual language for the knowledge unit to be used by its knowledge components.

Premises Conditions

Knowledge Inference

I I : KBI I inf.proc.1

I KB2 I I inf.proc.2 I I KB II--c_o_n_ce_p_;_:_fr_am_e __ ....

I KBn I I inf.proc.k I Conclusions

Fig. 1. The general structure of a knowledge unit specification

There are two types of tasks: the subproblem tasks to obtain some intermediate goals required by the main task and the control tasks to guide the search of the inference procedures of the main task. A control task receives as input, answers from different subtasks and generates, by combination or selection, answer scenarios to be the input of other subtasks. The basic tasks are described by rules, constraints or frames with its typical inference procedures.

According to Fig. 1, every knowledge unit is described by the premise conditions to be acceptable, the knowledge specification or conceptual framework (set of predicates and associated domains for its arguments and the set of basic knowledge units identification to be used in the task), the knowledge formulation

148 J. Cuena, A. Garcia-Serrano, M.P. Verdejo

in terms of rules, constraint or frame bases and finally the inference formulation describing the different inference procedures for the knowledge unit. The use of a knowledge unit may have the following format:

name-of-unit(premise-names,inference-procedure(parameters), conclusion-names)

The symbolic level specification must be further described at the implemen­tation level. Computable entities must perform the knowledge management and reasoning functions required to operate the symbolic level. For this implemen­tation level different alternatives can be considered such as logic or object oriented programing.

Alternative formulations could be considered such as the premise conditions distributed to the different inference procedures as specific preconditions for every one.

3.2 Two Case Studies

Let us assume that there is a platform of communications (such as e-mail) and data base sharing. On top of it academic consulting or co-editing of short and structured texts are the activities to be supported by the coordination system. First, a tool providing structured communication has to be defined (Verdejo, Abad 91). In any message there is a part to be understandable by the system and a textual part only by humans. From this passive support the design of the coordination system begins as an ideal intelligent agent. The design will be done in three levels: the knowledge level, the symbolic level and the implementation level.

At the knowledge level the designer describes the organization of the ideal agent in terms of what it knows and how it uses this knowledge to meet its goals.

3.2.1 Academic Consulting

A) Knowledge Level The coordination agent for academic consulting management must be able to perform the following operations, where goals, procedures and knowledge components are identified: - Question Classification

The first step is to classify the question. This requires a knowledge base where the criteria for classifying structured messages are represented.

- Management of a question This operation goes as follows:

*If the question can be answered in an automatic way then a message to the automatic source is produced.

*If the question is classified as requiring human analysis a selection of tutors must be made in order to forward the question to an appropriate human resource. The system supports this resource assignment according to different features which change with time. The mode of operation is as follows:

The Role of Knowledge Based Systems for Automatic Coordination 149

1. Identification and selection of the set of available tutors, according to their capacities and availability as described in the tutor knowledge base and the current state data base respectively.

2. Request an answer from the selected tutors. 3. Answers received.

- Decision oj an answer In general, different potential answers can be produced, so a last step is to evaluate them in order to give a final consistent answer to the student.

Given the classified question message, either a tutor's or automatic answer, the operation mode for selection follows:

1. Evaluate the different answers, taking into account the question. 2. If inconsistencies are identified then keep the consistent ones as

acceptable and reject the others. 3. If there are acceptable answers then a message with the question and

answers are sent to the subject coordinating person who decides which final answer message should be sent.

B) Symbolic Level From this natural language description of the academic consulting task --without symbolic or implementation references, the knowledge units and inference procedures identified as components of the task to be performed are:

- Four knowledge units: * A message classification unit. * An automatic answer unit responsible for the formulation of the questions

to the automatic sources. * A Tutor management unit responsible for selecting potential human

resources to answer the question. * An answer consistency unit.

- Three inference procedures: one for the question classification, another for question management and the third one to decide on the final answer to the student.

The structure is summarised in Fig. 2 where the different types of input and output messages, the knowledge units and the inference modes are shown. For illustration purposes, some details of the symbolic level description of this task are commented on next. The main task in our case study is the academic consult­ing task. The inference procedure of this main task is a sequence of subtask calls:

question classification( question, jorward(parameters), classified question)

question management ((classified question, tutors current state), injerence-p(parameters), answer message)

answer consistence( answer message, injerence-p(parameters), answer)

The message classification unit receives the structured student question as input premises and generates as a conclusion the type of the message. The classification task may be formulated by applying a rule base to attribute-object­value triples or by using a frame hierarchy with an inference procedure based on pattern matching and hierarchy exploration. Several possible classifications of a

150 1. euena, A. Garda-Serrano, M.F. Verdejo

Premises Question characteristics Educators answers

Knowledge Inference

Message classification unit Question

classification Automatic

answer unit Question management

Tutors manage-ment unit Answer

decision Answer consis-

tence unit

Conclusion Final answers

Fig. 2. Knowledge structure for the academic knowledge unit

question may be obtained (as the knowledge for classification may be incomplete). This design of the classification activity allows for the customization in different scenarios, so the message classification unit is a generic task to be instantiated in every case.

The tutor management unit has as input premise the classified question and the tutors situation data and, as a conclusion, the message answer. It may be formed from two types of knowledge, one for modelling the tutors capabilities (for answering a given question) and another to represent the criteria to assign questions to tutors.

The first class of knowledge may be formulated by frames about tutors and the second one may be formulated by a rule base to infer the degree of adequacy of every tutor to answer a given question. As in the first case, several assignments may be generated if the knowledge available is not discriminatory enough.

3.2.2 Co-editing

In the case of co-editing, the operation goes as follows: first, a phase of requirements specification should be carried out. We call this task document format definition. The goal is to fix aspects such as number, type, and topic of either questions, exercises or problems, their degree of difficulty, etc ... Once the format is established, then the work of writing parts according to the format has to be distributed i.e. a decision about who has to write what has to be taken. After proposals are written and communicated to the group, a further step consists of gathering alternatives, looking for different compatible partial versions. Then,

The Role of Knowledge Based Systems for Automatic Coordination 151

alternatives should be evaluated and selected in order to compose the final document for examination. All these operations should be performed in a collaborative framework, supported by an intelligent agent for coordination (facilitator).

Let us describe in more detail the functionalities provided by the facilitator in order to carry out the co-editing activities. Our scenario includes a group of teachers involved in assessing students following a particular course. Usually one of the teachers assumes the responsibilities of group coordination in order to produce an examination text in due time.

A) Knowledge Level The intelligent agent, at this level, must be able to perform the following operations:

Document format generation *The responsible person starts the co-editing process by sending a message to

the other teachers/authors to propose an initial exam format. Reactions to this proposal are then received.

* A group discussion to reach an agreement should then be managed. Here the facilitator can 1. organize and structure comments and modifications received from the

group members to design several constraints to the exam format. 2. evaluate arguments in a qualitative way so as to reach a final decision.

Document part content production *Once an exam format is agreed a negotiation step is carried out to distribute

the work of producing the different exercises and document parts, under the restrictions given by the exam format. The facilitator acts here as a kind of secretary for the coordinator, for instance annotating volunteers, verifying that the workload is fully covered or proposing names for pending parts and different persons to submit proposals for specific exercises, problems or questions.

* A discussion step is started where proposals are evaluated with the same approach as that proposed in the format exam consensus. As a result, for every component of the exam format an ordered list of specific candidates is produced.

Document composition *Finally, partial proposals have to be selected in order to compose the final

exam. This is a decision process similar to the format exam consensus decision, where the responsible person proposes candidates and the facilitator controls the group decision by evaluating annotations provided by the rest of the group.

B) Symbolic Level The structure is summarized in Fig. 3 where the different types of input and output messages, the knowledge components and the operation procedures are shown. The structure of the knowledge unit to be formulated at the symbolic level is based on five knowledge units:

- Format document negotiation unit - Format document consensus unit

152 J. Cuena, A. Garcfa-Serrano, M.F. Verdejo

- Authors assignment negotiation unit - Document part content consensus unit - Total document content consensus unit These five units involve inference procedures of two types: consensus and

negotiation. Therefore two kinds of knowledge must be modelled: knowledge for two types of decision making, both negotiation and consensual, and the specific domain knowledge, exam production in this case. The first type of knowledge can be formulated as a generic task. Next we elaborate on this aspect.

Premise exam request

Knowledge Inference

format document negotiation unit

format document document format

consensus unit generation

educators assign. document part negotiation unit content production

document part con- document I tent consensus unit composition

total document con-tent consensus unit

Conclusion final exam

Fig. 3. Knowledge structure for the exam co-edition knowledge unit

Negotiation is a way of cooperation that allows decision making by conflict resolution (euena, Garcia-Serrano 93). The negotiator facilitator has knowledge about the goal to decide, the kind of agents proposals (in a structured language) and about the possible conflict situations in order to make arguments to solve conflicts. The negotiation operation may be reactive to the changing features and make predictions by evaluating the different proposals in every step. The general human-computer operation procedure is as follows:

1. Given a set of agents to take place in the negotiation, teachers in the educators assignment negotiation, there is a step of distribution of the initial proposal. In the example the initial proposal is an initial format exam.

2. The different agents submit proposals for execution of some subtasks among the proposed. In the example proposals are about who will write which exercises. A proposal has three parts:

The Role of Knowledge Based Systems for Automatic Coordination 153

(P) the set of preconditions, (PS) the set of postconditions and (1) the text explaining the task content.

In our example (P) may be about exercise format conditions and (PS) the requirements verified by the exercise content described in (1).

3. The negotiator facilitator, using its knowledge, identifies the degree of satisfaction of every proposal, reacting in one of the following ways: - Call for proposals to fill the gaps. In the example the gaps could be

exercises not yet submitted from any teacher. - Ask for agreement from the agents for changes in the proposal. In this

case the facilitator could make an initial argument by applying predefined rules for conflict management.

- Decide to cancel a proposal (if no resolvable inconsistencies are detected). For example if there are more teachers to fill an exercise than the number allowed or because the initial format requiremen~ for the exercise are not accomplished.

- Decide to accept a proposal that meets the goal specifications. 4. The agents submit proposals to fill the gaps. 5. The facilitator reacts in the same way until the goal is reached (all the

exercises are assigned to a sufficient number of teachers).

Premises agent proposals goal constraints

Knowledge Inference

I argumentation unit

I proposal satis- negotiation faction unit manager

I proposal deci-sion unit

Conclusions final decisions proposals

Fig. 4. Knowledge structure for the negotiator facilitator

For design purposes it could be decided that agent intentions be represented by a preference list, conflicts and inconsistencies detected by constraint satisfaction according to the current educators state and the exercise constraints (identified in the format exam).

In the case of consensual decision making there is a knowledge unit that facilitates the cooperation to select an alternative among a given set of

154 J. Cuena, A. Garda-Serrano, M.F. Verdejo

alternatives. As described in Sybil (Lee, in Greif 1990), the facilitator may evaluate arguments in a qualitative way so as to reach a final decision.

The Sybil reasoning system is based mainly on the Mycin paradigm for evidence accumulation, but evidence here is linked to proposals of group members instead of facts or rules. When a goal is proposed with several alternative values, participants suggest possible subgoals of this main goal (based on alternative goal decomposition, or evaluations from different participants). Evidence is accumu­lated for these subgoals too, in such a way that the system keeps track of the tree of goals and subgoals and the different claims for and against, in order to propose a final balance of the different positions.

The operation mode of the facilitator goes as follows: 1. Given a set of agents to participate in the consensual decision making the

facilitator distributes a message with the different alternatives (with a set of qualitative or quantitative constraints).

2. Each agent answers with a message that contains the alternative position (positive or negative).

3. The facilitator uses a knowledge base to combine the different messages and proposes an ordered list of alternatives (for example the first three alternatives).

4. The previous steps are repeated until equilibrium is reached. The structure is summarized in Fig. 5.

set of agents Premises initial alternatives

goal constraints

Knowledge Inference

KB

conceptual frame

proposal combi- consensus nation rule base formation

message unders-tanding unit

Conclusion final decision

Fig. 5 Knowledge structure for ·the consensus formation facilitator

These two generic tasks, negotiation and consensus formation, will generate specific decision making tasks by instantiation. The instantiation must be done by introducing the following elements:

The Role of Knowledge Based Systems for Automatic Coordination 155

- The conceptual framework where the set of facts to be analyzed are syntac­tically predefined by predicates, frames or attribute-object-value triples. The rule base expressing criteria for combining the different proposals to obtain the resulting decisions.

4 Conclusions

A proposal for coordination modelling based on a knowledge-based approach has been presented. This approach is particularly appropriate for a distance learning envinlnrnentbecause:

1. Contents are highly revisable with time, allowing the system to be adapted to the experience and criteria modification of the coordinated group.

2. Explanation is a key feature both in learning and collaboration activities. As stated in the introduction, our goal is twofold: to improve group activity in

distance learning and to create a framework to define a cognitive architecture for active group support software design. A project is now on course to evaluate and refine these ideas.

References

Chandrasekaran, B. (1983) Towards a Taxonomy of Problem Solving Types. AI Magazine 4( 1) 9-17

Chandrasekaran, B. (1986) Generic Tasks in Knowledge Based Reasoning: High Level Building Blocks for Expert Systems Design, IEEE Expert

Chandrasekaran, B., Tood R., Smith J. (1992) Task Structure Analysis for Knowledge Modeling. Comm. of the ACM 35(9) 124-137

Cuena, J.(1993) Contributions to a Knowledge Oriented View of Software Design. In Knowledge Oriented Software Design (J. Cuena, Ed.). pp. 51-76. Elsevier

Cuena, J., Garcia-Serrano, A. (1993) Intelligent Computer Support. Chap. 5. In Cooperation among Organizations: The Potential of Computer Supported Coopemtive Work (R. Power, Ed.). pp. 72-102. Springer-Verlag

Falcone, C. (1990) Groupware for Educational Environments. Report WP 22. Delta Project 7002

Greenberg, S. (1991) Computer Supported Cooperative Work and Groupware. Chap. 1 (S. Greenberg, Ed.). Academic Press

Greif, I. (1988) Computer-Supported Cooperative Work: A Book of Readings (Irene Greif, Ed.). Morgan Kaufmann Publishers

Newell, A. (1982) The Knowledge Level. In Artificial Intelligence. Vol 18. pp. 87-127 Verdejo, M.F., Abad, M.T. (1991) Human-Human Communication in an Open Distance

Learning Environment. Proc of the 8th International Conference on Technology and Education

13 Supporting Collaborative Dialogues in Distance Learning

Carla Simone

Dipartimento di Scienze del1'Informazione, Universita di Milano, Via Comelico 39, 1-20135 Milano, Italy E-mail: simone@hermes.unimi.it

Abstract. Distance Learning is an activity showing different modalities of cooperation among teachers and students. Moreover, a relevant part of this activity is devoted to the organizational aspects supporting the main mission of teaching and learning. CSCW technologies can provide suitable tools for many Distance Learning activities. The paper presents a tool supporting collaborative dialogues through the exchange of semistructured messages and describes how the provided functionalities can be exploited in various activities required by Distance Learning. The main focus of the paper is on functionalities devoted, on the one hand, to the automatic management of parts of the dialogues, and on the other hand, to the maintenance of the organizational and cognitive contexts in which collaborative dialogues occur.

Keywords. Collaborative dialogues, CSCW, semistructured messages, user models

1 Introduction

In attempting to support Distance Learning by computer-based technologies one can try to simulate the framework of a traditional classroom, i.e., implement the co-presence of teachers and students in a virtual space, or try to deal with the peculiar case of a physical distance between them to identify functionalities suitable to support this 'loose' form of co-operation. Both approaches are sensible and can be taken into account in relation to the teaching activity under concern or to the capabilities provided by the available technology. In this paper the focus is on the second approach as we are interested in designing technological supports for the 'loose' cooperation that can be eventually integrated with the real-time technologies exploited by the ftrst approach.

In the framework of DL loose co-operation is determined by the following scenario:

- teachers and students are, in general, neither physically nor temporarily co­present; - teachers are working with many students and students are working with different teachers; - students meet teachers (and vice-versa) in isolation, i.e., not in a classroom, as it occurs in traditional learning, but for occasional or speciftc circumstances;

Supporting Collaborative Dialogues in Distance Learning 157

- the notion of class almost disappears, as there are no collective, interactive lectures; - the relationship between a teacher and one of hislher students is on the one hand very distributed in time and on the other very personalized, as each individual student can be considered as a unique individual, much more than in ordinary learning.

From the standpoint described above DL is not just an ordinary learning with some adjustment; on the contrary, it raises problems of a very different nature. First of all, communication among teachers and students is mainly asynchronous and, in general, it is not taking place within a learning/teaching process that is shared among all students. Hence, the fundamental part of the learning process consisting of "learning from the experience of the others" is very difficult to preserve and maintain in a DL setting.

Secondly, the learning/teaching processes require a high degree of self­organisation particularly for teachers but also for students. Both of them have to keep control over a relevant number of on-going activities (concerning different teachers and students, respectively) that are potentially very different. For example, different students can have reached different steps in the same course programme or be engaged in different syllabi within the same course, and different teachers can have different co-operation strategies with their students.

To sum up, DL has not only to deal with problems typical to traditional learning (such as, for instance, how a specific topic fits with the individual student's prerequisites, if the related notes and learning materials are or not of good quality, if the evaluation methods are adequate or not with respect to the learning objectives, and the like), but also with the complexity of a quite specific type of co-operation between teachers and students. In fact, we do believe that the best possible solution of the first type of problems does not avoid facing the latter: on the contrary, stimulating student learning and interests triggers a greater need for interaction between them and their teachers. In addition, the possibility of asynchronous interaction is one of the more fruitful features of DL.

The question we want to consider here is how co-operation technologies (the so called CSCW technologies) can help in solving the problems mentioned above. More specifically, we want to consider those technologies supporting the dialogue between co-operating people through the exchange of written messages. This choice is due, first of all, to the possibility of processing message contents provided by this kind of interaction, secondly, to our experience in developing one of these technologies, and finally to the fact that if and when other kind of communication supports will provide the same capability the approach we propose will be applicable to the new situation too.

The various teaching activities generate collaborative dialogues needed for different purposes: negotiating course contents and plans, tutoring and advising students, asking for information about the courses, checking and monitoring the student progress, and so on. The network of dialogues between students and teachers is very complex and covers different topics. Therefore, the design of an adequate support cannot be limited to an efficient transmission or record of messages. On the contrary, it is necessary to consider their content in order to support the dialogues on the basis of their meaning and to have them fully linked with the operational environment in which they happen.

158 C. Simone

The capability of processing message contents is a central point in the approach presented here and constitutes at the same time its strength and weakness. In fact, due to the well-recognized limits of the results on natural language processing, this capability is based on the use of conventions in the construction of the message content, like predefined structures and keywords, reference to a shared dictionary, and the like. This limits the range of applicability of the approach: for example, it is unlikely that people engaged in real-time message exchanges are keen to accept many of such conventions. On the other hand, when applicable, these conventions are the means for the acquisition of the knowledge necessary for a management of dialogues based on their semantics. The main focus of the paper is on functionalities devoted, on the one hand, to the automatic management of parts of the dialogues, and on the other hand, to the maintenance of the organizational and cognitive contexts in which collaborative dialogues occur.

The paper is organised as follows: the next section illustrates the main features of CHAOS (Commitment Handling Active Office System), the CSCW appli­cation we are developing. In Section 3 we show to what extent this kind of application may support DL, while in the concluding section we relate CHAOS to other CSCW proposals and alternative scenarios of (collaborative) learning.

2 CHAOS: Commitment Handling Active Office System

The motivations behind the design of CHAOS and the detailed description of its structure can be found in (De Cindio 1988,1989). For the purposes of this paper it is sufficient to provide an introduction to the general system functionalities related to potential user needs. Some aspects of CHAOS more related to the specific context of DL will be illustrated in the next section.

First of all, CHAOS provides an environment integrating communication, activity management and organisation structuring. This conceptual framework is also reflected by the underlying system architecture which is organised into four main modules, respectively: the Conversation Handler (CH), the Group Agenda Module (GAM), the Group Structure Module (GSM) and the Group Language Module (GLM). At the conceptual level, integration is based on a very general 'circularity' (De Cindio 1989): communication is the means by which commitments are mutually taken, both to define and execute activities, and to define and practise the organisational structure (i.e., role definition and role playing). In its tum, communication is affected by the roles and the experience of the participants and by the nature of the activities they are involved in. Communication, activity execution and role playing are glued together in the notion of 'conversation' (Winograd 1987), i.e., a structure of communication acts with the purpose of defining commitments and a structure of activities to fulfil them together with rules capturing the role of the involved people.

Secondly, CHAOS derives from monitoring the communication of the knowledge related to the above four modules and maintains it in as many knowledge bases, as will be discussed in the next section. Let me introduce now the two main types of conversation, i.e. conversation for action and for possibility.

Supporting Collaborative Dialogues in Distance Learning 159

CONVERSATION FOR ACTION

When a commitment concerns the execution of an activity, the related conver­sation is called conversation for action (CA). It consists of the following phases (Fig. l(A) shows a Petri net describing the various conversation phases): - opening by a request/offer of doing); - a negotiation of the commitment attributes; - a withdraw or an acceptance. - in the case of an agreement, the activation of the related activity during which a new negotiation can start to redefine the commitment.

Finally, the report by the doer of the activity to the asker who expects the results, so that s/he can evaluate them. In the positive case the conversation is over; otherwise a new negotiation/execution step starts with the above mentioned modalities. Example (A = asker; D = doer):

D: I'll prepare the programming exercises in two weeks. (offer) A: It is too late: the examination is next week, on Thursday. (negotiation) D: I'll do it for Monday. (negotiation) A: OK. (agreement) .... some days later .... D: Sorry, it will be ready on Wednesday. (negotiation) A: OK. (agreement) .... some days later .... D: Here you find the programming exercises. (Report) A: Well done! Thanks. (Positive evaluation)

CONVERSATION FOR POSSIBILITY

When the purpose of the conversation is to define the group structure, the conver­sation is called conversation for possibility (CP) and its structure is like the one described above except that no action has to be performed (Fig. l(B». The conversation ends by the acceptance or the withdrawing of the negotiated group structure. Example: (A = delegAnt; D = Delegated)

A: Do you want to be responsible of this group of students on the uses of Prolog? (Offer of delegation) You will be free to organize them as you want, but I want to have a report every week. (Protocol, see below) D: OK, but in my opinion it is more meaningful to do a report every month (negotiation) A: OK

The main aim of CPs is to define people Roles. A role in CHAOS is a set of protocols, i.e. a set of rules governing future behaviours (e.g., every time a certain situation occurs, an authorisation for doing something is required), defining rights and duties of a group member within Areas (i.e. units of organisation) and/or within Activities (i.e. units of work). Accepting a role makes the related protocols applicable to future behaviours: i.e., the protocols affect the domain of possibility of people within subsequent conversations.

160 C. Simone

execution

Report OK

(A)

Fig. 1. (A) Conversation for Action (B) Conversation for Possibility

W=withdraw A= agree N = negotiate D = declaration

(B)

At the architectural level, integration is achieved by concomitant accesses to the functionality provided by all three modules, and by a flow of knowledge automatically triggered by events originated within each of the three modules, that updates the knowledge bases handled by the other two.

The main knowledge flows are the following:

From Conversation Handler to Group Structure Module and vice-versa: - when a protocol is agreed upon, the related rules are passed to the GSM, in order to keep trace of the modification of the role definitions; - at any step in a conversation, the CH asks the GSM for the appropriate pattern of the current conversation and for its relationships to other conversations, according to the previously defined rules;

From Conversation Handler to Group Agenda Module and vice-versa: - when a request/offer of doing is accepted, then the system inserts the related commitment in the to-do-list of the performer at the date and time specified by the commitment temporal constraints: commitments are described by semi structures that allow the system to derive the essential information. Any modification occurring in the negotiation phases is recorded accordingly; - when the performer tells the system that the activity fulfilling the commitment is finished, in the to-do-list the commitment is substituted by a recall that the

Supporting Collaborative Dialogues in Distance Learning 161

report (i.e., the results) have to be delivered, again at the appropriate temporal point (if any); - if a protocol requires opening a new conversation within another one (e.g. for asking an authorisation), the CH inserts this action in the to-do-list of the appropriate user. Notice that a to-do-list contains both communication actions and activities; - then, by choosing the appropriate entry in hislber to-do-list, a user can decide to perform a communication action or an activity. In the first case, the CH is activated, the conversation containing the selected speech act is resumed, and a new step is executed. In the second case, two situations are possible: either the activity is elementary, i.e., the user did not associate any structure of smaller activities to it, or slbe did it, by using the activity structuring language provided by the GAM module.

Elementary activities allow for the insertion in the to-do-list of other commitments identified by the user in hislber on-going job. Structured activities provide the user with a stronger support, as they guide the execution along their description. The language for structuring activities focuses on the partial order among sub-activities, and on the communication of what is concerned with both the deadlines and the flow of information. An example of a structured activity and of system support in performing it is given in Section 3.2.

We conclude this short presentation of CHAOS by recalling that the user can define rules (Gasparotti 1990) for associating in an automatic way communication acts and manipulations of the knowledge managed by the system. For example, rules can define criteria for filtering incoming messages and for organising them into folders depending on their contents; for redirecting them to more appropriate receivers (e.g., an expert, a responsible) or for generating automatic answers; finally, for associating with predefined events (e.g., deadlines) communication actions (e.g., recurrent requests) or actions to communication events (e.g., store in a file pieces of information contained in a message).

In presenting CHAOS we focused on its direct tailorability by the user to the target environment. Such an activity requires as a precondition the definition of new organisations or work structures. Many authors claim that these are very expensive operations. In order to reduce the efforts required to the user, CHAOS contains mechanisms for recording and recalling recurrent patterns both in protocoVrole and activity definition (patterns can be specified at any moment) so that users are in charge of handling unusual situation only.

Because at any level user support still keeps open, for the user, the options for full control of the structures to be customized, we may claim that CHAOS' flexibility is appropriate for any context of use, from very unskilled to expert users, from stiff to flexible organizations.

3 CHAOS Supporting Distance Learning

CHAOS is a technology supporting generic co-operation among human beings through the handling of dialogues among them. Consequently, CHAOS can also provide DL with specific support concerned with the handling of the communication between students and teachers (and other agents, such as tutors, school managers, experts, secretaries, administrators, technicians ... ) and its pragmatic implications in terms of mutual commitments and acquisition of

162 C. Simone

knowledge. In our view and experience, this communication contains repetitive and standardised information together with unique and 'high-quality' knowledge exchange, and implies in both cases significant modifications of the cognitive and/or the organisational context in which people are called to interact. We believe that co-operation technologies like CHAOS are able to support the above aspects. We will demonstrate this by means of some examples suggested by our experience as teachers in a university characterized by phenomena that make our teachingnearning style quite appropriate for a DL context, even if formally our university is not organised as such. First of all, there are many students preparing their exams without following (all) the lectures and having very loose contact with other students and the educational staff. Secondly, the number of students abandoning the faculty or taking their degree after a period that extends over a number of years is almost double what is considered to be normal1. Finally, the regulations make it difficult to organise different teachingnearning modalities that improve cross-fertilisation among the contents of the various courses.

3.1 Supporting Repetitive and/or Standardised Communication

Let us consider the case where a student asks a teacher to apply to a course or to write a thesis under his supervision. In both cases it is likely that the teacher requires that the student has some prerequisites before committing himself to this new duty. Then the conversations started by student requests have an initial, very standardised part that is followed, if the prerequisites are met, by a high personalized dialogue depending on the specific characteristics of the student. CHAOS provides means for handling the first part of the conversation in an automatic way through the definition of some rules, as shown in Fig. 2, and to switch to the personalized part when needed. In this way, a teacher is relieved from duties requiring his presence but not his intelligence and can save time for performing activities of a higher quality, still keeping track and control of what happens without information gaps.

Another example of a feature provided by CHAOS rules is the possibility for automatically redirecting messages to other people: this is useful when a request for information concerns a topic that is better known by another member of the teacher pool or when the teacher is absent and he delegates student supervision to other people, with the possibility of selecting the more appropriate one for each student or topic.

Here, saving time is accompanied by a better service to students.

3.2 Supporting the Organisational Context of Co-operation

The degree of success of a co-operation depends heavily on the extent to what mutual commitments are satisfied by respecting mutual expectations, deadlines, contents, modalities and the like. These goals can be achieved if people activities are supported by tools able to derive from the dialogues the relevant organisational

1 Most of the Italian Universities are public Institutions that have (comparably) low enrolment fees and no numerus fixus; the access is therefore almost "free" of costs.

Supporting Collaborative Dialogues in Distance Learning 163

LL rule for DOMAIN: Thesis

IF From: student-code Speech act: Request

COMMIT TYPE: PhD Thesis-request Subject: Groupware Type: applicative Pending exams: NOT (8, basic)

THEN Set characteristic: group thesis-open Answer: Counter-offer

COMMIT TYPE: Thesis Academic conditions

Availability: ok Starting date: ~ 01-10-90 Basic prerequisites: LISP,

Winograd & Flores's book, LAN technology

Comment: please acknowledge agreement

Fig. 2. Rules in CHAOS

LL rule for DOMAIN: Thesis

IF From: student-code Speech act: Request

COMMIT TYPE: PhD Thesis-request

Subject: Groupware Type: applicative Pending exams: > 3

THEN Answer: Reject

COMMIT TYPE: Thesis refusal

Text: Request of Thesis i: possible only when pendin! exams are less then 3 and no one of them is basic. Please try again when such conditions are met.

KL rule for DOMAIN: PhD Thesis

WHEN (planned group thesis meeting) = true

IF Characteristic: group thesis-ok THEN Answer: Counter-offer

COMMIT TYPE: PhD Thesis starting meeting

Subject: Meeting date: 25.09.90 Place: room alfa Time: 10 a.m.

Comment: please ack. participation

infonnation. CHAOS provides two main tools: the user AgendaffoDoList and the Activity Manager. Let us focus on the second one, as the support provided by the fIrst one should be clear from previous sections.

Again, let us consider an example in the area of Education. Suppose that a teacher wants to monitor the extent to which a group of students masters the topics in his course. By using his Agenda he can plan some predefined dates when students can apply for and take an examination that consists of the preparation of a paper. Moreover, he can define an activity managing all the steps of this multi­party process: an activity instance is shown in Fig. 3.

164 C. Simone

The system automatically inserts the appropriate commitments in the teacher ToDoList at the appropriate date; when he selects it from the ToDoList, the activity is instantiated so that the teacher can start the process.

Activities can also model interaction modalities governing teacher and students co-operation concerning a course, a thesis, or the execution of other learning activities. Two aspects are relevant in how CHAOS allows one to conceive and perform activities. First of all, there is an implication that the communication and co-ordination of the activities are the focus of attention more than the execution of specific non-communicative tasks (that however are supported): in fact, the main attributes defining activities concern the activation and management of conversations and the synchronization of the involved steps. This is an emerging way to look at workflow within CSCW that contrasts (as pointed out in (Medina Mora 1992» with more traditional ones mainly based on document or object flow.

Secondly, the standardisation implicitly implied by the notion of activity is mitigated by the availability of the 'free' communication environment that can be utilised for handling possible exceptions. Any kind of 'free' conversation open within an activity is flagged by its name and rules like the one shown in Section 3.1 can be defined to propagate to the activity the outcomes of the exception handling (closing a specific conversation, modifying a commit, stopping part of the activity and the like). This is very important in order to make the learning modalities adaptive to the evolving context of use: exceptions, personalizations, new possibilities should be viewed and supported as a fundamental side effect of co-operative learning.

3.3 Maintaining the Cognitive Context of Co-operation

The effectiveness of a co-operation depends also on the effectiveness of the communication on which it is based, i.e., on the capability of the interlocutors to reduce and solve the ambiguities and misunderstandings naturally arising from communication. In its turn, this capability is based on the availability of a cognitive context rich enough to interpret communication contents properly. CHAOS pays big attention to these aspects by extracting from dialogues as much knowledge as possible in order to build and maintain their cognitive contexts.

First of all. CHAOS maintains a knowledge base of the conversations each user is (on-line) or has been (off-line) involved in. Actually, each user owns one of such knowledge bases. that is private to him. Then each step in a dialogue cah be contextualized both with respect to the related conversation, its sub-conversations and, possibly, the concluded conversations the user wants to resume for any reason. As dialogues in DL can take weeks and months. this support is basic to reconstruct the communication history to the extent it is needed. For example. a teacher can reconstruct the history of the dialogues activated with each one of his students during the preparation for an examination. This reconstruction can be based both on the network of conversations and subconversations and on some keywords that appear both in the more structured fields of the messages (like, the set of involved disciplines) and in the 'free' text of the messages (that are interpreted as it will be sketched later on).

Secondly. CHAOS maintains a knowledge base of the experience gained by its users when they fulfil successfully a commitment. In CHAOS roles and

Supporting Collaborative Dialogues in Distance Learning 165

t from ToDoList DATE-TIME=ok

from ToDoList PREPARE TESTS (Action)

_ tATE-TIM&ok DEADLINE: 10-1-92 OUTPUT: <tl, t2, t3>

"IIf1

EXAMINATION (Communication)

- INPUT: <tl, t2, t3> EVALUATION (Action) TYPE : request I""" INPUT: <f1, f2, f3> COMMIT: 'Perfonn Test (input)

DEADLINE: <22-1-92> WITHIN: <19.30 115-1-92>

SOLICIT: <2 days> SOLICIT: <30 min.> RULE: if TlME>WITHIN and NO-REPORT

OUTPUT: <r1,r2,r3>

then CLOSE CONVERSATION REPORT: <f1, f2, f3> OUTPUT: REPORT

I

LEGENDA: I

- : control flow RESULTS (Communication)

-:dataflow

~: inlouttrigge~ INPUT: <r1, r2, r3> REPORT EVALUATION: INPUT RULE: if SINGLE EV ALUATION=OK

then SEND(POSITIVE) else SEND(NEGA TIVE) + to ToDoList : DONE

Fig. 3. An Activity in CHAOS

experiences determine User Profiles (Divitini 92); moreover, experiences are organised along a network of disciplines characterising the application domain. Then, it not difficult to imagine one instance of such a structure specialised to any DL environment. Disciplines are explicitly mentioned within conversations and activities: hence, upon a declaration of satisfaction by the asker (e.g., the teacher) the doer experience in the related disciplines is updated, following some (user definable) rules.

This knowledge is important in various circumstances: for finding an appropriate person to redirect conversations to (see Section 3.1); for advising students/teachers to activate an interaction with suitable experts or with other students/teachers who share or have shared similar problems or needs or duties; or simply to understand better the background of people they are co-operating with.

Finally, CHAOS maintains structures called User Models (UM), one for each user, that are the core of the GLM knowledge base. To this aim CHAOS makes use of a Lexicon and a Dictionary (again specialised to the application domain)

166 C. Simone

where syntactic and semantic knowledge is recorded, respectively. Each step of a dialogue is processed in order to interpret messages. This process is based on the semistructured format of the message (Malone 1992) and on the application of NLP techniques for interpreting the unstructured parts of the message (Bignoli 1991). Each UM contains two kinds of knowledge. First of all, the knowledge about the entities of the application domain (i.e., objects, actions, activities) involved in the message content of the conversations involving the user (Personal User). Second, the meta-knowledge about what the user knows about the knowledge of other users, i.e., the knowledge referred to within dialogues between any two pairs of users (Consciously Shared Knowledge between user i and j, CSK(i,j)). Then, each UM reflects the user dynamic communication experience and is the cognitive context for interpreting message contents in conversations involving this user.

The knowledge contained in UMs is usable for different purposes (Bignoli 1991, Divitini 1993). First of all, to find the appropriate references of the entities mentioned within message contents: such references have to be looked for in a privileged UM, i.e., in the CSK of the interlocutors first and then in Personal Knowledge of the speaker. This process is performed by the system that is able to signal potential sources of misunderstanding when the search for the referents is not (fully) satisfactory, i.e., when the interlocutor communicative experiences are not able to establish a 'consensus domain' where the interpretation process is robust.

The output of this process, i.e., VMs linked to the record of the conversation histories2 is an excellent source of knowledge for reconstructing user attitudes in facing problems. Therefore, in the particular case of DL, student and teacher learning/teaching attitudes in their mutual interactions may be stored and retrieved. For example, one may first recall how they execute learning activities: the sequence of steps, the choices in case of possible alternatives, the exceptions they raise and their consequences, the evolution of their knowledge acquisition in performing learning activities. Secondly, UMs can be exploited, both by the users or by the system, to tailor the dialogue depending on both user experience (as evaluated in Section 3.2) and on specific UM knowledge contents. For example, the system can tailor answers to specific questions avoiding details with the experts and being more specific with novices; the teachers can avoid to proposing programmes whose content is already known from the combination of past experiences or recognise missing items in a topic that is part of the domain (s)he has taught.

In CHAOS users can query the system information that can be found in the VMs; the corresponding accesses are under system control. In the specific domain of DL this restriction could be relaxed in favour of the possibility of directly accessing individual UMs, perhaps under explicitly defined policies guaranteeing user privacy. For example, this would allows for a reconstruction of the dynamics of student knowledge acquisition along all their learning activities and for monitoring how knowledge is propagated and shared among group of students. This is very relevant information for organising student co-operation and for

2 In this way CHAOS implements the well known notions of implicit and explicit 'focus lists' [Grosz and Sidner 1986).

Supporting Collaborative Dialogues in Distance Learning 167

preventing the distortions typical of knowledge propagation among unskilled people.

4 Related Work, Implementation and Conclusions

CHAOS draws its inspiration from two CSCW applications: namely, the COORDINATOR (Winograd 1987), based on the language/action perspective (Searle 1969) and on the notion of conversation, and OVAL, the new version of ObjectLens (Malone 1992) proposing the notion of semi-structured message. Moreover, the idea of using the conversation structure for representing structured activities is reminiscent of the Action Workflow proposed in (Medina Mora 1992). However, what distinguishes CHAOS from the proposals above is the idea of deriving from handling the communication the basic knowledge about the group organisation and the on-going activity structure together with the group linguistic knowledge. In this way, the cognitive and organizational contexts ofthe dialogues are updated without requiring CHAOS users to input any further information but the one necessary to communicate.

CHAOS is still under development. The following parts are implemented in a quite stable prototype: the Conversation Handler with the filtering and redirecting capabilities, the Group Structure Module and the Group Agenda Module as far as the elementary activities are concerned. The early versions of the User Models, the group Lexicon and Dictionary are presently in a redesign phase and have to be integrated into the above mentioned prototype (De Cindio 1993). Strategies for further exploiting the linguistic knowledge acquired during the dialogues (Divitini 1993) and the notation to handle structured activities are the main concern of the current research and development activity.

Coming back to the DL framework, our proposal covers aspects that are not considered by approaches based on the so called 'traditional model of distance education' and by the collaborative models of distance learning based on real-time communication that can profitably exploit the variety of tools for co-operation based on the same type of interaction. In fact, on the one hand 'independent study supported by well developed learning materials' is still valid in some specific steps of the learning activities: the problem is then to integrate its technological support with the communicative one. One way could be the integration of the two related UMs so that they collect the most updated situation; another one, could be the integration of the 'independent study' with the communicative environment, by using the concept of 'exception handling' to solve problems and to identify new strategies, while keeping trace of all the knowledge exchanged and acquired in this alternative paths. On the other hand, if and when possible within the organisational and social structure of the participants, the real-time interaction is fundamental to share experiences with a low effort, to listen other people discussions, to drawn from other people experience alternative learning/teaching modalities. Obviously, in this scenario dialogues can become very complex and the exchange of knowledge can be based also on the real-time dynamics captured by shared multimedia environments. Building UMs in this scenario becomes very difficult and the advantage of a very creative framework is paid for by the difficulty of reconstructing, both from the student and the teacher points of view, the learning process that actually occurred together with the related knowledge acquisition and propagation.

168 C. Simone

Finally, our proposal could be profitably integrated with approaches like (Cerri 1992) as they can provide complementary tools and strategies to exploit the knowledge CHAOS is able to acquire from handling communication. For example, both approaches share the notion of UM and of role. The basic question is: are these notions used for the same purposes ?If not, as it seems at a first glance, how to integrate their use? Answering these question is the matter of a future work.

As good quality materials, collective activities and autonomous reflections all contribute in a fundamental way to a fruitful learning, so the related supporting technologies have to be integrated to provide an adequate support. This is not only a technological problem, but rather a modelling problem as an effective integration requires the understanding of the contribution each isolated and/or collective activity provides to the whole multi-party and multi-media learning process.

Acknowledgements

The author wants to thank all the members of CHAOS team over the years, in particular G. De Michelis and F. De Cindio with whom CHAOS was initially conceived, and P. Omodei Sale' whose contribution to CHAOS implementation is fundamental; moreover, Stefano Cerri for his interest in our work and his encouragement in connecting its results with Distance Learning themes, and for his careful reading of an early version of this paper. This research has been conducted in the frame of the Italian project "Information Systems and Parallel Computing", Research Track (7.2) 'Groupware' and the ESPRIT-BRA Project COMIC (#6225).

References

Bignoli, C., Simone, C.(1991) AI Techniques for supporting human to human communication in CHAOS. In: J.M. Bowers and S.D. Benford (eds.), Studies in Computer Supported Cooperative Work - Theory, Practice and Design. North­Holland, Amsterdam

Cerri, S.A.; Cheli, E.; McIntyre, A. (1992) Nobile: user model acquisition in a natural laboratory. In: Jones, M., Winne, P.H. (eds.) Adaptive Learning Environments. NATO ASI Series F, Vol. 85. Springer-Verlag,1992, pp. 325-347

De Cindio, F., Simone, C., Omodei Sale, G. (1993) CHAOS Modules Integration: the user interface, CNR "GROUPWARE 7.2" Technical Report 7/95 (2.0)

De Cindio, F., De Michelis, G., Simone, C. (1989) Groups in a Language/Action Perspective. Second European Meeting on Cognitive Science Approaches to Process Control. Proceedings, Siena

De Cindio, F., Simone, C., Vassallo, R., Zanaboni, A (1988) CHAOS: a Knowledge­Based System for Conversing Inside Office. In: W. Lamersdorf (ed.) IFIP TC8IWG8.4 International Workshop on Office Knowledge: Representation, Management and Utilization. Proceedings, North-Holland

Divitini, M., Omodei Sale, G., Pozzoli, A., Simone, C. (1993) Supporting the dynamics of knowledge sharing within organization. COCS 93. Proceedings, Milpitas, CA, 1993.

Supporting Collaborative Dialogues in Distance Learning 169

Divitini, M., Simone, c., Tinini, R. (1992) User Models and Profiles in CHAOS. Workshop on Man-Machine Interaction. Proceedings, Bari

Gasparotti, P., Simone, C. (1990) A User Defined Environment for Handling Conversations. In: S. Gibbs and A. Verrijn-Stuart (eds.) Multi-User Interfaces and Applications, Elsevier Science Publishers, North-Holland

Grosz, B.J., Sidner, C. (1986) Attention, Intention, and the Structure of Discourse. Computational Linguistics 12(3)

Malone, T.W., Fry, C., (1992) Experiments with OVAL: a radically tailorable tool for cooperative work. In: M Mantei, R. Baeker (eds) CSCW92. Proceedings, Toronto

Medina Mora, R., Winograd, T., Flores, R, Flores, F. (1992) The Action Workflow approach to workflow management technology. In: M Mantei, R Baeker (eds) CSCW92. Proceedings, Toronto

Searle, lR (1969) Speech Acts: an Essay in the Philosophy of Language. Cambridge University Press

Winograd, T., Flores, F. (1987) Understanding Computer and Cognition. A New Foundation For Design. Ablex Publishing Corporation, Norwood, NJ

14 Towards Models of Interaction Between an Artificial Agent and a Human One

Dominique Guin

ERES, Departement de MathcSmatiques, Universit~ Montpellier 2, Place E. Bataillon, F-34095 Montpellier cedex 5, France

Abstract. We present a transposition of a model of interaction in machine learning which may show the way to describe a wider range of interactions in learning environments. It takes into account the basic role of interaction and collaboration and the necessity of various teaching styles which occur at different moments in different domains. Our model provides a possible way to describe didactic interactions between two agents: the teacher is an artificial agent (agent A) whereas the learner is a human one (agent H). Flexibility and adaptibility of this model come from the distinction between an agent and the various roles successively played by him during the different phases of learning.

Keywords. Agent, behavioural level, protocol of communication, conceptions, dialogue, didactic interaction, interactive learning environment, knowledge level, role, metaknowledge, message, teaching style

1 Introduction

Our aim is to provide a model allowing the description of a wider range of interactions in learning environments instead of focusing on knowledge base, problem solving, or a particular learning theory. M.Elsom-Cook and P. Goodyear have pointed out the basic role of interaction and collaboration in learning environments with the necessity of taking into account various teaching styles which occur at different moments in different domains, and are successively reified into tutors, guided discovery or micro-world (Elsom-Cook 91, Goodyear 91).

We present a transposition of a model of interaction in machine learning MOSCA describing machine learning from examples. This transposition provides a possible way to describe didactic interactions between two agents: the teacher is an artificial agent (agent A) whereas the learner is a human one (agent H). Flexibility and adaptibility of this model come from the distinction between an agent and the various roles successively played by him during the different phases of learning.

Defining the didactic interaction has the explicit purpose of teaching for the artificial agent (agent A), of learning for the human agent (agent H). This transposition, aiming to describe didactic interaction, is focused on a typology of communications, in order to make computer programs more able to manage dialogues with learners. It requires us to consider mutual understanding during the

Towards Models of Interaction 171

entire dialogue: for this purpose, the description of the didactic interaction has to be made at a behavioural level in order to make explicit what is implicit in the communication. Then a protocol of communication may be defined allowing us to emphasize, in the interactive process, the significant roles of understanding and forecasting the action of the other partner.

Our intended point of view is pragmatic: it is a classification of behaviours along the spectrum of interaction between the two agents in order to describe them in various types of instructional situations defining the successive roles played by the agents during the interaction. Therefore, from the data of a particular domain and a given learning theory, it will be possible to describe the interaction in terms of roles, specifying roles and the control of the management of these roles for the two partners. We have to mention the system NOBILE (Cerri et al 92) where a different notion of role is introduced for negotiating the learner model with a human expert during dialogues where roles are inverted.

2 The Pentad MOSCA

In the situation of machine learning, the learner is an artificial agent and the expert is a human one. The MOSCA protocol (Reitz 92) comes from semi­empirical theories (Sallantin et al 91). It is an analysis of machine learning in terms of types of messages.

2.1 The Knowledge Level

The knowledge level (Newell 82) is a hypothesis which takes place in the continuity of the organization of computer systems at different levels: each system can be defined at a given level without reference to the other levels, and each level can be reduced to the lower one. This hypothesis supposes that there exists one upper level above the symbolic level allowing us to elaborate models without expressing them in a computer language. Knowledge is expressed in terms of goals and actions, systems in terms of agents, data processing is realized according to a behavioural law, the principle of rationality: actions are selected to reach goals.

Agents are entities able to act in an autonomous and a rational way, and to evaluate their potential action relative to their current state in order to make a decision for action. Moreover, they may have beliefs in the effects of their action. These models have to be able to provide the simulation processes that they are supposed to model, and to be understandable by humans in order to be shared between experts, cogniticians and knowledge engineering.

2.2 State and Transition Operator

In the MOSCA protocol, an agent is characterized by his current state (taken in a states space E). Then he evolves, in other words, he changes his state. This change is realized by the application of an operator of states transition (taken in a operators space 0, with 0 : E --> E).

172 D.Guin

2.3 Phase and Behaviour

The learner behaviour of an agent is defined as a path in a product space state­operator E x 0, an element of this space is named a phase. Then a phase of an agent is a pair (state, operator). The phase space is a subset of the product space E x O. Two successive phases (ei, 0i) and (ei+l> Oi+t> of a behaviour are naturally linked by the relationship:

2.4 Roles

A role is a subspace of the phase space: roles define a partition of the phase space. An agent plays a role during a behaviour if all the state-operator pairs characterizing his behaviour belong to this role. He plays only one role at a given moment, but he plays successive roles during his behaviour .

... } :; five w'"

Fig. 1. The behaviour of an agent, a path in the phase space

The notion of role aims to classify various behaviours of an agent for reducing its complexity. The MOSCA protocol is based on Lakatos' theory: "proofs and refutations" (Lakatos 84). Two agents communicate playing a role among the five roles: Master, Oracle, Probe, Questioner and Apprentice displayed in a pentad (in English: MOPQA).

Master

~ /

Probe

Fig. 2. The MOSCA pentad

Oracle

/ APprentke~

Questioner

Towards Models of Interaction 173

In this behavioural protocol, the type of message depends on the role:

* The oracle provides solved problems whose solution is not refutable, therefore assertions are always true.

* The questioner provides unsolved problems to the apprentice.

* The probe provides solved problems to the apprentice whose solution is refutable, therefore assertions may be false.

* The apprentice: - can solve problems given by the questioner, - can give an argument about the probe's proposition. - can ask some questions to the oracle, or put some request to the questioner or the probe.

* The master receives the apprentice's argument and criticizes it.

In this protocol of communication, there is a clear distinction between the agent and the role played by him at the instant t, in order to classify various behaviours of an agent for reducing its complexity.

3 Framework

3.1 Variety of Models and Ways for Designing Educational Systems

We do not want to give an overview on learning theories, but we will just give one example in order to express the diversity in designing, from them, adaptive learning environments:

In (Derry 92), S. J. Derry examines two models of interactive learning which emphasize metacognitive skills: a neo-piagetian model pointing out the role of exploration where the student is characterized as a reflective explorer, and a model based on Vygotskian psychology (Vygotski 78) where the teacher plays the central role of a mentor. Acting in the zone of proximal development, as the student's problem solving abilities improve, the mentor reduces his assistance encouraging the student to think independently, to use strategies and control processes. Finally, S. J. Derry selected the Vygotskian approach, designing a system which is a content dependent domain area (word additive problems).

P. Dillenbourg opted for a Piagetian orientation with some Vygotskian aspects (Dillen bourg 92): the instructional interaction can support the learner's psychological reflection, which is the cognitive activity that leads to some consciousness of one's own knowledge. Then he proposes a pyramid metaphor for designing educational systems that progressively transfer to the learner an increasing amount of control in the solving process. P. Dillenbourg does not propose a specific system, but rather a general architecture for metacognition.

174 D. Guin

3.2 A Didactic Approach

The didactician approach is that learning is an adaptive process of the learner to an environment in which he really takes part, but it stresses educational purposes on the content. In this constructivist approach, a conception is viewed as a mental representation of the learner about a concept (it may be initial, eventually erroneous in comparison with reference knowledge).

The organization of various types of knowledge may be pointed out by the notion of conceptual field (Vergnaud 90) providing a frame to organize various types of knowledge taking part in the formation of the signification of the concept as:

- problem situations, - procedures (describing the invariant organization of the behaviour in

terms of operative invariants and schemes), - signifyings, in order to describe these problems and procedures (for

example, various diagrams, representations or linguistic expressions).

Teaching may be viewed as a sequence of interactions between the teacher and the learner aiming to lead the learner to doubt (for example, when faced with a contradiction) and to revise his conceptions. All through the interaction, there is an evolution of the conceptions of the learning agent relative to these various types of knowledge in the learning process: faced with a new situation, his conceptions evolve, and the interaction is an expression, a manifestation at the actor level of this evolution, relative to the behavioural one.

The interest of the didacticians is essentially to study the articulation conceptions-situations in the learning process (Artigue 90). Our aim, then, is to elaborate interactive learning environments in order to facilitate this evolution of conceptions sharing the initiative between the two partners. For this purpose, we do not want to base instructional systems upon only one model of interactive learning, but rather would prefer to have the possibility to describe the interaction which will be successively based on various models.

3.3 Two Simultaneous Learnings

At a given moment, the behaviour of the learner results from his available schemes and conceptions (beliefs) which may be eventually erroneous about the conceptual field. We focus here on the articulation conception-interaction in the learning process. The interaction is closely linked to the epistemic model which is the internal model of the system representing conceptions of the artificial teacher (agent A) about the learner. This model requires an interpretation of the observable behaviour at the interface, but it depends on the choice of the observables at the interface. The epistemic model must be carefully distinguished from the searcher model, and also from the real learner's conceptions (Balacheff 91). For this purpose, we will try to integrate the distinction pointed in (Dillenbourg and Self 91) between beliefs of various agents (system or student).

Our main idea is that in an educational situation, two simultaneous learnings coexist: during the student's learning, the teacher also learns about student's

Towards Models of Intemction 175

conceptions. Then, he builds a learner model during the learning process and defines, from it, his didactic action: in other words, his didactic choices (teaching strategies) are linked to the learner model.

If we now consider the knowledge base of the artificial agent A, it includes a part of static knowledge about a given conceptual field. In this static knowledge, which is an expert knowledge, there is some declarative and procedural knowledge, but also some metaknowledge. Moreover, one can distinguish different kinds of metaknowledge: there is strategic knowledge relative to the conceptual field and general knowledge (domain independent) which is linked to the control of the use of knowledge (in Section 4 we will give examples of dialogues dealing with these two types of metaknowledge). Finally, the expert knowledge contains didactic choices, that we can express as pedagogical strategies.

On the other hand, the knowledge base of the teacher agent A has also an evolutive knowledge: the epistemic model relative to the conceptions of the agent A about the conceptions of the agent H (on a given conceptual field). This evolution will be described at the behavioural level by the application of operators.

The observable part of the behaviour of each agent may be described by a succession of dialogues from various roles (not necessarily five) he successively plays. Therefore, this observable part, linked to the evolution of his conceptions, may be set out in a multiad. There are two learnings on different knowledge domains, then two simultaneous reviews and evolutions of conceptions, then two multiads:

Linked to the evolution of the epistemic model

(learner model) Fig. 3. Two multiads in a teaching process

Linked to the evolution of the learner conceptions

4 Interactions Relative to Human Learning

In this section, we focus our attention on a pentad of the human agent, then we want to classify different types of flows circulating between the artificial agent A and the human learner H. Flows may transmit knowledge concerning various components of a given conceptual field, but also metaknowledge. Flows may also refer to the content of the screen or not. The linking of various roles played by the agent A is aimed to improve learning of the human agent H, but the control of

176 D. Guin

these flows will not be described here, because it depends on the chosen learning theory and teaching strategies. We will give examples of flows in mathematics, some of them are chosen in (Cauzinille 92). Other examples in various domains may be found in (Guin et al 93). In this description, the apprentice's role is always played by the agent H.

4.1 Oracle to Apprentice

This flow deals with the transmission of true knowledge to the learner (agent H). The oracle's role is always played by the teacher (agent A). This knowledge is relative to various components of a given conceptual field (situations, procedures and signifyings), but also changes of representations and metaknowledge. The oracle provides solved problems (in the broad sense: knowledge) whose solution is not refutable (assertions are always true) about various components of a given conceptual field. This flow may be activated by a request from the master or the apprentice himself. The choice of what and how knowledge will be transmitted is linked to the epistemic model, therefore it depends on the current state of the agent A. Here are examples of messages which may be transmitted:

Domain Geometry Algebra

Natural logic

Message This geometric diagram is a square. Factorizing this polynomial, we pass from step n to step n+ 1 (given steps) applying such identity, for example from (25x2- 36 - (5x+6) (x-2» to (5x+6) «5x-6) - (x-2» applying the identity a2_ b2 = (a+b)(a-b)) and the factorisation of (5x+6). The negation of this sentence is "there is at least one Cretan who tells the truth".

4.2 Apprentice to Oracle

The agent H (playing the apprentice's role) may take the initiative asking some question to the oracle (played by the agent A), more generally a request for help. This message may contain information taken as parameters for elaborating the inverse flow.

Domain Indeterminate Algebra

Geometry Indeterminate

Message I don't understand this word, what does it mean ? In this calculus that you give me, which identity is allowed to pass from this step to this other ? Give me another example of a parallelogram? What is the link between diagram 1 and diagram 2 ?

4.3 Questioner to Apprentice

If this role is played by the agent A, the aim may be to get more information about the diagnosis, i.e. the learner's conceptions of the agent H. The agent A may test the persistence of an erroneous conception from the agent H. Therefore, the choice of the content is linked to the knowledge base, and also to the

Towards Models of Interaction 177

epistemic model. On the other hand, the agent A may eventually propose tasks of comparison, and more generally encourage the agent H (playing the apprentice's role) to use metaknowledge.

The agent H may also take the initiative playing this role of questioner if he wants to evaluate himself or increase one's standing in front of the agent A on a subproblem or an analogous problem. The questioner provides unsolved problems to the apprentice, queries may deal with various components of a given conceptual field or metaknowledge.

Domain Algebra Natural logic Metaknowledge

Metaknowledge Metaknowledge Metaknowledge Metaknowledge Arithmetic Indeterminate

Message What about the polynomial 4x2 + 4x + I ? What is the negation of the sentence "All Cretan are liar" ? Look more closely at this step, what can you say? Can you see some contradiction ? Have you an idea how to solve this problem ? What can you use to convert this expression ? Compare your result with the starting point, aren't they similar? Are these two problems analogous ? What operation do you know? (help or diagnosis) Can you explain how you found this result? (diagnosis)

4.4 Apprentice to Questioner

Since we focus on the pentad of the agent H, the role of apprentice is always played by the human agent H. A prerequisite of such a message is a previous query from the questioner. The apprentice can solve problems given by the questioner, but he can also take an initiative requesting the agent A to ask something. This situation is different of the previous one (the agent H playing the role of questioner), because he does not choose the content of the question.

Domain Algebra Natural logic Metaknowledge

Metaknowledge Metaknowledge

Indeterminate

Message This polynomial 4x2 + 4x + I is a square. The negation is "There is at least one Cretan who tells the truth". Yes, indeed, there is a contradiction! I have probably made a mistake! I'll try to find it. Yes, indeed, I have gone round in a circle! In this diagram, I have a parallelogram, then I'll try to use this method which is based on parallelograms. Give me another example ...

4.5 Probe to Apprentice

The role of probe is always played by the agent A to force H to give an argued criticism (to test the epistemic model). These flows are similar to those of the oracle, but the probe provides solved problems to the apprentice whose solution is refutable (in other words, assertions may be false). It may be activated from the initiative of the agent A or H.

178 D.Guin

Domain Algebra Natural logic

Algebra Metaknowledge

Message The polynomial 4x2 + 4x + 4 is a square. The negation of the sentence "All Cretan are liar" is "All Cretan tells the truth". Here, there is a sum of 3 terms, do you agree ? This problem can be solved with this plan.

4.6 Apprentice to Probe

These flows are analogous to those of the apprentice to the oracle. To control his good understanding, the agent H playing the apprentice's role may query a question: there is still a shared initiative.

Domain Algebra

Metaknowledge

Message Give me another example of polynomial, I'll justify if it is a square or not. Give me another problem with a plan, I'll justify if it fits with the problem.

4.7 Apprentice to Master

This type of message has a prerequisite condition: it is an argument on an assertion transmitted by the probe in a previous message. The argument may be composed of knowledge of various types necessary to evaluate the acquisition of a concept. This argument has to fit with a granularity fixed by the teacher which depends on the epistemic model. Therefore, this argument may be incorrect because it is false or incomplete (relative to the granularity fixed by the agent A), and then it may induce an evolution of the epistemic model.

Domain Word additive problem

Arithmetic

Message To solve the following problem: Yesterday, Paul and Pierre played two games of marbles. On the first game, Paul won 5 marbles ; on the second game, Pierre lost 4 marbles. What's the result ? It is sufficient to replace in the second sentence "Pierre lost 4 marbles" by "Paul won 4 marbles". 4112 is less than V3 because V6 is less than V3.

4.8 Master to Apprentice

The agent A, after having received the apprentice's argument, decides if the explanation is correct (according to the didactic contract); eventually he may criticize it. Then the agent A plays the role of master, here are some examples of such criticisms:

Domain Indetenninate Arithmetic

Indeterminate Algebra

Towards Models of Interaction 179

Message This argument is not sufficient. Yes, but especially because the fraction 2/6 is equivalent to 1/3 and 1/3 less than 2/3! Good idea, but the plan is incomplete! One cannot add terms which are not similar! (in reaction to the erroneous procedure 3a + 2 = 5a).

5 Perspectives

Integrating didactic concepts, we have proposed a way to describe didactic interaction elaborated from a model of machine learning. The evolution of the behaviours of the agents will be represented by the action of operators depending on the learning domain of the agent (A or H). Then one can describe an educational situation with two protocols which interact, taking into account the variety and the organization of knowledge in the conceptual field, and the variety of teaching styles. This model based on functional roles will allow us to describe simultaneously the evolution of the conceptions of the learner and those of the epistemic model. The distinction between an agent and the role he plays at a given moment seems to be an interesting approach for conception Qf Interactive Learning Environments, but especially for collaborative dialogue technologies. Therefore, our purpose is to provide a language' of description of such dialogues in various application domains (mathematics, physics and biology, etc.), where the didactic interaction has to be shared between roles.

For this objective, it is necessary to evolve this proposition: from the given (a learning theory and a restricted conceptual field), different roles and a partition between these roles have to be defined and set out in a multiad. We have pointed out the necessity of managing simultaneously two multiads in order to describe interaction in learning environments. Then the management of the flows has to be expressed in the definition of the control which will depend on the chosen teaching style. Protocols of communication will be elaborated from the behavioural level of the agents all through the interaction. It may be considered as a first approach, providing a language of description at the knowledge level of the interaction in interactive learning environments while taking into account the specificity of the content.

From now on, we are aiming to test the validity of this proposition on various conceptual fields. Our first application domain is given by the system ELECTRE (Palies 88) which elaborates, from machine learning techniques, a learner model. Our objective is to take into account the elaborative and collaborative nature of the learning process defining an interaction to evolve the erroneous conceptions pointed out in ELECTRE. This interaction will be described in terms of behaviours of agents (cf. fig 1) based on roles and messages between roles which will be set out in two multiads.

Moreover, we make the assumption that this methodology may be used in various interactive situations (dialogue games, ... ) and we hope that experiments will be attempted in this way to elaborate models of interaction in learning environments.

180 D. Guin

References

Balacheff, N. (1991) Contribution de 1a didactique et de l'epistemologie aux recherches en EIAO. Actes des Xmemes journees francophones sur l'lnformatique. IMAG Grenoble: Bellissant C.

Cauzinille, E., Me1ot, A.-M. (1992) Explications et apprentissage: l'analyse d'un dialogue tutoriel. Actes du colloque Explication'92, Sophia-Antipolis, 43-63

Cerri S.A.; Cheli E.; McIntyre A. (1992) Nobile: user model acquisition in a natural laboratory. In: Jones, M., Winne, P.H. (eds.) Adaptive Learning 'Environments. NATO ASI Series F, Vol. 85. Berlin: Springer-Verlag ,1992, pp. 325-347

Derry, S. J. (1992) Metacognitive models of learning and instructional systems design. In: Jones, M., Winne, P.H. (eds.) Adaptative Learning Environments, NATO ASI Series F, Vol. 85. Berlin: Springer-Verlag, 1992, pp. 257-286

Dillenbourg, P. (1992) The language shift: A mechanism for triggering metacognitive activities. In: Jones, M., Winne, P.H. (eds.) Adaptative Learning Environments, NATO ASI Series F, Vol. 85. Berlin: Springer-Verlag, 1992, pp. 287-315

Dillenbourg, P., Self, J. (1991) A framework for learner modelling. Interactive Learning Environments 2(2)

Elsom-Cook, M. (1991) Dialogue and teaching styles. In: Goodyear, P. (ed.) Teaching Knowledge and Intelligent Tutoring, Ed P. Goodyear, New Jersey: Ablex

Goodyear, P. (1991) Research on teaching and the design of intelligent tutoring systems. In: Goodyear, P. (ed.) Teaching Knowledge and Intelligent Tutoring, New Jersey: Ablex

Guin, D. (1991) Necessite d'une specification didactique des environnements informatiques d'apprentissage, Actes des deuxiemes journees E.I.A.O Cachan 91, 253-259

Guin, D., Billet, S., Reitz, P., Herin-Aime, D. (1993) Protocole comportemental de l'interaction didactique entre un agent artificiel et un agent humain, Environnements Interactifs d'Apprentissage par Ordinateur, 193-205: Eyrolles

Lakatos, I. (1984) Preuves et Refutations. Paris: Hermann (translated from the English: Proofs and Refutations, Cambridge University Press)

Newell, A. (1982) The knowledge level. Artificial Intelligence 18,87-127 Palies, O. (1988) Meta-connaissances pour la modelisation de l'eleve, contribution au

diagnostic cognitif par systeme expert, These de doctorat de l'universite Paris VI. Reitz, P. (1992) Contribution a l'etude des environnements d'apprentissage.

Conceptualisation, specification et prototypage. These de l'Universite Montpellier II

Sallantin, J., Quinqueton, J., Aubert, J.P. (1991) Les thoories semi-empiriques: - Conceptualisation et illustrations, Revue Intelligence Artificielle 5/1, 69-92 - Elements de formalisation, Revue Intelligence Artificielle 511, 93-107.

Vergnaud, G. (1990) La thoorie des champs conceptuels. Recherches en Didactique des Mathematiques 10 (213), 133-170

Vygotsky, L.S. (1978) Mind in Society. The development of higher psychological processes. Cambridge, MA: Harward University Press

15 The "Natural Laboratory" Methodology Supporting Computer Mediated Generic Dialogues

Stefano A. Cerri

Dipartimento di Scienze dell'lnfonnazione, Universita di Milano, Via Comelico 39, 1-20135 Milano, Italy E-mail: cerri@hennes.mc.dsi. unimLit

Abstract. In previous papers we have presented the NAT*LAB methodology for student model acquisition and suggested potential extensions of parts of the methodology to knowledge acquisition and knowledge communication in Informative, Tutoring and Design dialogue management systems (Cerri and McIntyre, 1991). In this paper we wish to refine the potential extensions of the methodology for managing collaborative dialogues generic with respect to various dialogue types, partner's types (human or computer) and partner's location (local or remote). It is suggested that, although the methodology may be applied for achieving a wider range of functionalities, its major potential applicability exists in model construction and refinement by abduction, as is the case in diagnosis.

Keywords. Collaborative dialogues, CSCW, knowledge acquisition, model construction, diagnosis, user models

1 Introduction

1.1 Computer Mediated Dialogues

Computer mediated communication (CMC) is concerned with human-to-human communication at a distance by means of a telecommunication network. Examples are the electronic mail and systems for phone or video teleconferencing. The distinctive value added by CMC is that the network allows the time required for a message to be available for the receiver to be reduced independently of the distance.

In this paper we assume asynchronous written communication. From this we have several consequences (cf. Simone 94) including the following ones: (i) messages are separated and easily identified, (ii) we may treat single messages as semantically and pragmatically relevant and therefore classifiable.

When communication (a) occurs between two partners, (b) is bi-directional, then it may be called a dialogue. Dialogue is defined as an ordered sequence of messages, each of which has an emitter and a receiver where at each tum each partner plays alternatively the role of the emitter while the other partner plays that

182 S.A. Cerri

of the receiver (cf. Guin 94 for a formal definition of roles played by two agents in a dialogue).

In the following, we will therefore call Computer Mediated Dialogues (CMD) the subset of asynchronous, written CMC processes that respect these conditions.

As we want to consider not only humans but also computers as candidate partners entitled to play each of the two roles, we extend the previous definition of CMD to Human-Human, Human-Computer and Computer-Computer dialogues. We will call Conversations sets of Dialogues among more than two partners.

1.2 Human-Computer Dialogues

Human-Computer dialogues may be classified as one of three main types: Informative, Tutorial, and Design. The classification refers to the expected dominating purpose - in pragmatic terms - of the whole dialogue from one single partner's viewpoint; i.e. from the user's viewpoint... A dialogue with an Information System may be classified Informative, as the human user mainly puts queries and the system answers with the information available; a dialogue with a word processor or a CAD system or a programming language interpreter may be classified as Design because the human user (mainly) issues commands to be executed by the system; in a Tutorial dialogue the user mainly responds to questions by the system or performs the actions suggested I required by the system supposed to play the role of a teacher in the function of tutor I examiner. For each dialogue belonging to one of these dialogue types there are phases of the dialogue (sequences of turn taking by the partners) that may be classified to belong to another type: for instance during a Design phase, the user may need help and ask the system about one of its functionalities; therefore initiating a (sub)dialogue that may be classified as Informative or Tutorial (cf. Winkels 1992 for a report about different types associated with phases in dialogues where the user's purpose is to require help from the system)

1.3 Tutorial Dialogues

One of the most influential, historical papers concerning Tutoring Systems [Carbonell 70] claimed that Tutoring dialogues in realistic situations should be "mixed initiative"; i.e. the student user and the system teacher are entitled to take in turn the initiative to lead the following phase of the dialogue in terms of the main goal, plan, tasks to be executed by the partner, etc .. For instance, a student may wish to switch to an Informative phase putting queries and expecting the partner's answer; therefore modifying radically the expected next move of the system-teacher.

From those times, the "dialogue" stream of research l in Tutoring systems has more and more merged with research in Artificial Intelligence and Cognition

lTbe two main streams of research in Tutoring Systems or Information Technologies for Learning, may be associated to the "dialogue view" and to the "interface view" of the learning process. The first stream is interested in the pragmatics of dialogues the

The "Natural Laboratory" Methodology 183

because when a user puts a free query to a system, the latter should at least be able to understand the query. For a number of years people thought that "understanding" the user implied the analysis of the syntactic and semantic layers of the user's query, thus a so called "natural language interface". Later, the main focus shifted to the pragmatic layer, and especially to the model of errors, bugs and misconceptions, called "student model" (Self 74; Clancey 86; Self 90). The reason came from the (debated) hypothesis that error analysis, diagnosis and remedy are key issues in good teaching, as we learned from Plato and his teacher Socrates. These endeavours were pioneering - in the 70' s - the attitude to consider the pragmatics of dialogue in System design in general, not only in Tutoring.

However, due to the extraordinary difficulty of modelling human incorrect knowledge and reasoning strategies, the success of systems incorporating student models was limited to a low number operating in very narrow domains. In order to bypass the almost intractable problem of student modelling, people observed that perhaps good teaching is neither necessary nor sufficient for learning. But yet. the problem of understanding the user in dialogues remains central.

1.4 Knowledge Acquisition

At the same time people recognised that also models for correct knowledge are difficult to be build, because they are not only domain but also goal-specific and therefore the knowledge acquisition process can hardly be reused. For instance: when people tried to use mature expert systems as a source of expertise to be associated to a tutoring system, - the domain component - they discovered that the expert knowledge necessary to teach how to solve a problem is to a great extent different from the expert knowledge used to solve the problem and not simply an add-on 2.

The knowledge acquisition bottleneck stimulated methodologies for knowledge acquisition, formalisation and reuse. These methodologies (similarly to software engineering methodologies) consist of a set of computer-supported prescriptions for the designer of a system that help to identify the relevant issue in design and try to consolidate standard formalisations of the result of interviews, i.e. of dialogues between the designer and the expert (cf. McConnell, 94 for data emerging from experimental interviews in distance learning). All the methodo-

learner is engaged in, independently of the technology supporting the dialogue while the second stream concentrates on the technology for interfaces and/or for communication, in short: multi - hypermedia and telecommunication networks. The classification between the two main requisites of new Technologies supporting learning into systems for flexible and systems for distance learning does not fully clarify the different cultural approaches described above. 2 From this observation, currently accepted, Radboud Winkel's 5th proposition in his Ph.D thesis (Winkels 92): "All knowledge-based systems should be developed as if they are the domain component of an Intelligent Tutoring System". We agree with this view. Notice that the statement comes from one of the key software developers in EUROHELP, perhaps the largest European ESPRIT project in the domain of Tutoring Systems for industry, active in the years 1985-91.

184 S.A. Cerri

logies assume that the result of design (and the subsequent implementation) is a more or less "mature" version of the system to be produced: the user of version n of the system cannot modify the behaviour of version n according to his needs; in case he wishes, he is expected to notify the designer who will decide if and how the system will be modified for a better version n+ 1.

1.5 Motivation for the Work and Structure of the Chapter

We have addressed the issue of a methodology for knowledge acquisition in a quite different way. We have supposed that it is not impossible to think of the user as a co-author 3 of very constrained components of the system, so that the subsequent user-system dialogues may generate refined versions of the system. Differently from other Authoring projects, the modification we envisioned occurs at run time, while using the system. The rationale for setting this target was to facilitate the construction and refinement of student models by letting the expert-teacher intervene directly - and not through the system designer - in limited aspects of the system development process. At least, we wanted to asses to what extent a similar scenario is feasible (cf. the "reification of interactions" in Boder and Gardiol, 94).

As a result of these activities, mainly concentrated around a CEC DELTA Exploratory Phase project called NAT*LAB: Student Model Acquisition in the "NATural LABoratory" 4, we have produced a generic program called NOBILE: Nat*lab Object Based Interactive Learning Environment 5. The system may be considered a "shell" supporting our methodology for student model acquisition. Although NOBILE may be interesting as a computer-supported methodology for new training systems, we maintain that its abstractions and generalisations may be directly applicable also to a variety of CSCW situations.

As much as the suggestions of the paper will be convincing, they may be considered a supportive evidence for the potentially fundamental nature of research and developments in CDTDL: from the analysis and synthesis of one concrete stereotypical communicative situation (i.e. that of learning) we induce guidelines suited for abstraction and generalization on a wider range of communicative situations.

In the following I will first briefly describe the NA T*LAB methodology supported by NOBILE and then suggest its potential abstractions and generalis­ations to Computer Mediated Dialogues.

3 Generic co-authoring systems (in the domain of Education) are currently an issue of active research, for instance in the CEC DELTA OSCAR (and the related DISCOURSE) projects (cf. Ulloa et al. 1994). These systems have the purpose of providing groups of authors - assumed to be unskilled in programming - tools for direct authoring not only of the interfaces (potentially complex multimedia documents) but also of the knowledge structure (potentially complex dialogue management systems). 4 "Natural" as we thought that a relevant contribution to system development would occur in the natural context where systems are used. 5 For better references to NAT*LAB, see the acknowledgements at the end of the paper.

The "Natural Laboratory" Methodology 185

2 The Natural Laboratory Methodology

2.1 Parameters of a Dialogue

Suppose we describe formally a concrete dialogue management system D as a function that manages dialogues, i.e. sequences of turn taking, each of which is a pair of messages. Each partner in a dialogue possesses such a dialogue management system in order to engage in the dialogue. Therefore we assume that two D exist: if the dialogue is a human-computer dialogue, one D belongs to the human, the other to the computer.

The D associated with a computer generates effective procedures once its parameters are specified and the computer starts the generation process. We will think of the following parameters for D:

PI: the agent6 that produces the first message mi to be sent to the other partner; PII: his unique identifier; P12: his physical location; P13: the role PI assumes to have in the dialogue; PI4: the knowledge base PI uses; including the strategy for conducting the dialogue; PIS: the time of occurrence of the dialogue.

P2: the agent that receives mifrom PI and produces m2 to be sent to PI; P2 is defined by P2I->P25 similar to PI.

We will call "partner model" the information available to each D belonging to one of the partners about the other partner. If, for instance, we consider a computer with D's parameters such that PI 1 is its own identifier while P2I is the identifier of a user, then in D the data associated to P2 are the partner model. If the partner is a leamer, and the computer is running D with teacher as a role, then in D the P2 information will be the student model, including the "traditional" description of the student's state of domain knowledge that is described - for PI -by P24.

2.2 Dialogue Types, Roles, Strategies

In the specific situation of Tutoring, the system has usually the "strategy" of dialogue management that consists of a sequence of presentations to the human­student of some domain of knowledge or expertise, followed by a sequence of queries by the system and answers by the student that aim at assessing if the

6 In the following we will also call agent a partner in the dialogue in order to suggest: a. a potential generality to multi-agent conversations; b. the underlying object­oriented architecture that facilitates the construction of programs simulating autonomous agents, as it is now common use in the literature in order to separate the logical properties from the implementation details. For other similar proposal of an interpretation framework based on agents, cf. euena, Garcia-Serrano and Verdejo, 1994; Guin, 1994.

186 S.A. Cerri

student has acquired (or constructed) the knowledge from the teacher's presentation and consolidating his expertise, followed by a phase of test where each answer is scored for an evaluation of the student's performance. Such a dialogue management system may be considered to possess a "default" strategy. Sometimes these strategies are flexible. For instance, in the middle of an exchange the student may interrupt his teacher and ask for explanation about a topic.

Similarly, a typical Information system includes a dialogue management subsystem that controls the dialogue with the user. This subsystem has a default strategy that may sound as follows: "after an initial phase where the user is required to enter his name, password, and other administrative information, the user will be expected to put queries and the system will answer each query; if the user asks for help, then a help module is called and, once finished, the system returns to the previous point".

The default strategy of a Design system may look as follows: "after an initial phase where the user is required to enter his identification (name, password, and other administrative information), the user will be expected to issue orders and the system will execute each order; if the user asks for help, then a help module is called and, once finished, the system returns to the previous point".

In brief, Tutoring systems are the ones where the user plays the role of a student and the system plays the role of a teacher; Information systems are the ones where the user plays the role of the partner-to-be-informed and the system plays the role of the informing-partner; design systems are the ones where the user plays the role of the issuer-of-orders and the system plays the role of the executor­of-orders. Concluding: we may associate default roles of agents to dialogue types.

2.3 Inverted Dialogues

One may revert these default roles. A system may play the role of the student provided he is given the knowledge of a student. For instance, a Data Entry is a kind of inverted Informative dialogue where the system queries the user and the user answers queries therefore causing the system to update archives. Knowledge Acquisition dialogues are Informative dialogues; if a system would be able to engage in such a dialogue with an expert, we would be able to talk of an inverted Informative dialogue (similar in pragmatics terms, to a Data Entry dialogue).

Inverted dialogues are therefore considered to be those Human-Computer dialogues where the traditional, expected, default role of each partner is inverted. We have called those dialogues "inverted" because we thought mainly of teacher­learner situations, but real situations may be more complex and deserve a classification of more than two roles potentially played by partners in dialogue. For instance, in (Ouin 94) we have a classification of 5. x 4 = 20 potential pairs of roles that may be associated with the Human and with the Artificial agent. In general, we will continue to call Inverted Dialogue that dialogue where each partner is aware of the fact that he and/or the other partner may play consciously for a particular purpose a role different from the one expected by the nature of the dialogue. .

Let us consider the case of Tutoring Systems. A system may play the role of a student and engage in an inverted dialogue provided it has the knowledge of a

The "Natural Laboratory" Methodology 187

student. Which student? A hypothetical student or a real student? If the query of the user-teacher is equal to the query the system (in the role of a teacher) had previously put to student S, then the system may play back the recorded answer of S to the query. Otherwise, the system may use a student model - that is a kind of knowledge base - to generate an answer from the model.

Therefore we envisage at least three types of abstract "students": human students, records of human students and students simulated by a model. One may think of a fourth type, i.e. a record of a simulated student, but this does not make much sense unless the model simulating the student has an internal learning component, i.e. produces time-varying behaviours. Similarly, a teacher may be a human, a recorded human or a generated tutor with a default tutoring strategy such as the one described above (presentation, assessment, test).

2.4 The Kernel of the Methodology

In the NAT*LAB methodology supported by NOBILE we have evaluated the feasibility and assessed the potential value of a dialogue management strategy supporting sequences of Human-Computer interactions (in the domain of Tutoring) that include "inverted dialogues" for the purpose of student model acquisition. In particular, we have evaluated two particular sequences, called Inverted Dialogue Method 1 and 2 (101 and 102).

As an example, consider the simple case that a tutoring system querying a student to add 10 to 5 receives 50 as an answer. How can we decide to infer (abduce) from this single information that the student exchanges addition and multiplication or, instead, that he simply substitutes the first figure of the first number with the figure in the second number or any other potential, compatible incorrect knowledge about the algorithm for addition?

A (human, female) teacher may wish to look at the student (human, male) behaviour and play back the system with reverted roles loading the student record (Step 1, Fig. 17). After a number of interactions with the student record the teacher may have elaborated a model, among the many possible ones, that is certain or, at least, the most plausible one. The teacher, from her consciousness of the model, may be able to foresee the student behaviour but may not be able to formally describe the model, or she may not be willing to do it for several reasons. Or else, she may be willing simply to check first if the model has already been "programmed" in the system, or if one of the available student models may be refined to fit her own discovered model.

In order to confirm the match between her (conceptual) model and one or more of the existing, available, formal, runnable student models in the system, the teacher would now start the direct dialogue (step 2 in Fig. 1) where the role of the system is that of a teacher putting questions, and the role of the user (the real teacher) is that of the student modelled according to her understanding of the

7 Notice that in Fig. 1 the single teacher plays the role of the teacher in step 1 and simulates that of the student in phase 2. This "student" may only be simulated by the teacher, once the teacher has built a conceptual model that is plausible to explain the student mistakes she saw during phase 1.

188 S.A. Cerri

Step 1: Acquiring the Model Step 2: Confirming Understanding

~ ~~ ~ \:::::.,J ~ ~ \::7

0 0 ~ 0 ~ 0 ,,~~ 0

0 • R 0

4 .P ~J!I )!l1? $ Fig. 1. A two-stage model ac..,uisition dialogue

student's behaviour she has previously examined. The user (real teacher), playing the role of the student in Step 2, answers questions as if she was the student; these answers are compared by the system to the answers that are generated by each of the student models available and if one model passes the test, this is a candidates to become "the" model , in the sense that its behaviour fits with a teacher's expectations. The selected model is therefore considered as one effectively verified by the concrete expert's evaluation.

If none exists, then a highly specialised model editor is offered to the teacher­user to edit one of the available models so that possibly a new model will be available for the system in the future by refinement of one of the existing ones. If the teacher's model cannot be derived refining an existing model, the teacher has to ask the help of the system designer (by appending notes to the dialogue trace, or as she wishes). Therefore, refinement on initiative of the expert is a direct manipulation, when possible, otherwise it proceeds in the more traditional way.

Model·based acquisition

IdenldlClbon T eache Leame r

~m Confirmation Teache Leamer

m~

Key: rHl Domain ~ specialisl

~ Human tmJ studenl

["rnl Studenl

l§]J record

fl@l Student gmodel

fIDlTutoring g strategy

Student·basei acquisition

Teache Leamer

~ ~ Teache Leamer

~ ~ Teacher Leamer

~

~ m Fig. 2. Role assignments in model-acquisition dialogues

Behaviour acquidion

lCenbfation

Confirmation

t.todoI malchng

t.todoI verificltion

The "Natural Laboratory" Methodology 189

While the example is extremely simple, it is very complicated to describe. It suggests how simple dialogues may be complex to manage. Although the two "inverted dialogue" strategies (or methods) implemented and experimented in NOBILE, i.e. ID 1 and ID2 are pre-programmed variations of these sequences of interactions, more complicated model acquisition strategies have been identified 8.

For instance, see Fig. 2 interpreting it with the help of the interpretation key.

2.5 Focus on Diagnosis

Although NOBILE supports any kind of tutorial dialogue, the specific advantage of it can be appreciated in situations where error analysis, diagnosis and remedy of misconceptions makes sense. As we will see, diagnosis is a quite pervasive goal in problem solving, and therefore we will suggest generalising the methodology to diagnostic situations outside of the tutoring domain.

2.6 Human-Computer Dialogues in Nobile

The dialogue management strategies described above have been implemented by abstracting the logical association of the teacher agent and the learner agent with respect to the human agent and system agent. In the specific case that the system agent is associated to the student, a further distinction can be made between a student record and a model. Therefore, a human student may interact with the (traditional) system tutor via the system; and a human teacher may interact with a student record and a system (student) model; further: at any time any human user may switch the role.

In Fig. 3 we present a classification of potential fillers of the teacher and of the student roles in NOBILE. The same figure suggests that some of these logical, abstract agents may be generic with respect to the domain of dialogue, others are domain specific.

2.7 Other Functionalities Emerging from Inverted Dialogues

We conceived the sequences of direct and inverted interactions described above in order to develop the methodology for the specific functionality of acquiring knowledge about student models.

However, other potential functionalities are offered by the "inverted dialogue" approach, such as the ones outlined briefly in Fig. 4, applicable when the Tutoring System includes a knowledge based system for teaching Diagnosis.

8 The results emerging from CHAOS [Simone 93] may be integrated at this point in order to extend the potential definition of dialogue management methods including multi-agent conversations. For example, the model verification strategy may be scheduled to be requested simultaneously to more than one expert-teacher, each supposed to be particularly skilled in different aspects of the domain to be learned.

190 S.A. Cerri

(.) ::: 'u Q) Q.

W c: 'iii E o c

.2 13 QI Co

en c 'iii E o c

Fig. 3. A classification of potential fillers of the teacher and of the student role in NOBILE.

These functionalities show that the approach allows te relaxation of the unrealistic constraint that either the user is a student (assumed to be unskilled in the domain of expertise) or the user is a teacher (supposed to be a perfect expert). On the contrary, a spectrum of expertise may be associated with the user, who can profit from the system for different purposes each of which requires a different degree of expertise.

User Most Expert

Least Expert

The "Natural Laboratory" Methodology 191

D' 1 la ogue et 0 M h d I Computer = Student

User = Teacher +

I User = Student Coml!uter = Teacher

I Computer = Student LJsll[ = Ie:af:ibet

+ I User = Student Computer = Teacher

Computer = Student llsec - TeaCher

+

User = Student Comptlter :::: TeaCher

User = Teacher Computer = Expert

I I I I

Functionality

Student Model Acquisition

Training in Diagnosis

Reasoning about Misconceptions

Discovery Learning

Fig. 4. Examples of functionalities offered by the Inverted Dialogues for purposes also different from student model acquisition. The "user" is classified according to his/her level of expertise in diagnosis.

2.8 Object Oriented Architecture of NOBILE

The "agents" referred above recall an object oriented architecture of the system 9.

An example of one of the message passing algorithms used by objects in NOBILE appears in Fig. 5.

Further to other properties we have briefly indicated before - such as abstractions at the domain level - this architecture allowed us to explore a few more potential generalisations that I wish to present in the following section.

2.9 Human-to-Human and Computer-to-Computer Dialogues

In particular, as the Dialogue Management System allows - by definition - the association of logical roles with physical agents dynamically, it allows also unfamiliar dialogues such as "computer-to-computer" dialogues and "human-to­human" dialogues to be run.

9 NOBILE runs on a Macintosh. It is written in Common Lisp. KRS (Van Marcke 88) and GTE [Van Marcke 90]. Because both KRS and GTE are in Common Lisp. the kernel is easily portable to any platform. KRS and GTE are joint property of Knowledge Technologies and DIDA*EL; both systems have been ported to UNIX and Windows -based platforms. GTE is currently extended with partial support of the CEC DELTA DISCOURSE and OSCAR projects.

192 S.A. Cerri

r=:::l See Problem ~ ~ .~

~:Iern t 1 Problem ! Present

[prob~m ) (TeaCher) Available J Problems All Problems S · 11presentatlon

tnng +String

(context I

Fig. 5. Message passing in NOBILE. 1. a System Strategy (method) asks the teacher agent for a problem; 2. the teacher agent asks the domain for available problems; 3. the System Strategy asks the learner agent to see the problem; 4. the learner asks the problem to present itself; 5. the problem asks its context for the information it needs to present itself.

For instance, a specific System tutor may be ordered to run a dialogue with a specific student record in order to gather data that have been lost; or run the set of all records in order to make statistics or check some particular responses in an automatic way. A System tutor may run one or more student models for similar reasons. Each message exchange may be monitored and reused for particular purposes (e.g.: statistics, case analysis, etc.) that may be important also for other agent types in the Educational scenario (e.g.: administrators, trainers, teachers of other disciplines, peer students, etc.).

Similarly, a human expert playing the role of a teacher may engage in a dialogue with a human playing the role of a student. The physical location of the two humans may be the same or different; in the first case the two humans may coincide, and at each turn the same person is expected to change role.

While the advantages for potential applications to eM Human-Human Dialogues at a distance are intuitive, if the two agents coincide in the same location the purpose of the dialogue may require a justification. An example of such a justification may be the need for a teacher to evaluate the System strategy and eventually annotate, for the system designer, specific suggestions. This is facilitated by the strong contextualisation of the teacher's annotations, so that, once the designer will re-run the annotated dialogue, a minimal effort will be needed to understand the teacher's suggestions.

The dialogue may be monitored by the system, so that interesting aspects may be later identified and further processed. to

10 Monitoring message exchange is crucial also for achieving co-ordination in group work (Simone, 93).

Teacher I:il Humin Expert

~ tOU

The "Natural Laboratory" Methodology 193

Learner

~Humln_t ~PWoQ,--

Fig. 6. The Nobile Control Panel suggests how settings may be selected in the system by the user in order to define roles and the associated parameters in the dialogue. Notice that (a) as a Teacher, the user may select a Human Expert; a System Tutor (currently only Alice, a toy system teaching foreign language conjunctions) or an IDM: Inverted Dialogue Method, i.e. IDI or ID2, to be applied on the System Tutor selected; (b) as a Learner: a Human Student, a Student Record or a System (student) Model The Student button pre selects the "default" setting, i.e. System Tutor as Teacher and user as Learner. The Expert button enters the browsing and editing tools of NOBILE.

2.10 Computer Mediated Dialogues at a Distance

CMD refers to the computer support to dialogue. The above described NOBILE system can be described as a CMD system that supports various types of dialogues in the domain of Tutoring. The system in fact acts as a mediator and a communication manager between agents that may activate a "traditional" Tutoring System as the selected filler of System tutor, or sequences of Direct and Inverted Dialogues, such as those associated to the Inverted Dialogue Methods ID 1 and ID2.

This Tutoring System will have a few defined Student Models and, as a consequence of interactions, more can be added also by the user when he activates the corresponding editor. Usage of the System Strategy also generates Student Records that are added for further use. The potential value of Inverted Dialogues Models (lDM) and pre-programmed sequences of different types of dialogues has been presented. These pre-programmed sequences become on their own a specialisation of "teacher agents", in the sense that they are strategies for conducting the dialogue by the abstract teacher agent with the abstract learner agent.

In the case that Teacher and Student are both humans, their physical locations do not necessarily coincide: the abstract agents may be assigned to remotely located sites. Further to the previously described potential functionalities of the methodology, one may use the CMD system also to monitor the message

194 S.A. Cerri

exchange of the agents: any monitoring function can be realised by extending the existing one, that builds student records.

In Fig. 7 the high level architecture of the system is presented, in order to clarify the feature of NOBILE as a mediator between two abstract agents, each of which is instantiated dynamically by selecting the corresponding properties.

Teacher 1-----·1 Student

Teacher Agents Leamer Agents

Fig. 7. NOBILE as a Computer Mediated Dialogue manager

3 Abstractions and Generalisations

3.1 Requirements for Abstractions

Although more generic "NOBILE-like" systems may be welcome for a number of reasons different from the following one, we think that the main value added by the NOBILE-supported, NAT*LAB methodology has to be found in problem solving tasks requiring diagnosis as a component. As far as Teaching is considered to be such a task, the methodology is applicable to Teaching.

Diagnosis is an abductive process: from symptoms one has to develop a model of the causes of symptoms without having access inside the malfunctioning system that remains a "black box"lI.

Because automatic diagnosis is difficult, even if the system is not as complex as a student, the NA T*LAB methodology assumes that part of the process may be fruitfully carried out by a human expert. But how can a human expert verify the adequacy of his diagnosis? The proposed methodology suggests the expert "to

II Notice that the modularity of systems developed using Object Oriented Programming methodologies relies on the same assumption: object specifications are identified in terms of behaviour at the interface, not in terms of the implementation algorithms inside the "black box".

The "Natural Laboratory" Methodology 195

behave as if he was simulating the malfunctioning system" in order to check the correspondence of his conceptual model with the real malfunctioning system's behaviour, or, if available, with behaviours generated by the computer runnable models of the system.

Therefore, for the purpose of diagnosis, inverted dialogues allow the separation of the model conceptual identification and verification -that is a task better done by the expert - from the model description, specification and implementation -that may require the expert to collaborate with the system designer, though this is not always necessary 12.

In case the expert is able to edit a new model by refining an existing one, the advantage is relevant: not only the traditional knowledge acquisition process is simplified, but the expert is confronted with a context relevant for him, a lexicon significant to his expertise, a situation (in our case a teaching-learning situation) that is specified in detail, unambiguously. A very different environment from the usual one of interviews. A much more "natural " environment.

In summary, a generic methodology emerging from the one described above may be useful if applied to situations that have the following properties:

a. There is the need for model construction from external evidences (e.g.: errors by students, symptoms in patients, bugs in programs, malfunctioning in circuitry'S, ... ). The evidences are not necessarily symptoms of incorrect functioning, they may also be symptoms of alternative, unforeseen but correct functioning (according to an underlying structure that is not known).

b. External evidences are available (a student may be queried, a patient may be analysed, a chip may be tested, a program may be designed with traps and hooks that allow to test its variables at run time), though the internal structure is not accessible.

c. There is a non empty set of (partial) models initially available (as the initial set of student models in NOBILE). Models are runnable simulators of deviating behaviours.

d. At least one human expert is available. The human expert is supposed to be "better", at least in some way, as a model construction agent, than any available artificial agent.

In Fig. 8 the flow of the abstract methodology for system design and refinement is shown.

12 Notice that the specialised editor for refining student models in NOBILE is applicable under the assumption that a teacher, not expert in programming, is able to "author" -under limitations - a program (a knowledge base). This assumption is fundamental for projects in advanced authoring systems, such as GTE. The ultimate goal is to offer experts in any domain, customised tools for direct authoring in that domain. Once these tools will be available (and evidences are promising) then this authoring facility of NOBILE may be enhanced in a significant way. Conceptually, editing a correct or incorrect knowledge base are similar tasks; practically they are feasible only if the editor power fits the author's authoring expertise.

196 S.A. Cerri

Initial Domain Representation

Initial Set of Models

Extended Set of Models

Dialogues

Final Set of Models

1. Construction of domain representation by designer

2. Development of minimal initial set of domain models by designer

3. Incremental refinement and extension of error set by a series of dialogues.

4. Inclusion of the models developed in the finished system.

Fig. 8. System development in the Natural Laboratory

3.2 Scenarios for Generalisations

Let us call System strategy any program associated with the logical Teacher agent.

3.2.1 Diagnosis of Patients; Diagnosis of Circuits

First the "System strategy" puts queries to the "human patient II artificial circuit"; data are gathered in the "patient record II circuit record" and one or more "patient models II circuit models" are generated as a possible diagnosis.

Then a human doctor /I human expert in circuits looks into the patient record II circuit record.

If and when he has build a conceptual diagnosis he activates the inverted dialogue phase, where the System strategy puts to him questions, and he "acts as if he was the patient II malfunctioning circuit" he has just diagnosed. During this

The "Natural Laboratory" Methodology 197

phase the human doctorlhuman expert in circuitry verifies his diagnosis by checking the behavioural match, the effects, not the causes.

A potential subsequent match with automatically diagnosed error models and model refinement may be thought in the same way as it was described before for the case of student models.

3.2.2 Verification of a Simulator; Verification of OOP Modules

We may envision Computer-to-Computer dialogues with the simulator as a "student", the real device as the expert component of the System strategy; the human expert as a "model constructor" of potential malfunctioning of the simulator.

In a sense, an object in Object Oriented Programming (OOP) may be considered a simulator. It performs like any other algorithm, but one is not expected to have access into the internal OOP methods implementing the algorithm; the only thing of interests is the behaviour at the interface. The scenario is then similar to the one envisioned for other simulators.

3.2.3 Plan Debugging in the Access to Information Systems

Suppose the scenario is to be analysed, reflecting a common case in the practice of Information Systems, that is described in the following.

During the first phase an agent A (the person that has an informative need) contacts another agent B(the person supposed to help) in order to identify the specific Information System that may be the information source he needs. A asks B to load a specific Information System - say I and puts I his queries.

After an Informative dialogue session between A and I, A is not satisfied because he either did not succeed or he succeeded partially. Because the trace of the dialogue is recorded as "user record" (corresponding to our student record), A may ask B to engage in a dialogue with the user record as if B was I (that is: try to play the role of I).

During this phase B may check (on another window on the screen) if his own answers would "fit" I's answers. If it is the case, then A's lack of satisfaction may be due to A's incorrect formulation of the queries, if that is not the case, B may discover that in order to fulfil A's need, I is insufficient and another (perhaps complementary) Information System has to be substituted to I and indicated to A for a more successful search. What in fact B may understand from the trace of the session is the plan A has for fulfilling his need using I. .

Because automatic plan understanding is as difficult as automatic user modelling, the methodology may help a human (B) to recognise a plan, eventually debug a few steps of the previously executed plan and transfer to A the corrected plan to be used. But: until now we have only suggested B to look at a trace of the dialogue between A and I, a quite straight suggestion. Let us check if our methodology can help us better.

Once B thinks he has recognised A's plan he may debug it in his mind according to his knowledge about how to access Information Systems, and particularly the specific Information System I. The "correct plan" - in the

198 S.A. Cerri

Tutoring metaphor - is similar to a "correct student model" conceived by a skilled teacher that has previously looked into a student record. As before it was the case, the first validation of B's correct plan recognition and debugging may occur without formalising B's conception, but may be achieved by a session where B acts as if he would use the incorrect planning algorithm for satisfying A's queries by accessing I or another Information System.

Again, B's understanding of A's incorrect behaviour is validated by behavioural matches, not necessarily by matching a formal behavioural model. B may ask A to put B other queries of the same kind: B would collect the queries and, once he has debugged them, put the corrected sequence of queries to the adequate Information System, get the answers and present them to A. If A is now satisfied, then B may describe A in some more or less formal way the correct plan to follow in the future.

The scenario reflects an application of the Methodology that uses Human-to­Human dialogues. The mediation of a computer in those dialogues seems important for two reasons: as a monitoring and archiving system (even if no formal model is build for the incorrect planning by A-types of users) and as a semi-automatic plan debugging system, if any model is build or refined during the process.

3.2.4 Collaborative Software Engineering

Another example may be that of an office, where Employee A issues a request to Employees B, C, ... about a goal (a complex CAD software to be built) and some resources to be used in order to achieve the goal. We assume that A is a manager and B,C, ... are system programmers. A knows programming, but does not know the CAD programming language used by a few programmers, the Windows programming language used by others, the C++ programming language used by yet other programmers.13 The communication problem is twofold. First: B,C, ... have to "understand" what A wishes (and orders to do). Usually there are software specifications for this, but not always. Then, A has to "understand and verify" B,C, ... products (the programs) and their integration. Suppose all agents work at a distance through a network.

The first dialogue between A and each of his programmers (B,C, ... ) may consist of a presentation, a series of assessment questions and a test. A would play the role of a Teacher; B,C, ... that of students. From the dialogues, only "student records" are collected. Then A looks at the dialogues reordered by question. For each question, A looks at all the answers (of B,C,D, ... ). If some answers were deviating from the ones he was expecting, the corresponding programmer did not understand or assumes information that was not assumed by other programmers; this misunderstanding or unexpected reaction may be due to

13 In the example (as well as in the previous ones) we assume full collaboration between agents, in order to avoid side effects, such as "I do not tell you X. even if I know you need to know X in order to solve our common problem, because I do not wish you to take advantage from the knowledge of X for your own purposes".

The "Natural Laboratory" Methodology 199

simple slips or to deep discrepancies, for instance a different lexicon. One concrete example of lexical misunderstanding in programming may be the name "object" that usually refers to "object in Object Oriented programming" (call it: object#l) but also to "datum associated to a variable name (constant or procedural object)" (call it: object#2). Suppose C is the programmer that produced one unexpected, deviating or incorrect response. Suppose A comes to the hypothesis that C believes that "object" denoted object#l while A's intention was to talk about object#2. Naturally, A may meet C and fix the problem. Or call C by phone. Time consuming operations in the case that the hypothesis is incorrect. A better verification strategy consists in simulating (by A) the consequences of the inconsistency: if A plays C in a CMC dialogue where the system puts the same questions A had put C previously, then the inconsistency may be disconfirmed or confIrmed by subsequent matches of the two sets of responses.

Once A's understanding has been certified, it makes sense that A contacts directly C for remediating the problem.

4. Conclusions

Any potential generalisation of the methodology requires one to enter into the specificity of the application. What may be reused at an abstract level are two notions: a. the feasibility of potential controllers of the dialogues (programs) that guide different phases through different role playing by the partners according to the purpose; b. the integration of human and artificial agents in the processes, again, according to the purpose; and the recognized better performance of each of the agents in specific tasks. For instance: a human expert is better at diagnosing, even if he does not necessarily know all the detailed causes of malfunctioning: a doctor may produce excellent diagnoses, even if he would not be able to provide for all the physiological causes of the illness.

If any of those generalizations will be realized, we will have indicated a potential fall-out of R&D in Technologies for Human Learning for other applications. Apart from the interests of the new application, this seems important at the moment for our own domain of investigation. We do not wish to be pushed to a role that was never appropriate historically, i.e. an area of applications both for technologies and for pedagogical theories. Even if applications are important, our conviction is that significant Technologies for Human Learning cannot emerge from a purely applicative process, simply because we don't know much about Human Learning. Instead, we can develop theories and models of Human Learning with the help of experiments about human learning, where our experimental laboratory is mainly equipped with computers and networks. If that is the case, our attitude has to be mainly that of experimental research; therefore aiming at building models and theories and abstract and generalize them for our own as well as for other application domains.

200 S.A. Cerri

Acknowledgements

The initial idea of Inverted Dialogues as a knowledge acquisition methodology was generated by the author when he looked, in 1986, at a demo of BUGGY, an Intelligent Tutoring System developed by J.S. Brown and R. Burton at XEROX PARCo The first project, called "Laboratorio Naturale" reported a first application in PROLOG to teaching Italian verbs (Saetti & Cerri 89).

The development of the Nobile prototype described herein was made possible by the support given to the Nat*Lab project (Delta P70041D1016) by the CEC, and by Dida*El s.r.l., the main contractor. The author would like to thank Angus Mcintyre, the main scientific and technical contributor to the developments around NOBILE, who deserves also the credit of having produced all the intuitive figures about the system presented here and elsewhere. Thanks are due also the other members of the Nat*Lab consortium - Knowledge Technologies & the Vrije Universiteit Brussel (B), IDSIA (CH), Siemens AG (D), HiTec (GR), the Universities of Udine and Trieste (I) and the Universities of Lancaster & Leeds (UK) - and past and present workers at Dida*el and Knowledge Technologies (including Philippe Alcouffe, Franco Benetti Genolini, Horacio Arlo Costa, Elena Cheli, Anna De Censi, Nur Erol, Camillo Falcone, Claudio Guggiari, Rod Moyse, Nicola Olivetti, Marina Salvato, Roberto Serra, Karen Valley, Kris Van Marcke, Danielle Verstegen and Giorgio Vezzino) who have given help and suggestions.

The abstraction and generalisation scenarios, instead, reflect only the author's views.

Finally, a special thank to Carla Simone to have accepted this initiative (cf. Simone, 94), attempting at bridging the gap between two different - but hopefully synergetic - research experiences.

References

Boder, A; Gardiol" C. (1994) Building an Evolving Knowledge-Base from Computer Teleconferencing. This volume

Carbonell, J. (1970) AI in CAl: an Artificial Intelligence approach to computer assisted instruction. IEEE Transactions on Man-Machine Systems MMS-ll,4, 190-201

Cerri, S.A.; Cheli, E.; McIntyre, A. (1992) Nobile: object-based user model acquisition for second language learning. In: Swartz, M.L., Yazdani, M. (eds.) Intelligent Tutoring Systems for Foreign Language Learning, The Bridge to International Communication. NATO ASI Series F, Vol. 80. Berlin: Springer­Verlag, 1992. pp. 171-190

Cerri, S.A; Cheli, E.; McIntyre, A (1992) Nobile: user model acquisition in a natural laboratory. In: Jones, M., Winne, P.H. (eds.) Adaptive Learning Environments. Foundations and Frontiers. NATO ASI Series F, Vol. 85 Berlin: Springer-Verlag, 1992, pp. 325-347

Cerri, S.A; McIntyre,A(1991) Knowledge Communication Systems. Invited Paper at the 8eme Congres Reconnaissance des formes et Intelligence Artificielle, Lyon -Villeurbanne; 25-29 November 1991; Proceedings edited by AFCET; AFIA-ARC­IRIA

The "Natural Laboratory" Methodology 201

Clancey, W.J. (1986) Qualitative student models. Am. Rev. Comput. Sci. 1986 (1): 381-450.

Cuena, J., Garcia-Serrano,A., Verdejo,F. (1994) The Role of Knowledge Based Systems for Automatic Coordination in Distance Learning. This volume.

Guin, Dominique (1994) Towards Models of Interaction between an Artificial Agent and a Human one. This volume

McConnell,D. (1994) Learning in Groups: Some Experiences of Online Work. This volume

McIntyre, A. & Cheli, E, (1990) NOBILE: An Object-Based User Model Acquisition Shell. In: Cerri,S.A.; Whiting,J. (eds.). Learning Technology in the European Communities. Proc. of the DELTA Conference on Research and Development; The Hague, 18-19 October 1990. Kluwer Academic Publishers; 1992

McIntyre, A. (1991) A strategy for the design and development of complex knowledge-based systems. In: Proceedings of AI*IA Conference , Palermo, October 1991

The Nat*Lab Consortium (1991). Industrial feasibility of goals, methods and tools. CEC DELTA project P7004IDI016 (NAT*LAB) deliverable WPNmg 2. -NAT*LABIND4, (Public)

The Nat*Lab Consortium (1991). The Methodology for the Acquisition of Student Models .. CEC DELTA project P7004/DlOI6 (NAT*LAB) deliverable WpNmg.4-NAT*LABIND8 (Public)

Saetti L., Cerri S.A (1989). MAESTRO! e il laboratorio naturale In: Proceedings of DIDAMATICA, AICA, Bari, May 1989, pp. 217-225

Self, J.A. (1974) Student models in computer-aided instruction. International Journal of Man-Machine Studies, 6: 261-276, 1974

Self, J.A (1990) Theoretical Foundations of Intelligent Tutoring Systems. Journal of Artificial Intelligence in Education, 1(4): 3-14, 1990

Simone C. (1994) Supporting collaborative dialogues in distance learning. This volume

Ulloa, A, De Girolamo, A, Delaney, S. (1994) OSCAR: A System for Collaborative Distributed Authoring of Multimedia Training Materials. This volume

Van Marcke K.(1988) KRS: An Object-Oriented Representation Language. Revue d'Intelligence Artificielle, Vol. 1, No.4 1988

Van Marcke, K. (1990) A Generic Tutoring Environment, Proceedings of the ECAI-90, 1990

Winkels, R.(1992) Explorations in Intelligent Tutoring and Help. lOS Press, Amsterdam, 1992

16 Building an Evolving Knowledge-Base from Computer Teleconferencing 1

Andre Boder and Christine Gardiol

Neurope Lab, International Business Park, F-7 4166 Archamps, France, Tel: (33 50) 315 620 fax: (33 50) 315 630 E-mail: boder@neuropelab.unige.ch

Abstract. We present a model called JITOL (Just in Time Open Learning). The method underlying open learning in JITOL is based on the idea that collaborative learning is essential, particularly for the professional development of individuals. Then, it emphasizes knowledge stemming from the debates between professionals and provides methods for capitalizing on it by a recursive process, called the reification of interactions.

Keywords. Computer teleconferencing, heuristic, interaction, knowledge acquisition, knowledge bases, open learning, professional development.

1 Introduction

The ideas behind the JITOL model stem from several sources, including a DELTA experiment called JITOL described in (Lewis, R., Goodyear, P. & Boder, A. 1992). The central objective of the nTOL project is to experiment with and evaluate a learning environment for the professional development of individuals. There are two specific foci within the JITOL model: a) the human-human interaction process; b) the progressive construction of appropriate knowledge bases, built by constantly integrating users' requirements. They are called "Evolving Knowledge Bases".

The rationale behind JITOL is that punctual and structured courses do not correspond to the very nature of learning. A permanent and interactive environment is more adequate. Among other things, interactions yielding proactive attitudes of users allow them to make their own representation and strategies explicit. This, in tum will help find the appropriate knowledge to solve their problems.

In addition, today's state of knowledge is becoming increasingly complex, quantitatively difficult to apprehend and constantly changing. Therefore, it is more

1 Some elements of this text are taken from: Boder, A. (1992) The Process of knowledge reification in human-human interaction, in Journal of Computer Assisted Learning, 8, 177-185

Building an Evolving Knowledge-Base 203

and more a challenge to select the appropriate knowledge for individuals to enhance their knowledge or to solve specific problems. In consequence, only knowledge bases which progressively integrate users' requirements have a chance to be appropriate. This is true both for learners at the stage of initial training or for updating of highly-skilled professionals.

The evolving knowledge bases are based on the idea that representations of the interactions between participants in the system should be stored and thereby become additional knowledge resources for subsequent use. Since the interactions will mostly emerge from problems that arise in the course of individuals' professional work, indexing of these knowledge bases can be problem-oriented. In this way, users will benefit both from actual interactions on the network and from subsequent updating and customizing of resources. The process of building the evolving knowledge bases is called reification.

A major idea behind the concept of reification is that there exists a major source of knowledge which is not accounted for in the traditional scientific community and particularly in the world of professional applications. Namely, experience gained by professionals, which is not explicit and not formalized. The hypothesis then is that this knowledge can be accessed through professionals' interactions and should then be capitalized on with the reification process.

2 The Four-Windows Interface

The environment basically consists of a "four-window" interface in the form of a cartesian representation. One entry defines public versus private items. The other entry defines knowledge bases versus debates.

In the upper-left window is the Evolving Knowledge Base. In the upper-right window, electronic debates (conferences) take place. In the lower-left window, users are provided with a note-book. In the lower-right window, e-mail is available to them.

2.1 The E-Mail Window

This is a traditional electronic mail; it allows the users to communicate personally with someone else.

2.2 The Note-Book Window

This is a space where users may make and build their own knowledge, as they cut and paste anything from the three other windows. They may also take some notes.

2.3 The Evolving Knowledge Base Window

This is a knowledge base made up of texts, figures, scanned images and soon video-clips. Its main principles are: - the relevance (in relation to the professional needs of the users) rather than exhaustiveness,

204 A. Boder, C. Gardiol

- a modular approach, - the progressive construction of this knowledge base thanks to the "reification" of the interactions which take place in the electronic debates window, - an evolving principle: the knowledge is not fixed once for all, on the contrary it is open to discussion, - a training tool but also a professional tool to be used in the everyday professional practice (especially, thanks to the implementation of a computerized medical file about patients and to the presence of sixty typical images of diabetic patients' feet to be used as references for the medical practice).

The unit in the knowledge base is the "topic". Each topic is approached through a certain number of modules (which are the same for each topic). These modules are the following: - Presentation of the topic: brief description of the topic; - Points of view: a variety of points of view on the topic (scientifically validated or not), from different authors and experts. The goal is to present different opinions on the topic so as to provide material with which to discuss the topic in the electronic debates window; - Cases and situations: examples of practical situations and patient cases which illustrate the topic. The idea is to give concrete and practical material to understand the topic better; - Glossary: critical terms with the equivalent in the different languages of the partners, and with a definition of these terms; - References: bibliographical references with abstract; - Critical issues and practical solutions: most problematic issues raised by the participants in the electronic debates, along with different solutions to solve them. - Debate summary: synthesised presentation of the debate in progress within the topic. Analysis giving emphasis to the claim and the different people's arguments, counter-arguments and supporting evidences.

Apart from these different modules which offer different entry points in the topics, the Evolving Knowledge Base provides a "Semantic Search Network" which offers the users a semantic representation of a concept and which allows an inter-topic navigation (across the different modules of the topics in the Evolving Knowledge Base). The Semantic Search Network is based on the idea that many points of view, provided by cross-references in a hypertext-like environment provide a better representation. This is similar to semantic networks which help construct an overall representation of a problem or a situation.

Its basic mechanism is to allow users to "button-click" on words, which then generate a graphical representation of possible links with other modules or other concepts. The links are provided with semantic value. So one can access any related node by finding support or evidence for the given word, or an alternative view or yet an example.

2.4 Electronic Debates Window

Each topic in the debate corresponds to a topic in the Evolving Knowledge base. The relation between the knowledge base and the debates is thus very close, the first is a support for the second and vice versa. Some debates are launched thanks

Building an Evolving Knowledge-Base 205

to some reflective activities about typical situations. A synthetic version of the debates is available in the Evolving Knowledge Base.

Regularly, the interactions in the debates are "reified", in other words, they are analyzed, organized and put in the Evolving Knowledge Base. These information do not acquire however a fixed and final status; as a principle, they are open to re­discussion. This strategy allows to be as close as possible to the constant evolution occurring in professional practices.

3 Building an Evolving Knowledge-Base

3.1 The Reirication Process

The purpose of reification is to build up knowledge bases in such a way that knowledge can be easily retrieved and accessed "just in time". Knowledge must be relevant for professional use and therefore appropriately organized (most likely in a modular and task oriented form).

Hence, qualitative analysis of interactions is required, because it shows more clearly how content dependent knowledge is used and for which task a particular piece of knowledge is useful to solve a problem. If the final goal is to organize knowledge according to its use, detailed analysis of how content dependent knowledge is used in problem-like situations is mandatory. Such an analysis is in fact a conceptual analysis both of the content itself and of the strategies with which the invaluable heuristic phase of knowledge elaboration is achieved. Such strategies include comparing items, arguing, providing evidence, making new hypothesis etc. This is typically the process of reification.

In order to implement the whole process of reification, two parts are required; a) selecting relevant items from the file containing users' interaction, b) analyzing and re-organizing knowledge to build up the knowledge bases. This process has to be seen as a recursive one, since knowledge bases are constantly updated.

3.1.1 Selection

There are two aspects to be focused upon. One is the principle of selection of relevant pieces of knowledge. The other is the way selection is actually carried out.

The basic source is the file of captured users' interactions. From this file, items are left aside. Others are kept for analysis. There is of course a subjective aspect in the selection process. Nevertheless, it is important to remember the rationale, in order not to end up with too much controversy at this stage.

The purpose of reification of interactions is not to build up a scientifically formalized and validated core of knowledge. It is to capitalize on those items in the process of knowledge acquisition which have a heuristic value. This knowledge is dynamic by definition because it has a heuristic function. But it must be made explicit and therefore included int a knowledge base which constitutes the first step of knowledge acquisition. This step is usually eluded in science. When one deals with knowledge focusing on experience and professional expertise, this step is

206 A. Boder, C. Gardiol

particularly important and should not be eluded. In the elaboration of scientific knowledge, validation then takes place in a second step. But this is beyond the scope of nTOL.

A theoretical parallel needs to be made here. At the individual and developmental level, Piaget (1975) put forward a major process called "reflective abstraction", which helps focus upon one's own cognitive tools and recursively elaborate more sophisticated and more abstract pieces of knowledge. Reification in nTOL serves the same purpose both at an individual and at a collective level.

The key point in the process of selection is to dispose of an interdisciplinary team. The work achieved here is in fact rather paradigmatic of what is an interdisciplinary task. Precisely because the function of the evolving knowledge bases is heuristic, one does not want to use only domain specific criteria in the selection process. Otherwise, invaluable items sometimes epistemologically erroneous but heuristically fundamental would surely not be taken into account. But one still needs domain specific validation, to ensure scientific relevance. Consequently, collaborative work between cognitive scientists (including psy­chologists, linguists, etc ... ) and domain specific experts is mandatory at this stage.

Another issue is to decide when to actually proceed in selection. This depends entirely upon the nature and the state of the current debate. Debates in nTOL revolve around themes. These are rather well defined. This avoids long and ill­oriented debates. Some debates are very quickly and extensively nourished by users. Some not. If a debate starts to include a great deal of interesting ideas, identified as key issues (KI), then reification, and therefore selection will take place. Delays may vary between a couple of days and several weeks.

The selection process yields a set of issues, still in the form of interaction between several users. These must then be analyzed according to several parameters described below.

3.1.2 Analysis and Reorganization

The analysis is achieved on the basis of how material will be organized into the knowledge bases. There are three parameters; 1) usability of the knowledge bases, 2) rhetoric of the debate, 3) navigation in the knowledge-base.

3.1.3 Usability of the Knowledge Bases

Because the purpose of JITOL is to provide just-in-time knowledge to professionals, the first parameter is to define how easy and relevant it is to use the knowledge bases.

The "usability analysis" consists in defining with precision how a matching can be obtained between the needs of a category of professional in the domain concerned by the debate and the items which have been selected as relevant from the interaction file.

lt is typically a two-way analysis in which bottom up information from these needs guides the way key issues are indexed and the way suggestions for access and potential use of knowledge are made.

Building an Evolving Knowledge-Base 207

Indeed, the work consists mainly in indexing key issues in such a way that each of them relates to already known difficulties in a given professional practice. Suggestions are made as to how difficulties can be overcome. But elaboration of suggestions is restricted to a couple of ideas and consists mainly in referring to either other debates or other pieces of knowledge on the network or outside the network. References to other data bases or more traditional media such as books or scientific papers is particularly developed under these suggestions.

3.1.4 Rhetoric of the Debate

The second parameter relates to how arguments, counter arguments, and other rhetorical aspects have modelled the debate. It is important for the user to find out and keep track of how an issue has evolved in a debate, why it has taken such or such orientation and why it yield such or such conclusion. The reason why this is important also relates to the fact that such knowledge mainly has a heuristic value.

The rhetoric analysis has a lot of specific features most of which are not exposed here. Among them are features such as "refutation of an argument", "generalization from a suggestion", "alternative to an hypothesis", etc ...

Rhetoric analysis is a major element of reification because the logic of collective arguing in debates usually remains implicit, which prevents coherent elaboration. Also, debating with teleconferencing as a support is very different from face-to-face debates (McConnell 1990). In fact electronically supported debates lose some elements of face-to-face debates, but gain some other elements, not present in face-to-face debates.

Among those are two aspects. One is the fact that because debates are asynchronous, users have the time to think before actually engaging in the debate. Another one is the fact that users' interventions, in electronically supported debates, must be rather limited (debates with long interventions tend to induce loss of interest among users). Therefore, analyzing interventions with a rhetorical indexing helps reorganizing debates for further interactions, after feedback is taken from within the knowledge bases. This suggests that users must actually use both the debate facility and the knowledge bases in parallel.

3.1.5 Navigation in the Knowledge-Base

At the stage when knowledge is acquired, categories used to organize the elements of the knowledge do not necessarily correspond to well formalized and scientific criteria. It has been demonstrated that, while in the process of elaborating new concepts, even scientific researchers use common knowledge, usually based upon metaphorical ideas (DiSessa 1986).

The mental schemes mentioned above in fact playa role in apprehending new pieces of knowledge and navigating through a knowledge-base. They help linking new knowledge with already known phenomenon. Therefore, because JITOL is concerned with knowledge in elaboration, it is worth providing elements to categorize it in a form which corresponds to the way users actually apprehend it.

208 A. Boder, C. Gardiol

The major difficulty with this approach is that every learner uses rather different ways to apprehend the same piece of knowledge. Therefore, the method proposed is to provide a basic conceptual framework to guide apprehension of difficult key issues. This is done by a semantic help in the navigation through the knowledge­base.

The way this device is conceived is by providing a semantic search network which allows the different pieces of knowledge located in the different modules to be linked. Each individual may then search for links between pieces of knowledge by calling either an example of a concept (reached with a hotword in a hypertext­like style), or an element to support or yet an argument in opposition to the one suggested in the hotword.

4 A Medical Application

4.1 Context

There is a sector in the medical field, the one of chronic diseases, where the quality of the interactions between doctors and patients has a direct impact on the treatment of the latter (Assal 1989). Indeed, for this kind of patient who has to be treated and live everyday with their disease, they have to learn how to manage it. They have to know it, and to understand clearly what the doctor tells them.

This implies, for the doctor, a special approach where the bio-medical aspect is not the only one to take into account. The doctor has to think of strategies to allow his patient to understand and to manage his disease. Such a reorientation implies that the doctor should reconsider the type of knowledge which he may provide to his patient.

The critical issue is that the type of knowledge which is relevant is not one which is formalized and which doctors may find in classical medical textbooks or even online medical databases. In fact, the most relevant knowledge is one which stems from doctors' interactions, where personal skills and experience are central.

The JITOL model aims to create the favorable conditions that well allow interactions to take place and then to capture and capitalize on the knowledge stemming out of the interactions. The domain selected is the one of diabetes.

4.2 Reification in Chronic Diseases

Capitalization on experience and expertise in the field of diabetes appears to be particularly relevant, considering the conditions mentioned above.

Within this field, two major subjects have been selected, mainly because non­formalized experience is a rich source of knowledge: the issue of motivation of the medical team and the issue of prevention of amputations which is a major consequence of diabetes (Schoenenweid & Assal 1992).

The JITOL model has been applied to five medical centres in European countries, each of which is specialized in the treatment and education of diabetic patients. The leading center (in the Hospital of Geneva) is internationally known for its approach in the field. Before debates actually run, initial material is

Building an Evolving Knowledge-Base 209

provided by this center. Later on in the course of the debates, additional knowledge in included, stemming from the reification of interactions between medical doctors.

Usability of the new knowledge is particularly taken care of. For instance, consultation with diabetic patients requires a lot of attention and experience, especially for prevention of amputations. Computerized images of wounded feet are included in the knowledge bases and difficult key issues arising in the course of the consultation are carefully analyzed.

Results from the tests run in the first period of the application show that an incredible amount of informal experience (i.e. strategies doctors personally apply) come out and is shared by other medical team members. The reason is that, unlike these diseases requiring only bio-medical action, chronic diseases also require knowledge, which applies more broadly to the whole follow up of the patient. This knowledge also bears on the transferring of responsibility to the patient.

The conclusion which can be drawn shows that reification works particularly well in a case where informal knowledge plays an important role. If one generalizes from this conclusion, one can suggest that reification is powerful in cases where knowledge is at a stage of elaboration. This is true because then informal knowledge gained by making explicit both experts' expertise and learners' representation of a problem is mostly productive.

References

Assai, J.Ph. (1989) Education des diabetiques. Connaltre sa maladie et gerer son traitement: la formation medicale des patients. Encycl. Medic. Chir. Paris. Therapeutique. 25191 E20 11. 12 p.

Boder, A. (1992) The Process of Knowledge Reification in Human-human Interaction. In: Journal of Computer Assisted Learning. 8. 177-185

DiSessa, A. (1986) Towards an Epistemology of Intuitive Knowledge in Physics. Report of the School of Education. University of Californian at Berkeley

Lewis, R.. Goodyear, P., Boder, A. (1992) Just-In-Time Open Learning - A DELTA project outline. Occasional paper, NU1I92. Neurope Lab. Archamps.

McConnell, D. (1990) Case Study: The Educational Use of Computer Conferencing. ETTI 27.2.

Piaget. J. (1975) Equilibration des Structures Cognitives. Etudes d'Epistemologie Genetique. XXXIII. P.U.F. Paris. (Also: The Development of Thought: Equilibration and Cognitive Structures. Blackwell. Oxford, 1987).

SchOnenweid. C. & Assai, J.Ph. (1992) Le pied diabetique a risque: vers une prevention des ulcerations. STY 5. vol. 4. 289 - 295

17 Educational Scenarios for Telecommunication Applications

Peter Zorkoczy

EPOS International, Switzerland

Abstract. Attention is drawn to the crucial role which scenarios play in the analysis and design of technology-supported learning systems. Some examples of such scenarios are given which have been used in the derivation of general and specific learning system architectures. Directions for further research in the area of scenario construction and utilization are proposed.

Keywords. Education, training, telecommunications, distance learning, scenarios, training system architecture, technology-supported learning.

1 Introduction and Background

The rapid technological development and the wider accessibility of high-quality telecommunication links are poised to bring about a significant change in distance learning: it is now becoming possible to link together learner with tutor and learner with learner at a distance in a directly-interactive mode, cost-effectively. Contrast this with the situation just a few years ago, when distance learners had to rely on a centralized, largely uni-directional system, based on mailed and broadcast materials, and only occasionally supplemented by face-to-face tutorials. Opportunities for direct interaction, including group activities were few and far between, and usually implied a time and cost penalty.

Importantly, the change towards more direct human communication in the distance-learning process is happening at a time of increasing demand for people trained to work successfully in teams. The necessary communication and management skills can only be acquired through experience gained in realistic environments and involving direct interaction with other people. While book-based and stand-alone computer-based learning have proved quite effective in conceptual, knowledge-oriented courses, they did not properly cater for educational techniques aimed at fostering collaborative skills. The addition of real-time two- and multi­way telecommunication links to the conventional distance-learning environment may provide the missing ingredient.

Educational Scenarios for Telecommunication Applications 211

The potential for collaborative, interactive working promises a significant change not only for the learner and the tutor, but for the entire distance-learning enterprise. The course team approach to learning material development, successfully applied at the British Open University and elsewhere, stands to benefit from telecommunication-supported collaborative authoring methods. The management of learning, e.g. selection of appropriate courses, enrolment, financial transactions, enquiries, and individual problem handling, can be made more responsive and more convenient through interaction supported by telecommunication.

These changes, however, are today largely in the realm of possibilities in most distance-learning organizations. We are at the threshold but we have not yet crossed it. Interactive telecommunications are not yet widely or routinely used in distance learning. The investment of resources needed to make the change is significant. But the change in educational and working methods needed to make the most of these possibilities is even more significant.

This paper therefore takes the view that a thorough analysis of the potential effects of the changes, and of possible change processes, is needed before the threshold can be confidently crossed. The tool it uses is the educational scenario, and its immediate objective is the review of the role and organization of scenarios. It then goes on to describe the conversational aspects of some related scenarios. It looks at some specific scenarios used in past and current projects. Finally, it turns\ to the implications of telecommunication-oriented scenarios for research and development. The paper builds on a line of work (Zorkoczy 1989, Simon 1991) and that of many past and present co-workers, all of whom I wish to thank for the intellectual stimulus within a collaborative working environment.

2 Why Scenarios?

The well-established meaning of the word 'scenario' is a summary of a sequence of events or of a situation, including information about its setting, participants, their roles and interactions, etc. Its purpose is to help visualize and explore events or situations which, on the basis of reasonable assumptions, could have occurred in the past or could occur in the future.

First and foremost, a scenario is a communication device: it enables its creator(s) to represent and convey ideas about settings, relationships and behaviours in a concrete, easily visualizable, rather than abstract form. It is often employed in a creative group environment as a means of making specific the ideas of group members about design alternatives. A scenario may be developed and refined (e.g. made more realistic) in the course of its use by the group. It then comes to represent the joint visualisation of the group members of a specific way, or ways, in which their design may be used. This, in turn, leads to the use of the scenario as a "blue-print" against which the implementation of a design may be tested in the

212 P. Zorkoczy

course of its development. ("Is the behaviour of the system still compatible with the role allocated to it in the scenario?")

In the analysis of technology applications and, in particular, telecom­munication-supported applications, scenarios can be used to reflect organizational (social) structures in which the technology will be employed. They can also include, as part of the setting of the scenario, the underlying architecture of the technology. (Examples of such scenarios will be given below.) Thus, in one particular current project, CfA (Verreck 1993), the scenario approach is being used to define the sphere of interest of technology-supported educational and training systems, with the view to identifying those components and interfaces in such systems which may benefit from a degree of standardization.

3 Frameworks for Scenarios

Scenarios are usually developed individually, in the context of specific projects, rather than as general purpose reference models. In some instances, however, like the standardization-oriented project referred to above, the context of the project is so broad that it justifies the consideration of a whole range of scenarios. The two examples which are described below represent an early and a recent attempt to generate frameworks for scenarios involving telecommunication-supported learning environments.

3.1 Example 1: Scenarios for In-Company Training

These scenarios (Zorkoczy 1989) were developed in the context of designing a technology architecture for a large hi-tech company, with its own internal training organization.

The 'cast of the scenarios included:

- Learners - Tutors - Course designers - Course developers

Subject-matter experts - Project managers (of course development) - Education managers - Line managers (of learners and others) - Administrators - External people and organizations (vendors, customers, etc.)

The framework was expressed in terms of 'settings', or locations where the training-related activities take place. These settings included:

- workplace 'on-the-move'

- group-room

Educational Scenarios for Telecommunication Applications 213

large classroom training-site! campus

- national! international network of training sites.

The scenarios related to training-related actions and interactions which may take place in the various settings between the actors, or between actors and systems. As an example, group-rooms were envisaged as providing access to learning- or course-production systems similar to the ones that a learner, tutor, course developer, etc. would find in their individual work place-based environments. For group activities, however, the setting would provide appropriate support. For example, in the small group-room:

co-operative problem-solving, decision-making games, simulations involving team effort, e.g. large-scale (joint) fault diagnosis and repair,

- presentations by learners and tutors, possibly using multiple media, - recording the details or results of work, for later group review, analysis or

assessment - linking up with other groups for discussions, joint presentations, etc.

would receive support from computer-based and telecommunication-based systems. In the larger classrooms, the setting would give access to facilities that are too

complex or too costly to provide on an individual or small-group basis. This, in the conventional training mode includes specialist lecturers or tutors who may be available only on an occasional basis and therefore must be used to maximum effect, e.g. with the largest possible group when the occasion arises. Similar situations can arise with equipment, such as expensive computer equipment or digital switching centres which form the subject of the training.

Interlinked classrooms (small or large) extend the facilities to larger numbers of participants, allowing the flexible formation of groups of appropriate size for particular activities. In addition, interlinking of people with systems in a building-, campus- or multi-site setting provides support for activities requiring access to (multi-media) learning materials not available locally, resource materials, databases, help desks, etc.

In this example, the set of scenarios was used as a guide to the design of both the organizational and the technical architecture underlying the training operations of a company. In the next example, the scenarios relate not to just a single company but more generally to technology-supported learning environments.

214 P. Zorkoczy

3.2 Example 2: 'Common Training Architecture'

The framework used in this example (Verreck 1993» is in many ways similar to the first example, but it makes explicit some of the detailed characteristics of the 'macro' constituents, i.e. of

- people, - settings, and - facilities

which formed the basis of the scenarios in the first example. Thus,

people are further differentiated in terms of their age group and group size, the objective of their activities (e.g. training. course development. management. tutoring). their job function or roles. For example. in terms of group size the scenarios identify:

individual - minimal group ( 2 people) - work group (5 people)

project group (10 people) - small class (20 people) - class (40 people)

large class (60 people) audience (100 people)

- large audience (>200 people);

- settings are expressed in terms of location (e.g. work place. training centre. institution), and the time dimension (synchronous/asynchronous or real­time/deferred time activities).

- facilities are subdivided on the basis of materials (e.g. in-house developed. assembled. third-party developed. commercial; modular or monolithic, single­or multi-media) and support technologies. The latter are further classified along the lines of: - telecommunications, - computation, - storage, and - user interface.

By re-interpreting actual educational and training scenarios in these terms, the objective of this approach is to expose interfaces where some degree of standardization is already present (and assessing its impact) or where standardization is desirable in terms of market development. quality assurance and other considerations.

Although scenario frameworks can be very useful in technical standarization work for the purposes described earlier. it is questionable to what extent the detailed

Educational Scenarios for Telecommunication Applications 215

contents of the scenarios themselves fonn a suitable subject of standardization, in view of the rapidly changing nature of infonnation handling.

4 Conversational Aspects of Scenarios

The process of identification and development of realistic scenarios has no well­established methodology. A frequently used approach involves the observation and recording of 'representative' situations, and the extraction of features which are particularly relevant to the context in which the scenario is to be used. What is 'representative' and which are the particularly relevant features are, of course, a matter of judgement.

An approach which has been successfully used by the author in relation to scenarios for telecommunications applications for education and training built on conversations between educational and technical specialists. They aimed to arrive at joint decisions on the identification and selection of scenarios for the design of applications. Of particular interest to telecommunication-supported learning is the approach related to action coordination and decision-making within groups.

This is based around ideas of Flores and Winograd, and described by (Winograd 1988). The work builds extensively on scenarios and case studies reported by (Kaasboll 1986). The scenarios relate to medical staff working in a hospital in Norway. However, since much of learning is based on, or is re-inforced by action, understanding of the perfonnance and co-ordination of action is of direct importance to training. Apart from its use in the process of derivation of scenarios, the support of action coordination, as proposed by Winograd, is also relevant in the context of telecommunication-supported learning. Specifically, the concept of an action­oriented process as a negotiation, or conversation, leading to a "fundamental unit of work" is of interest in the implementation of remote-tutoring systems in distance learning.

5 Implications for Telecommunication-Supported Training

In learning oriented scenarios of telecommunication-supported training the assumption is usually made that the main role of telecommunication support is to re-create the classroom or campus environment for the remote learner or tutor. Thus, provision is made for:

- delivery of course materials, and - two- or multi-way conversations.

Conversation support can, in tum, be provided for

216 P. Zorkoczy

- real-time interaction (e.g. by voice, vision, data, or a combination of these), am - deferred-time interaction (e.g. by sending messages on a one-to-one or

conference basis, again through the use of one or more media).

The educational justification for the introduction of technology in such situations is that it makes possible the same modes of learning as are employed in the classroom or on the campus. This must be moderated, however, by the understanding of the differences in the capability of telecommunications to support

- the transfer of information, and - social interaction.

Telecommunication technology is becoming ever more efficient and cost effective for the former task (e.g. through phone and fax, terrestrial and satellite data networks, etc.)

For the support of social interaction the (not unreasonable) assumption is made that the more dimensions of human communication can be transferred simultaneously, i.e. through multi-media communication systems, the closer the approximation becomes to direct human interaction. Advanced current and planned telecommunication applications in training and elsewhere reflect this trend (e.g. through video-conferencing, screen-sharing with voice links, etc.)

But recent telecommunication-supported learning scenarios aim at more than just the re-creation of the 'conventional' learning situation: the learner and the tutor are envisaged as being able, for example

- to gain access to course materials originating from anywhere around the world; - to browse in and retrieve documents from electronic libraries specializing in

particular topics to a much greater depth than they could in most company, university or public libraries;

- to 'participate' in remote international tutorials seminars, conferences; - to have a 'conducted tour' of exhibitions and other events relevant to the topic

of learning, as they happen and wherever they happen; - to obtain and give advice and support on professional matters from an

international network (social network) of colleagues.

In coordination -oriented scenarios the coordinating organization is enabled to work in an increasingly distributed way. That is, its operations are not restricted to a particular physical location, or even a particular country. The telecommunication infrastructure provides for:

the electronic distribution of documents, - the support of meetings with dispersed participants (visual, voice and data

links), - remotely accessible databases of course materials, administrative data,

instantaneous response to enquiries, requests for help on specific issues, etc.

Educational Scenarios for Telecommunication Applications 217

- electronic financial and commercial transactions, etc.

In co-authoring scenarios joint production of documents (e.g. course materials) by authors located in different places has been a realistic possibility for a number of years. More recent scenarios include, among others:

joint authoring of multi-media documents by authors in different locations; desktop access to remote databases of modular, multi-media resource materials; browsing/sampling in such databases for modular materials from which new course configurations can be constructed;

- on-line ordering and immediate downloading of suitable modules; rapid authoring of short course segments to deal with specific problems (the materials then being available for immediate distribution to remote learners or tutors).

An example of a project in which such scenarios have been employed is ECOLE. This is a research and development project in the current DELTA Programme of the European Community. It involves seven European telecommunication network operators working together with three major computer companies (Bull, mM and Siemens) on the technologies which support collaborative learning. The scenarios which were used to develop and field test the support tools are described in greater detail in (Steinbeck 1993).

They all represent actual distributed company training situations and include:

- 'virtual teams' (for the training of project managers), - 'just-in-time training' (for field service engineers), and

'the distributed classroom' (for professional up-dating).

6 Research Directions

As noted earlier in this paper, the construction of explicit scenarios is but one step towards the utilization in distance learning of recent developments in telecommunications. It needs to be followed by a thorough analysis of the implications of various scenarios for the effectiveness and efficiency of the learning, coordination and authoring processes, both from the human point of view and in regard to the introduction and operation of the necessary technical infrastructure. Only when the analysis yields satisfactory answers to these issues can one proceed confidently to the implementation of the changes.

The tasks to be accomplished as part of the analysis include:

218 P. Zorkoczy

- the generation and exploration of architectures for the training enterprise, - the definition of changes in the job content, task structures and modes of

interaction of people involved in the distance-learning enterprise; - the definition of changes in the patterns of information flow between these

people; - the definition of the implications of these changes on the needs for particular

features of the technological support systems; - the generation and exploration of architectures for the technical infrastructure; - the definition of requirements for standards which ensure the quality and

compatibility of the new system components, as well as of the system as a whole.

References

Kaasboll, S. (1986) Intentional development of professional language. IFlP Working Conference on System Design, Amsterdam

Simon, C., Zorkoczy, P. (1991) Towards an Architecture of Technology-based Training Systems. Proceedings of COTEP-2 Conference, Enschede, The Netherlands

Steinbeck, W., Sandvoss, 1., Schuett, T., Dietel, C., Ardaens, P., Zorkoczy, P. (1993) A Multimedia Environment for Distributed Learning. Conf. Proc. "Multimedia Communications '93", Banff, Canada

Verreck, W.A., Collins, 1.R., Pratt, 1.M., van Bruggen, 1. (1993) A Framework for Education and Training Models and Scenarios. Deliverable 3, DELTA CTA Project, Heerlen

Winograd, T. (1988) A Language/Action Perspective on the Design of Cooperative Work, Human Computer Interaction, Vol. 3, No. 1,3 - 30

Zorkoczy, P. (1989) Development of an Instructional Technology Architecture and Associated Strategies. Technical Report, Centre for Electronic Education, Open University, Milton Keynes, UK

18 Developing a Tool to Support Collaborative Dialogues and Graphical Representation of Ideas

M.B. Twidale, T. Rodden, I. Sommerville

Computing Department, Lancaster University, Lancaster LAl 4YR, UK Email: mbt@uk.ac.lancs.comp

Abstract. We describe the development of the Designers' NotePad; a tool to support the rapid creation and refinement of ideas. The tool supports collaborative dialogues with features that enable idea fragments to be rapidly entered and revised. The core features are entities and links. Users are able to gradually add more structure while still keeping the freedom to rapidly refine and without being required to make premature commitments about classification. An iterative design approach was used involving frequent studies of the tool in use. Particular attention is given to the impact of the interface on the usability and effectiveness of the system.

Keywords. Computer supported cooperative work, rapid prototyping/iterative design, cognitive maps/design, brainstorming, collaborative learning, formative evaluation, computer mediated communication.

1 Introduction

The aim of this paper is twofold: firstly to describe the Designers' NotePad (DNP) and its potential as an advanced form of Computer Mediated Communication (CMC) for Distance Education, and secondly to describe the techniques used in the development of the DNP that may be helpful in the development of other CMC systems. With regard to the former, we believe that the DNP offers considerable potential in supporting collaborative dialogues by its use of 'graphical text', i.e. short text fragments positioned two dimensionally and supported by graphical notations such as various kinds of links, surrounding shapes and the use of colour. Using the system, idea structures can be developed and refined with great ease and speed and used as the basis of a discussion for further developing the ideas.

With regard to the experiences and techniques used, we wish to emphasise the substantial influence of the interface on usability and effectiveness of the system. This effect can completely swamp the effects of other carefully prepared features of the system to support learning. Therefore we advocate devoting considerable resources to the iterative design and evolving refinement of an interface that is easy to learn and easy to use.

220 M.B.Twidale, T.Rodden, I. Sommerville

2 Background

The Designers' Notepad was developed as part of a joint project between the Computing and Sociology departments at Lancaster University. The aim of the project is to investigate computer support for collaborative software design, in particular the early stages of the design process which existing CASE tools offer little support for. Part of the project involves undertaking an ethnographic study of the software design process in order to inform the development of computer tools which can better support this process (Sommerville et al. 1993). However ethnographic study and analysis is inevitably time-consuming and so it was necessary to develop a prototype system before the full results of the study emerged. Therefore a rapid prototyping approach was chosen to facilitate the incorporation of the results and recommendations of the study as they became available. The intention was to develop a simple to use core system that could be quickly tested on users. The experience of use then informed subsequent development.

3 Overview of the Designers' NotePad

Figure 1 shows the DNP interface. In order to make the learning of the system easy, there are a small set of core features which are sufficient to enable productive activity to take place in a short time. The basic unit is a design: rather a misnomer here, essentially it is a window into which idea elements can be placed. The design in the figure was used in preparing this paper.

The user creates an entity by typing in a design window. She may then move it with the mouse. Linking is done by selecting one entity and then clicking on the entity one wants to link to with the shift key down. We also provided Textnotes, which are note pads based on the Post-It Note metaphor (see Fig. 1) and allow users to attach one or more notes to an entity. These can be used for more textual comments, ideas, opinions, paragraphs of a final document, references etc. A variety of Textnote types are provided and users may define their own (including form-like structures). An entity with Textnotes has an icon attached (e.g .. the entity 'small scale' in Fig. 1) and the notes can be examined by clicking on the icon. Designs can be saved and loaded from a file and a paper report may be created containing a screendump of the design and a list of the entities and their Textnotes.

Each entity may itself be expanded to become a subdesign. A new window is opened and entities and links can be created in the normal way. Subdesigns may contain entities that are themselves subdesigns. A loose form of typing for entities and links is provided using colour, shape and labels. The user controls the degree to which she wants to use this typing. The type of an entity or link can easily be changed at any time. The use of typing here is closer to the concept of styles in a wordprocessor than types in a programming language. Eventually a software designer may wish to use more formal and rigorous typing in order to benefit from type checking, but that is appropriate for the later stages of software design; it too restrictive for the early stages.

~

~

~

~ "' o ~ J g ~

~.

~ ~ '" ciQ'

::s ~ '" 8- s· g ] ~. o ::s o ..., 8: '" '0 i

•• -1

-1 •••

••••

•••

• -8

E

J8

a

Set D

lrect

OlY

Mak

e N

ew D

esig

n

Des

ign

From

File

Top

Des

igns

All

Des

igns

Text

Not

es

SaYe

Det

ails

Load

Det

ails

Shap

e

defin

e by

neg

atio

n

~ 0'

'9. ::s O

Q

~ ~ 8 en

~ :g ::\. 2 f. ~ Sf.

e:.. ~ ~

tv

tv

222 M.B.Twidale, T.Rodden, I. Sommerville

Entities can be easily moved within and between designs. Additional features enable the design to be annotated. These include a framing facility for grouping related collections of entities. Groups can then be moved en masse to assist rearrangement. In order to control the complexity of an evolving design, a group can be 'pushed down' to form a new subdesign initially containing the members of the group and replacing the group in its originating design with a single named entity. In Fig. 2 the group of entities from Fig. 1 relating to requirements have been pushed down to form a subdesign, with a new entity 'System Requirements' replacing them in the parent design.

define by negation

Fig. 2. The subdesign created by pushing down a group from the main design

4 Educational Potential of DNP

The system was initially developed to support the early stages of design, which involves the following activities:

• Brainstorming • Ideas organising • Refining • Revising • Rearranging • Gradually adding detail

These are activities that occur during learning in many domains and so the DNP has far wider applicability as a tool for supporting learning activities. Furthermore, the development of a minimal core system for DNP as part of the rapid prototyping approach meant that the simplified features provided were necessarily not domain specific. Note that this claim for domain independence is unlike those normally made for domain-independent learning environments. These are generally complex systems where domain independence is due to a generic architecture that enables domain specific elements to be slotted in. For example they may include an expert system shell into which domain specific knowledge may be inserted. Our system's generality is due to the nature of the task to be

Developing a Tool to Support Collaborative Dialogues 223

supported (early idea organisation) and the simplicity of the core features. Also our focus is on the provision of appropriate tools to support users rather than the attempt to automate certain activities.

Thus the DNP can be regarded as an educational tool supporting concept mapping and other techniques for developing and refining ideas. Its key advantage is ease of learning and use. Vague, semi-formed ideas can be entered quickly and easily. If the user believes that all entries are easy to modify and correct, she need not concern herself initially with their ultimate form. This can prevent the block of premature commitment (Shipman & Marshall 1993). If she is free from commitment to form, she can begin by focussing on the broader issues. This inevitably involves some ambiguity, but the time to disambiguate is later on in the process. There are two reasons for the existence of ambiguity: firstly, in order to establish any form of overview, it is necessary to make gross approximations, 'broad brush strokes' are a useful metaphor. Secondly, in cases both of design and learning, the user is not merely keying into the system a pre-existing mental representation. The structure evolves as a result of its construction. So subtleties gradually emerge and are handled. As a map grows it eventually becomes clear that some terms are ambiguous and need to be refined (contrast this with the first reason where ambiguous terms are deliberately chosen).

5 Rapid Prototyping

During the early stages of software design, designers are particularly creative and necessarily handle ambiguous ideas. Little is known about the nature of this activity and consequently how best to support it. The same is true of learning activities involving creative idea organisation, and collaborative learning in general. Asking the users for requirements for such a system may be of little use even if they are experts in the activity. It is a notorious problem in the twin fields of requirements capture (Sommerville 1992) and knowledge acquisition (McGraw & Harbison-Briggs 1989) that experts have difficulty in articulating what they are actually doing, let alone what they would like from a system in order to do it better.

For both these reasons, the process of rapid proto typing for systems development is particularly appropriate. Our approach was to develop a very simple core system (principally; boxes, links, subdesigns and textnotes) and then to observe its use in authentic tasks. Although users find it difficult to articulate new requirements to support their areas of expertise, they can be most effective (and vociferous) in saying what is wrong or missing from a provided system or prototype.

In all our studies we were concerned to test the system on authentic tasks. That is, we asked our volunteers to bring along a piece of work (such as preparing a design, talk or paper) that they would have to do anyway. We asked them to try using DNP until they felt that it was no longer of use to them in solving their problem. We observed them using the DNP to work on this problem, noting their patterns of activity, what they found difficult and the features they would need to have added to DNP in order to progress further. After the study we discussed

224 M.B.Twidale, T.Rodden, I. Sommerville

informally with the volunteers their experiences of using the system and their comments on improvements.

The advantages of using authentic tasks rather than a uniform task of our own invention that we present to all volunteers are:

• It is naturally much more engaging for the user. • They ideally focus on their problem rather than our system, which is how the

final system will be used. If the interface is clumsy and fails to allow this focussing we can spot what needs correcting.

• Real world problems are different from artificial tasks: they are messy, do not fit into neat classifications, always contain exceptional features, are open­ended, contain elements of ambiguity and have a history and a future (so the system must cope with existing documents, ideas etc. and provide some product that can be used in later design stages or activities).

• There are inevitably implicit assumptions about the nature and style of use (and about the user, task etc.) in the design of the tool. In making up test problems, developers are in danger of incorporating these same assumptions. Thus the study will fail to reveal them. Outsiders' problems need not contain such assumptions and so serve as a better test.

This approach to development draws heavily on the work in participatory design; and approach to general computer systems development involving close participation with end users (Ehn 1989). Informal testing is cheap and relatively easy to arrange, but it should be begun as early as possible in the design stage and be repeated frequently throughout the continuing evolution of the system.

6 Gradualist Development

A consequence of the prototyping approach is that improvements to the system will be incremental. This, coupled with the need for a very simple core system with which to begin the testing cycle, implies a need for a gradualist strategy for testing and developing the functionality of the system. That is, we began by studying the use by individuals and small groups of users simultaneously working on the core system at a single workstation. This revealed the grosser interface errors and the improvements immediately needed for enabling progress with such activities. The changes were incorporated and re-tested.

For the early testing of the system it is quite in order for the developers to test it themselves. This might seem somewhat futile, but provided the activity is authentic (i.e. independent of the need merely to test the system) it can yield valuable results in a very rapid and economical way. For example, we used the DNP to plan the early stages of talks, reports and papers. Testing on developers is acceptable at the early stages because we are seeking to detect failure, not success. We can confidently predict that any feature that developers find difficult or clumsy to use will also be at least as awkward for normal users. In all cases interface failures 'scale out' to more complex contexts, although there is no guarantee that successes will. In the process of scaling out, new features and functionalities will certainly need to be provided, but there seems little point in worrying about these until the more basic problems detected in the simpler cases have been remedied.

Developing a Tool to Support Collaborative Dialogues 225

Later testing of the system involved a process of 'pushing out' its coverage to more general contexts. This includes different activity types brought along by volunteers and different user types who will share fewer implicit assumptions about the DNP. So far we have had the following kinds of volunteer: system developers, people involved in the project but not directly involved in its development, postgraduate computing students, colleagues from Sociology, visiting academics, undergraduate computer scientists and undergraduates from other disciplines. The tasks that users have used it for include software design, project planning, organising research ideas and preparing essays, papers, reports and talks. The DNP has now been installed at a number of other sites and we are waiting to hear comments and criticisms from users who may not share our assumptions about what a tool must provide to be effective. We have also pushed out the context of usage by studying cases where the same groups and individuals continued to use the system over a number of design sessions separated by up to a week.

In all cases, observations of practice lead to requirements to better support the particular activity. In addition it may yield further requirements for the simpler cases considered earlier. Only when the most glaring errors and omissions are identified and remedied is it possible to spot the more subtle ones that are nonetheless crucial in their influence on usability. Similarly, it is easier to spot an issue in the simplest context in which it can arise, rather than always testing in the ultimate intended context which is usually an environment with so many issues that it hard to distinguish between them and their relative importance.

Fig. 3 illustrates the idea of extending outward the base of testing. Local, cheap and easy to arrange tests on a basic system can catch the grosser errors and omissions. Only when these are eliminated can the more subtle ones become apparent. Some of these can also be detected by the same types of user. But others (to the right of the first dotted line) can only be detected by extending the user base and context of use. So we then extend testing to more diverse users who can catch even more subtle errors (but also others that our limited local testing could have detected but just happened to fail to do). Each type of user can detect errors and omissions to their left in the diagram, but generally will only be able to detect the more subtle of them after the others have been eliminated. This procedure can continue for any number of layers of increasing generality; the figure only shows three for simplicity. Each time the system is changed in the light of the testing, re-testing is necessary both to find the next most subtle errors and omissions and to ensure that the change in the system has not introduced new problems. An additional benefit of this approach is that at all stages after the development of the core system, one has a working prototype that is usable by some subset of people for some subset of the intended contexts.

We next intend to 'push out' the coverage of the system to more complex contexts: asynchronous working by different group members on the same design, distributed but asynchronous working (where designers email the evolving design or its changes back and forth) and finally synchronous, distributed activities. Again the gradualist approach applies: problems that a single user has will scale

226 M.B.Twidale, T.Rodden, I. Sommerville

Errors and omissions

Simple Test Cages

I More Complex I

- - - - - - - - ~_ - - -I

Advanced Cages

--------------------~

Increasing prototype sophistication More complex features and contexts More diverse users and scenarios

More subtle errors and omissions are smaller. There are unlikely to be detected until the less subtle ones have been spotted and rectified. Some simple test cases do not involve them and so cannot detect them.

Fig. 3. Illustration of the gradualist approach to developing and testing: using simpler cases first

out to multiple users at a single workstation, and the problems of the latter will scale out to cases of remote collaboration.

This approach is essentially complementary to the design work of for example Ishi and Arita (1991), Lu and Mantei (1991), Bly and Minneman (1990) and Tang (1991) who have studied users working on a shared drawing surface while in separate locations.

7 Results of the Studies

The system is now usable by individuals and small groups at a single workstation working on small scale projects. The users do not need to be computing experts. Fig. 4 illustrates a design by an undergraduate majoring in Religious Studies who used the DNP to plan an essay about Witch Doctors. The various results are reported in detail elsewhere (Twidale et al. 1993a and b). Of particular interest for this paper is the degree to which the interface needed to be modified in response to the observations of use and the emerging requirements of users. This is despite the fact that we consider ourselves as having some experience in interface design. We believe that predicting what users will find difficult is an inherent problem in the development of any system of CMC which has many potential uses and users (Grudin 1988). For such systems it is extremely difficult to predict what users will do and hence to design an interface to best support that activity. We have observed a variety of styles of use of the system. For example, some people make great use of colours, others hardly at all. This variety may be due to three reasons:

Developing a Tool to Support Collaborative Dialogues 227

I meaning and svmbOl1

I existential approach I I private and public I

I retrospective H not mantic I I concepts I I cognition and emotion I I mysticism II context threatening I

scepticism

I deception ~ tricksters I

Fig. 4. Preparation for an essay by a Religious Studies major. Seven colours were used.

1) Users have different styles of use. Both the process and product of designing vary greatly. This may be due to learning styles (Pask 1976), personal preference, academic background, etc.

2) We have observed the same user adopting different styles. This may be due to their working through an activity cycle such as brainstorm, rest, tidy up, reflect, brainstorm.

3) The nature of the task may determine the employment of a particular style.

Therefore the system and in particular its interface should be designed to accommodate a variety of styles of use.

228 M.B.Twidale, T.Rodden, I. Sommerville

7.1 The Effect of the Interface

A poor interface can have a substantial effect on performance, completely swamping the effect of other features of the system. This can make evaluation of a system particularly difficult (Twidale 1993). An improved system with far greater functionality may fail to display the expected gains in terms of say greater student learning due to a poorly designed interface which requires the student to spend more time on learning and concentrating on how to operate the system and consequently less time on the subject domain.

When the interface for a feature is improved one can observe a dramatic increase in its usage. We found such a case for the Textnote feature. This usage change can in turn lead to a change in requirements. So with Textnotes easier to manipulate, some users created many more of them and then wanted additional features to search and manipulate them.

7.2 The Co-evolution of System Functionality and User Requirements

It seems that users' requirements co-evolve with the developing system: as one feature is imprOVed it changes the nature of the activity being supported and this in turn leads to a new requirement. This is analogous to the way word-processing features have developed. Word processors were developed to support the writing process, but when a writer is provided with an easy to use word-processor, its effectiveness changes the nature of the writing process (Norman 1988). This in turn leads to new requirements for word processors which when implemented will in turn affect the writing process, and so on.

Therefore user requirements and interface issues of usability are not just difficult to predict, they are inherently unpredictable and so continuous testing is a necessary approach to developing a successful interface.

7.3 Coping with a Never-Ending List of Requirements

Partly as a result of this co-evolution and because of the ease of requirements capture by prototype criticism it seems that user requirements can become near infinite. Once you have a core system that is easy to use so that users can get started early on their designs they seem to articulate a never-ending stream of requests!

In particular it is important to discriminate between general and particular suggestions. Perhaps surprisingly, we believe it is more productive to heed the particular, especially at the early stages of design. A general suggestion has the characteristic form of "Wouldn't it be good if .... " whereas a specific suggestion is more like "What I want to do now is somehow to .... ". The specific suggestion is couched in terms of the user's particular activity and needs to be translated into some feature that would allow the user to achieve that desired action (and to a form that is useful to others). General suggestions are particularly common from people observing a demonstration of the system but not using it to solve a problem of their own. A particularly recurrent suggestion is "Wouldn't it be good if you could

Developing a Tool to Support Collaborative Dialogues 229

do rough free-hand sketches?" Now this is a perfectly sensible suggestion and it is clear that this would be a valuable additional feature. However none of the volunteers using the system to date requested it, although they did ask for a host of other particular features.

It is our contention that this difference implies that it is better to improve the system's existing functionality of text plus simple annotations (so that users do not abandon using the tool for handling problems that are amenable to that approach) before extending the coverage to those problem types which require quite new functions such as sketching. Therefore for example the DNP would not be of much use to a mechanical engineer doing design because sketching is a crucial part of her design activity. However it is of use to many other kinds of user (computer scientists, artslhumanities undergraduates, project planners etc) who may occasionally employ sketches but for whom this is not a core activity. The reason it does appear to be useful to such people is that we have considered and implemented many of their specific suggestions that are generalisable and so useful to many kinds of user, but are incremental improvements of existing functionality rather than being completely new features. For example the framing option described in the system overview that allows you to group entities, move them and collapse them down to form a new subdesign (thUS providing an abstraction mechanism) arose from a particular suggestion from a user who had a cluttered design after a bout of additional brainstorming.

It should be noted that implementing the particular suggestions is not a trivial panacea. As mentioned there are far too many of them. Therefore implementation needs to be prioritised according to several (potentially conflicting) criteria. These include the degree to which the suggestion might be generally useful (and the designer will need to consider how to make it so), its importance in impeding or accelerating a user's progress, its ease of implementation and the effect it may have on overall system performance, reliability and ease of use.

Although it might sound like mere common sense to focus on what people complain about before adding in more features that they might want, there is a strong pressure to concentrate on the latter. Systems are more frequently compared by functionality (the number and variety of features, flexibility, generality etc) both in the commercial and the academic worlds. In the commercial world, after consideration of price, purchasers of systems are inclined to opt for the system with the most functionality and pay less attention to usability. This is understandable as 'usability' is hard to assess; functionality can be measured by looking at the list on the back of the box in many cases. Likewise in the academic world, a paper about a CMC that has more features, has a brand new feature, is more generic etc. is likely to appear more impressive than a paper that describes a 'mere' reimplementation, but this time attempting to make the system usable outside the lab it was developed in. The latter may be dismissed as development rather than 'proper' research.

7.4 Collaborative Dialogues Using DNP

A study of the educational potential of DNP was undertaken (Twidale et al. 1993b) using four groups of four computing undergraduate volunteers. These groups chose to use DNP to help them work on a piece of coursework which required

230 M.B.Twidale, T.Rodden, I. Sommerville

them to work collaboratively on the specification of a deliberately open-ended and ambiguous remit. The intention of the coursework was for them to experience some of the creative aspects and problems of real world design. Of particular interest here is how groups of four users managed to successfully use a single workstation. The screen and the evolving design on it became the focus of interest and discussion. Although only one person could use the keyboard at a time and frequently one person used it for most of a session, all group members contributed to the problem and ideas once articulated on the screen seemed to become common property. That is they appeared not to be identified with the originator because all members participated in moving, shaping, colouring, positing or even renaming it by suggestions or consent.

Users' comments on their experience of using the tool often referred to its superiority over the confrontational nature of sitting round a table facing each other and writing on a piece of paper. When sitting 'shoulder to shoulder' round a screen, it was easy for any group member to take a turn in the conversation by leaning over and pointing at the screen. Even users who were sitting almost behind others would interrupt by leaning over and pointing. Because elements were easy to revise (rather than having to completely redraw them, as on paper), comments could be positive about how to change what existed. This differs from the pen and paper situation, which led to more negative comments in order to justify the effort of completely rewriting a design, or an attitude of deciding to put up with the current version rather than bother to change it. The ease of entry meant that the users claimed that they produced more and better ideas than working face to face.

8 Implications for CMC

8.1 Implications of the Results

Although the DNP provides a new way of working, users were able to very rapidly learn to use it and organise their work to make productive use of it. This bodes well for novel ways of cooperative design and learning. However the very success of the shared screen in becoming a focus for cooperative dialogue and activity poses a challenge. For remote cooperation, it will be far harder to achieve this sense of a shared focus. Much research remains to be done on this.

The importance of tum taking by leaning and pointing at the shared screen implies that attempts to support synchronous distributed cooperation need to pay particular attention to the turn taking mechanisms employed. The lean and point method conveys at least two meanings: 'I want to talk now' and 'I want to talk about THAT'. Note that in some ways this turn-taking is more fair than in face­to-face discussion round a table where it involves split-second timing at the moment a speaker finishes. This can be particularly difficult for participants not using their native language.

Users also commented on the ease and speed of entering features and the lack of commitment with comments like 'Just type it in, we can always change it later'. If the system feels sluggish to the user it completely changes the nature of the interaction, particularly during periods of rapid revision activity. Users become

Developing a Tool to Support Collaborative Dialogues 231

aware of the system rather than their problem and are inclined to revert to using pen and paper. This effect will also need to be considered by developers of synchronous remote systems where computational delays are likely. Note that this reveals a potential danger of prototyping: a cut-down prototype may have an acceptably rapid response that the full version fails to achieve and so the final version may be less acceptable than the prototype.

The way that ease of revision enabled ideas to mutate and all participants to be involved in the activity meant that ideas were not usually associated with individuals but with the group as a whole. This has many advantages for criticising and further adapting the idea as well as supporting general group dynamics. There is a danger that the named contributions to conferences lack this advantage. Conference organisers may wish to consider how to support a feeling of common ownership of an evolving idea within existing conferencing contexts. Whereas DNP only shows the final emergent design but not the process that led to it (unless partial designs are saved under different names), the text-conference is itself the change record, but lacks a summary of the final consensus (unless someone chooses to write this up and contribute it). It is likely that both the change record and a representation of the most recent idea structure would be useful.

8.2 Potential of the DNP in Computer Mediated Communication

The DNP offers users a means of rapidly articulating semi-formed ideas. It was intended to support the early stages of design and so was developed to allow software engineers to express vague creative initial ideas about a piece of software. Early designs are necessarily ambiguous and imprecise, but they are gradually refined to take on the structure and formalisms necessary to implement them in software. The same process can be used for many other creative activities including the exploration of ideas preparatory to writing about them.

Unlike pure text systems such as existing computer conferencing, the two dimensionality of the representations allows ideas to be expressed more succinctly. Links between words or their spatial connection can express subtle connections between them that have not yet been fully articulated (or even consciously considered yet). For example, two entities moved close to each other may mean; 'These are, or should be, somehow related, but I'm not sure how yet'. Later the connection can be considered and made more precise using labels on links, colour, shape, renaming the entities, creating intermediate entities, sUbdesigns etc.

The way that some conferencing systems use conferences and items to support different threads of conversation and the advantage this offers over a simple email distribution list is analogous to this, but in DNP the concept is taken further to within a single discussion. Indeed with the use of subdesigns, the system has features of a three dimensional textual representation, and a newly installed cross­referencing option (which creates a virtual link between entities in different designs) enables the construction of hypertext. Our results already show how this representation of ideas can be a powerful support for collaborative dialogues between groups of up to four people sitting round one screen. The screen becomes a focus of discussion and mutual misunderstandings can be detected and resolved by referring to and modifying the design. Yoder et al. (1989) have noted how a

232 M.B.Twidale, T.Rodden, I. Sommerville

shared-database hypennedia system with an effective interface provides a powerful foundation for collaboration.

We next need to investigate the features that will need to be added to the system so that it can be effective for collaboration at a distance. A start has been made on this, using a distributed version of DNP which runs on two workstations. Using our gradualist approach, we placed the workstations next to each other and had the two volunteers sitting side by side, each using a separate machine. The idea was that this meant they had perfect audio and video bandwidth (they could look and talk to each other nonnally) and if necessary lean over and point to the other's screen. In fact. after deciding on how to proceed, they worked in silence. They read and attached Textnotes to an existing design structure they had collaboratively developed earlier. They worked by reading the other's attached Textnotes and writing and attaching responses on new Textnotes. Using a CMC viewpoint one could regard them as using the DNP as a computer conferencing system.

We next wish to investigate the use of DNP in asynchronous distributed collaboration. The designs can be converted into a textual fonn and sent as conventional email to be reconstituted at a remote machine. However it is likely that additional channels of communication would be needed to explain and debate a user's design ·or contribution. We can choose a simple case first of using the textnote facility to enable discussion of changes made and the meanings of designs, in a similar manner to that just described.

The next stage would be to accompany the DNP with a conventional textual computer conference. It could be eventually integrated into the conferencing tool so providing a means of communicating ideas in addition to conventional text. This would support asynchronous distributed collaboration. Once the interface issues for this scenario had been resolved to a satisfactory level, we could consider the synchronous distributed approach. This is likely to involve the integration of a shared design with video and audio links (Derycke & Vieville 1993).

8.2 Potential of the Methodology

The employment of the DNP for remote synchronous CMC is a long term activity. However the results of our current research can offer a more immediate contribution to the design of CMC systems. Firstly the tool can be used as a planning tool for designing CMC courses. Course design is a creative design activity involving the management of multiple constraints just like software design. A colleague in the Music Department at Lancaster is currently evaluating the use of DNP in designing hypertext modules for teaching elements of the undergraduate Music curriculum.

Also, our experiences lead to a set of recommendations for developers of CMC technology and course developers. The chief recommendation is to pay particular attention to the interface; systems developers should allocate a substantial proportion of resources to the evolutionary development and testing of the interface. The system developed by Alexander et al. (1993) is a good example of a careful approach, focussing on the development of appropriate interface features. Testing should begin at an early stage and involve regular and in-depth analysis of failures: features that users find hard to use. These tests need not be large scale; our experience is that it only requires a study of a few users to reveal glaring

Developing a Tool to Support Collaborative Dialogues 233

problems, but they should be frequent as the design evolves and interface errors are corrected. In order to begin testing as soon as possible, a minimal system should fIrst be developed. Although this will naturally lack many features intended for the fInal system, it may still reveal signifIcant problems which will be much easier to correct at such an early stage. (Some usage problems will be due to the absence of the full system functionality and so must be discounted, but our experience is that other, unexpected issues are revealed at a fortuitously early stage). The tests should be authentic; as close as possible in nature and style to the expected usage, even if of more limited functionality. This is important to reveal hidden assumptions in the design. An example of a suitable test for a distance education context would be to observe a volunteer attempting to dial in via a modem from their own home or office rather than in the ideal conditions of a development laboratory .

We have claimed that interface effects can swamp more subtle ones such as features of the system that can support collaborative dialogues. Therefore researchers wishing to investigate such features need to consider the interface with care lest any results they obtain are due more to an unexpected interface feature than to something more profound about the nature or problems of working with CMC. This swamping effect has already been observed in the domain of interactive learning environments (Twidale 1991) where a poorly worded prompt produced a whole new set of student errors that it would be easy to interpret as a substantial and widespread learner misconception about the subject domain.

Those introducing a CMC course who have the opportunity to choose the kind of conferencing software used should consider interface issues when making the software selection. It may be that a system with an easy to use interface will lead to better learning than one that has more sophisticated features but is less easy to learn and use. Whichever system is chosen, we would advocate increased proportion of resources to teaching users how to use the system. Ideas such as Carroll's Minimal Manual (Carroll 1990) would seem to be particularly relevant. Naturally if students do not learn how to use a CMC system effectively they are not going to learn the subject of the course effectively. We do not wish to repeat the early days of commercial word-processors where many such machines were abandoned in the corners of offices because secretaries received inadequate training in their use so that their manifest advantages over typewriters were not realised.

9 Conclusion

When developing and testing the DNP we are always aware that our system is in direct competition with pieces of paper (or maybe a whiteboard). The early stages of design are normally done on paper and designers are experienced at using paper. Our system is in competition with pen and paper in the same way that early commercial word processors were in competition with typewriters. If at any stage our volunteers (and ultimately users of the completed system) find the DNP awkward to use so that its longer-term benefits (of ease of revision etc.) are outweighed by short-term costs of clumsiness of use, they will abandon it and revert to using pen and paper. We have observed this in studies. For development purposes this is no problem: we just examine the situation when they gave up and try to improve the interface of the features they were then trying to use and match

234 M.B.Twidale, T.Rodden, I. Sommerville

to their requirements. Outside user testing it is a problem. It does however impose a useful discipline on development in that we continually have to consider whether additional features might lead the user to give up and to go back to using paper or whether the feature postpones that moment.

In many current circumstances CMC systems do not have such direct competition. Users have to use the single system provided for the course. It may be that this lack of competition is why to date the evolution of interfaces for such systems has been slow. Instead, resources are devoted to adding features and enabling connection to many networks. These are both laudable aims, but they should not be pursued to the exclusion of interface features.

We can hope that the interest in interfaces will change in a manner analogous to that for Personal Computers. Users of DOS-based systems had to put up with the interface provided in the absence of anything better, because many businesses required the purchase and use of PCs rather than say a Macintosh with its superior interface. With the introduction of Windows, the take-up has been extremely rapid and considerable debate now occurs on the development of better interfaces to applications running under windows.

Acknowledgements

Michael Twidale is a Science and Engineering Research Council Junior Research Fellow. The development of the Designers' Notepad was funded by the Joint Council Initiative in Cognitive Science and Human Computer Interaction.

References

Alexander, G., Lefrere, P., Matheson, S. (1993) Towards collaborative learning at a distance. In: F. Verdejo and S.A. Cerri (Eds.), Collaborative dialogue technologies in distance learning (This volume)

Bly, S. A., & Minneman, S. L. (1990) Commune: A shared drawing surface. In: Proceedings of the Conference on Office Information Systems, (pp. 184-192). Boston

Carroll, J. M. (1990) The Nurnberg Funnel. Cambridge, MA: MIT Press Derycke, A. C., & Vieville, C. (1993). Real-time multimedia conferencing system and

collaborative learning. In: F. Verdejo and S.A. Cerri (Eds.) Collaborative dialogue technologies in distance learning (This volume)

Ehn, P. (1989) Work-oriented design of computer artifacts. Hillsdale, NJ: Lawrence Erlbaum Associates

Grudin, J. (1988) Why CSCW applications fail: problems in the design and evaluation of organizational interfaces. In: Proceedings of the Conference on Computer Supported Cooperative Work (CSCW '88), Portland, Oregon,1988,85-93: ACM Press

Ishii, H., & Arita, K. (1991) ClearFace: Translucent multiuser interfce for TeamWorkstation (Research report, NTT Human Interface Laboratories)

Lu, I., & Mantei, M. (1991) Idea management in a shared drawing tool. In R. M. Bannon L. Schmidt K. (Ed.), ECSCW'91, Amsterdam: Kluwer

McGraw, K. L. & Harbison-Briggs, K. (1989) Knowledge Acquisition: Principles and Guidelines. Prentice-Hall, New Jersey

Developing a Tool to Support Collaborative Dialogues 235

Norman, D. A. (1988).The Psychology of Everyday Things. Basic Books Pask, G. (1976) Styles and strategies of learning. British Journal of Educational

Psychology, 46, 128-148 Shipman, F. M., & Marshall, C. C. (1993). Formality considered harmful:

experiences, emerging themes and directions. Proceedings, InterCHI '93. Sommerville, I. (1992) Software Engineering. Fourth Edition. Addison Wesley,

Wokingham Sommerville, I., Rodden, T., Sawyer, P., Bentley, R., & Twidale, M. B. (1993)

Integrating ethnography into the requirements engineering process. In 1st International Conference on Requirements Engineering, San Diego: IEEE Press

Tang, J. C. (1991) Findings from observational studies of collaborative work. International Journal of Man Machine Studies, 34(2), 143-160

Twidale, M. B, (1991) Student activity in an Intelligent Learning Environment. Intelligent Tutoring Media, 2(3/4), 113-127

Twidale, M. B. (1993) Redressing the balance: the advantages of informal evaluation techniques for Intelligent Learning Environments. Journal of Artificial Intelligence and Education

Twidale, M. B., Rodden, T., & Sommerville, I. (1993a) The Designers' Notepad: Supporting and understanding cooperative design. In C. Simone (Ed.), ECSCW93, Milan

Twidale, M. B., Rodden, T., & Sommerville, I. (1993b) Investigating the use of a computational tool to support the refinement of ideas. Submitted to Computers and Education

Yoder, E., Akscyn, R., & McCracken, D. (1989) Collaboration in KMS, A Shared Hypermedia System. In Proceedings of ACM CHI'89 Conference on Human Factors in Computing Systems, 37-42.

19 Real-Time Multimedia Conferencing System and Collaborative Learning

A.C. Derycke and C.Vieville

University of Lille I, TRIGONE and LIFL Laboratories F-59655 Villeneuve d'Ascq, France Fax: +33 20436967 Email: derycke@trigone.citilille.fr

Abstract. Our previous experiment with the use of Local Area Networks to favour collaborative learning has led us to design and implement an open architecture to support intentional collaborative learning activities such as multi-party simulation and games for cooperative problem solving. This effort is now included in a DELTA European project called CO-LEARN which aims to support various collaborative situations and to give a right place to all the educational agents: tutors, teachers, colleagues at work and others learners. The fields of open and distance learning will be used as an experiment to try remote access through ISDN.

Keywords. Groupware, shared space, WYSIWIS, Audio Conferencing, Multiparty simulation, Teamwork, Conversation space, distance Learning.

1 Introduction

We have developed collaborative learning strategies using learning technologies and communication networks over a long period of time. In the framework of continuing education of adults we try to apply to the education process a general philosophy which can be considered as the "Cooperative Autonomy" concept : new learning technologies are used to open the learning and to favour self-paced learning strategies and the development of autonomous skills. The individual workstations are connected to a network so as to improve cooperation between learners and tutors. By cooperation or collaborative learning, we mean not only sharing, or exchanging of knowledge between the learners themselves but also with the various educational agents : teachers, tutors, experts, managers ...

This concept was first applied in the classroom with the advent of Local Area Networks (LAN).

From 80 to 85, we have been designing a global solution called the Nanoreseau which linked several learners workstations to a file and printer server by a low cost LAN (Derycke, Loosfelt and Comilliet 1986), (Derycke et al. 1988). This system offered some primitive tools which allowed some

Real-Time Multimedia Conferencing System 237

exchanges of information between the learner's workstation and we have applied pedagogical scenarios which mixed autonomous work and cooperative activities.

Our approach has been initiated by an ethical motivation : it happens that we are often involved in the retraining of adults with a low level of education and often with social problems (unemployment) and we use education for the socialization of the learners. On top of that, we know that collaborative learning has also benefited from the pedagogical or cognitive viewpoints, because it has been demonstrated that knowledge acquisition is also a social construction (Mugny 1989, Vygotsky 1962).

Moreover, it is recognized that social relationships, which can be developed through collaborative learning, are also a required cognitive capacity in human organizations, especially in the workplace (Bjorn-Anderson and Ginnerup 1991).

So we have claimed on some occasions that teamwork can improve the learning process and that it is also important to learn to work within a team. Of course, we are trying to transfer our general philosophy of learning to the distance learning case.

First, we have used asynchronous Computer Mediated Communication. We have experimented with a Computer Conferencing System adapted for the French Videotex Network, Teletel, in various collaborative learning activities such as problem solving or case studies. This led us to draw several issues in the design of a new system in order to support a wide range of collaborative learning activities and communication models (synchronous and asynchronous, text or multimedia oriented) (Derycke 1992).

This brought us to set up a research and development program for the design and experimentation of a Computer Supported Collaborative Learning Environment. Of course we borrowed many ideas from experiences in Computer Supported Cooperative Work (CSCW) which are essentially oriented toward Office Automation and the professional workplace.

Several actions have taken place in parallel, such as the design of a real time multimedia conferencing system devoted to educational activities (Derycke, Vieville and Vilers 1990) and some experiments with the "collectivization" of traditional computer software such as Hypercard used in an educational process (Clement, Vieville and Vilers 1992).

Since the beginning of 1992, our research and development work is, partly, integrated into a more ambitious project CO-LEARN which is funded by the DELTA CEC program and involves several partners in various European countries (Delta 1992).

238 A.C. Derycke, C.Vieville

2 The CO-LEARN Project: Architecture and Tools

2.1 Objectives

The project is concerned with a wide range of flexible and distance learning and training environments and focuses particularly on the access to human resources : tutors, experts, working colleagues, and peers.

Whether as it is usual in an active pedagogy based on the learner's participation and autonomy or in a more traditional one such as Teleteaching, the objective is to create the necessary conditions for a high quality communication at a distance using multimedia channels or message exchanges.

2.2 The Main Technological Choices

There will be a strong emphasis on certain elements or technological concepts in the project:

- the audio channel; numerous experiences have shown the great importance of sound quality in interpersonal communication: this quality must be excellent, which implies the use of 7 kHz sound, the development of audio-bridge applications, and the acoustic quality of the rooms and equipment;

- the multiwindowing WYSIWIS concept, because it makes the transition from the private domain to the public one ; each user will be able to work at the same time in a private space defined by one or several windows, and in a public space defined by other windows; the networks are used as "pipes", transferring the precise content of a local window to that a remote user;

- the ISDN network; it is well suited to support multimedia exchanges between users and integrated communications through the transfer of multimedia objects between applications. On top of that, it is international (we will put emphasis on the handling of problems related to the interconnection of ISDN networks) ; it is also a low cost alternative or complement to satellite systems.

The systems to be developed will provide the means for communication based on high quality sound, photographic and graphic images and text (or text images). These multimedia objects will be digitized on standalone workstations.

Exchanges will take place in real time or in batch mode ; they will be enhanced by access to a multimedia document base (instructors may wish to store their slides in such abase), which will also store complex objects including multimedia messages exchanged during conferences, or supporting material, teachware or complete synchronised lectures.

2.3 The Four Modes of Communication

In the CO-LEARN project there are four modes of communication which have been identified depending on the educational activities (teaching, learning, tutoring) and on the location and time of interaction (synchronous vs. asynchronous).

Real-Time Multimedia Conferencing System 239

( 1) Real-time tele-teaching This is based on the Tete-Amphi prototype (produced by Telesysteme), which supports remote lecturing from one to several sites, up to four, in conjunction with courseware (slides, diagrams, etc.) that has been downloaded to, or installed on , the learner workstations before the lecture, thus providing a remote "electronic blackboard" facility, via local video-projectors. The trainer can control the presentation on the remote blackboard, and point at different parts of the screen. After the formal lecture phase, users will be able to ask questions, via an audio channel, turn taking being controlled by the host computer.

(2) Real-time tele-assistance This permits direct connections between a central workstation and up to four "slave" workstations. The tutor's central screen will be able to represent the state of each learner workstation, and the tutor can help one learner at a time, or monitor up to four learners simultaneously.

(3) Real-time multi-media conferencing This is designed for a small group of participants (maximum 6), using the metaphor of the virtual meeting room. Each participant has a private space for documents etc., and can move them into the public discussion space for synchronous voice discussion. One channel of the basic access ISDN is used for voice, the other for data. If an access to a broad-band network is available, then video-conferencing would also be possible. The prototype already exists on an Ethernet LAN (Derycke 1990). Participants need a high-resolution screen, a keyboard and mouse, headphones and microphone (and a camera and video monitor, or a video card, if video-conferencing is to be included).

(4) Asynchronous computer-mediated conferencing and electronic mail This is a text-based computer conferencing tool specially adapted to the type of exchanges that occur during educational situations. It supports particular roles and types of messages which are built into the design of the user interface. It would include voice annotation of messages, and attachments of graphics and other files to messages, if these features seem desirable. At the user interface level, there will be a linkage with the multi-media real-time conferencing system, to allow easy transfer from one mode to the other.

2.4 Integration and Tailorability of the CO-LEARN Environment

In most real educational situations there will be a combination of the various communication modes and tools which are used to achieve the pedagogical goals.

This leads us to design a CO-LEARN environment in which the various components are well integrated, and which gives the possibility of tailoring this integration, to cope with the users' requirements.

The integration is achieved by a clear and coherent approach which stands on a solid background. The most important characteristics of this approach are:

240 A.C. Derycke, C.Vi~ville

- a single latent metaphor for the design of the overall communication system : the tool perspective. In each communication mode there is a collection of tools (symbolised by an ICON) which is engaged in order to support the conversation;

- a powerful metaphor for the design of the user's interface (the "look and feel"). In order to be consistent with the windowing system which is already present on the workstation, the CO-LEARN environment is presented as a virtual learning centre following the ROOMS interface metaphor (Henderson and Card 1986).

There are rooms with apparatus to speak, to teach using a virtual overhead projector, a silent lecture room, a library ....

- the main underlying paradigm is the conversation paradigm with a clear separation of the elements of the conversation or discourse (verbal utterance in real time, message in deferred time) and of the objects of the conversation (the documents);

- in each situation (i.e. in different rooms) the user can have an access to a common information space which can contain documents about the topic which is being learned: multimedia documents, courseware, simulation tools ... These documents can be accessed alone for self-study or in a learning group for cooperative activities either in real time (intelligent white-board following WYSIWIS Concept) or in deferred time.

The tailorability of the system is achieved by the use of a design methodology which:

- first, allows the implementation of an open system i.e. extensible and reusable. This is due to an object oriented approach either in the analysis phase or in the programming phase;

- second, allows the possibility to change the various controls of the communication in accordance with the role of the different "locutors". This is done by a separation of the part of the code which controls the interaction (the"floor Algorithm") from the remainder of the code and by offering tools to edit and change the rules which govern this control.

3 A Real-Time Multimedia Conferencing System for Collaborative Learning

In the following we will focus only on one of the communication modes proposed by CO-LEARN: the real time multimedia conferencing.

Real-Time Multimedia Conferencing System 241

3.1 Collaborative Learning Activities Characteristics

We can analyse the characteristics of the most appropriate Collaborative Learning Activities (CLA's) to be supported by the real time multimedia conferencing mode. In open and distant learning this mode must be kept for CLAs which must work in a small team in real-time, i.e. CLAs which take place, in the traditional Universities, in the lecture or meeting rooms. Group problem solving and case-studies are the CLAs which imply discussions in a small group around a common goal or assignment. But a multiparty simulation or an educational game can be other appropriate CLAs for this kind of tools.

The main characteristics of these CLAs are:

- requirement to share the same information space: for example some documents which present the case-study objectives or documents produced by the members of the team;

- support of a conversation space. In real time the conversation is almost supported by the voice exchange (audioconferencing);

- a shared focus on some parts of a document or on some kind of intelligent white-board which can be seen by all the participants in real time (the WYSIWIS "What You See Is What I See" concept);

- a coordination process which is often relatively complex with various phases: from individual work done in parallel to group work on the same document or object. We must remember there are various strategies in the organization of collaborative learning (Johnson and Johnson 1991).

3.2 The CSCW Underlying Models

The choice of methodologies or models is not an easy task due to the nature of CSCW. Bannon and Schmidt have isolated the care issues for CSCW (Bannon and Schmidt 1991). They said that "CSCW needs to address the following specific requirements of cooperative work:

- articulating cooperati ve work; - sharing an information space; - adapting the technology to the organization and vice-versa".

The first two requirements are relevant to an information viewpoint of CSCW system. The third requirement is relevant to the organizational viewpoint of the system.

A detailed analysis of this model is beyond the scope of this paper (for more information see CoLearn 1992). We can just resume the main features of the models:

i) There is a separation between the objects of the discourse (the multimedia documents or the content of the intelligent white-board) and the elements of the discourse (the different utterances of the conversation).

242 A.C. Derycke, C.Vi~ville

ii) Articulating cooperative work, or learning implies coordination of tasks. This can be supported either by an Activity Oriented model/system or by a Conversation Oriented model/system:

- Activity oriented model/systems are designed to support the cooperative execution of a goal oriented task; i.e. a task where the structure of the communication process can be described in advance;

- Conversation oriented models/systems are designed to support less formal activities based on a conversation analysis and often derived from the Speech Act Theory (Winograd and Hores 1986).

In the context of distance education the Activity models correspond to the directive phases of the learning process, for example organising a round table to collect all the participant opinions, and the Conversation models corresponds to non-directive phases of the learning process (free discussion).

iii) The main component of these models are the roles, the message-objects, the functions, and the rules following, for instance, the AMIGO Activity Model (AAM) taxonomy (Pankoke-Babatz 1989). This model is used to specify: - the participants, their rights and their duties (the participating roles); - the objects which can be used to communicate (the message object); - the operations that are to be executed (the functions); - the regulation governing; which operation is executed on which object by who (the rules).

These underlying CSCW models are very important because they give a clear view of the requirements and they allow the design of a very flexible architecture. Because the roles represent the dynamic capacities of a participant in a conference, the system is tailorable and can support many kinds of cooperative activities.

3.3 The Network Architecture of the Real-Time Multimedia Conferencing System

The network architecture is constrained by the nature of the open and distance learning field of applications.

Effectively we are addressing an educational process which must link several learners and educational agents dispersed around various remote sites. This implies several consequences:

- potential users do not necessarily have their own private workstation with a large hard disk and private database which can store their own system environ­ment;

- each learner can access the system from different locations: from his enter­prise, from the local learning resource centre, and eventually from home;

- as opposed to the office automation case, the members of a group network are not members of the same company or institution. So they do not have the same social obligation and the organizational model can be complex (Derycke

Real-Time Multimedia Conferencing System 243

and Kaye 1993). This implies protection of the resources: privacy, control of the access rights ...

In its first release, our system will not address the design of a national or European global open university network. The management of multi­conferences with very large potential audiences is not our concern.

For the real-time multimedia conferencing system, our requirement is the support of less than 6 users for one real-time conference (a group) and no more than four conferences/group simultaneously. In fact, we will see that the limit will be imposed by the support of the voice channels.

Our fIrst prototyping efforts have been done on Unix Workstation Linked by an Ethernet LAN. We have taken advantage of the communication facilities offered by Unix (especially the "socket" mechanism) and of an integrated prototyping object oriented environment, Smalltalk-80 that we have extended to support remote access to a collection of objects and distribution of objects.

However, for distance learning this scheme could not be applied because:

- we must choose more useful workstations for the learners. In spite of its flaws we have chosen a PC/386 with Microsoft Windows 3.1 because it is relatively cheap and offers many tools to support others aspects of the learning process such as authoring systems, hypermedia, word-processing, electronic spreadsheet;

- we must use the available public network. In spite of the relatively slow development and of some interconnection problems in Europe, it seems that Integrated System Data Network (ISDN) will be potentially the best solution. Effectively, a real time multimedia collaborative learning process requires some interactions between the participants either by conversation using a voice channel or by a shared interactive information space (for example a multi-user drawing board).

The introduction of a temporal media, i.e., the voice, in the system, is one of major feature of CO-LEARN: in an asynchronous mode it is used for audio annotation in messages, in synchronous mode it is used to transport the teacher voice (teleteaching) or the group discussion (audioconferencing).

From the groupware point of view our system must support multiparty interactions on a relative peer to peer equal access (no master-slave relationship between the users even if one is a tutor, with social obligation, and the others are learners/customers). The central concept of this kind of collaborative learning is the group: a well identifIed collection of learners who are involved in the satisfaction of more or less common learning goals. We must remember that a group exists when its members perceive themselves as a "we" (Bannon and Schmidt 1991). From the organizational viewpoint the group is a social object manipulated as such.

For the system requirements, this implies that everyone can have a conversation not only with another participant but also with the group entity which looks like a virtual participant (i.e., use of broadcast facilities to transmit in real time not only the speaker's voice but also the action on the shared information space to all the group members).

244 A.C. Derycke, C.Vieville

All the aforementioned requirements and technical constraints lead us to adopt a relatively centralised architecture: a powerful Unix server hold the "kernel" of a "virtual learning resource centre". All the users, even the educational agents, can use the system from a remote location. In fact as we can see on the figure (1) the Conference server is a bridge between the users: it supports the audioconference via an audiobridge, a conference manager and a shared information space.

However we tried to relax some of the drawbacks of this centralized system, especially its relatively slowness for interactive actions, such as drawing using a direct manipulation interface (mouse and pointer).

DEC 5100

32 No + Nrs Acceleraf.or

Elhernet

Remote Worlr .... llon

( up to e simultaneously»

IS D N

(NUIoIERIS)

_.

Dete + Voice ( 7 KHz )

SO(2B+D)

) ....

SUN JPC

lo Campua F.O Nteworll

Fig. 1: Real time Multimedia Conferencing system based on narrow-band ISDN

In fact our architecture is hybrid (Lauwers, Lantz and Romanow 1990), most of the objects (program code) needed by the user application are present on the user's workstation. The conference server holds only some kernels (conference manager, and applications models) which are necessary to maintain the semantic consistency among the user actions and to solve the synchronisation problem at the beginning of a conference session (ability to accept latecomers ... ).

Real-Time Multimedia Conferencing System 245

3.4 Software Architecture of the Real-Time Multimedia Conferencing System

It is out of scope of this paper, which is oriented toward learning aspects, to give a detailed analysis of the software architecture. Some of the features, however, can be described in order to show the versatility and openness of our architecture.

- We use intensively the property of the object oriented approach for analysis, designing and programming. This allows flexibility of prototyping, reuse of components and openness because we offer several object classes to the designer of specific collaborative learning tools. Our software architecture can be considered a framework for extensibility by means of designing new groupware.

- The general aim of CO-LEARN is the integration of the different tools and communication modes described in the first part of this paper, in the same coherent environment. This means that the asynchronous and synchronous modes share some data about the information model (roles, messages, rules ... ). This is achieved by using a central object repository implemented on top of the GEMSTONE object oriented database.

- At the interface level the integration is achieved by the using the same general metaphor. We use the "Rooms" model: each group work takes place in virtual rooms which are placed where conversations occur (either synchronous by voice or asynchronous by multimedia messages) and which contain all the public tools and shared information engaged in the group activities.

- We don't postulate any assumption about the style of conversation (game of roles inside the group) which could be used. In the conference manager the group interaction is supported by a "floor algorithm". But in our case this "floor algorithm" depends on the dynamic role of the participants and on scripts which describe the rules. So the interaction scenario can be easily changed. Our approach is similar to that of T. Malone and his co-workers (Malone et al. 1987) or (HammaInen et al.1987).

In the first version of our prototype (Unix on Ethernet for classroom uses) the software has been produced by the Smalltalk 80 environment. In the CO­LEARN project, the work is done in a mix of C++ (for the implementation of various tools such as a shared telepointer) and of Smalltalk 80 either for the interface integration (the "Rooms" implementation) or for the centralised object repository. The interaction between the various tools is achieved by the use of the DDE exchange mechanism of Windows 3.1.

3.5 The User's Interface of the Real-Time Conferencing System

Figure 2 gives a view of the user's interface.

246 A.C. Derycke, C.Vieville

Fig. 2. A view of the user's interface

There are three kinds of windows:

- private windows which are only controlled by the user owner of the workstations. They represent tools used for private annotation, spreadsheet, word processor; - one public window which supports the WYSIWIS concept. In most cases it can be seen as an intelligent white-board. We use the word "intelligent" because the application (groupware) which underlies the white-board can react to user actions and perform powerful processing of the underlying objects and their graphical or textual presentation. This public window is very important for the "conversational space". Due to its uniqueness, it focuses all the participant's attention on some shared objects/documents which are the objects of the conversation. The elements of the conversation, which must be distinguished are exchanged by means of the voice channels and the audiobridge. However the WYSIWIS window also conveys some part of the conversation because some gestures of the locutor can be transmitted by the movement and appearance of the tele-pointer.

Of course there are some facilities to exchange between the private space and the public space. Initially a clipboard mechanism was present, but currently we use the DDE and OLE facilities of Window 3.1. For example we want to use an available drawing software such as Designer 3.1 and allow for the possibility of composing a picture which will be put in the public space and still allow it to be editable. - One conference manager window which gives feedback to the user of the coordination process rule by the floor algorithm (who has the control, how to get the control...) and of the activity of the others users: who is present, hislher name ... - Some dashboards which contain the various tools which are symbolized by the Icons (for example the Telepointer).

Real-Time Multimedia Conferencing System 247

4 An Example of Integrated Collaborative Learning Environment Using the Real-Time Multimedia Conference System: The Kanban Simulation

4.1 Objectives

With the availability of the first prototype system (Derycke et al. 1990) we decided to implement some dedicated collaborative learning applications, i.e. specializations of the "intelligent white-board". In order to verify the functionality of the conference manager, some very simple applications have been designed such as a simple drawing board or some educational games which involve a multiparty interaction.

However, we wanted also to check the flexibility of the architecture and to investigate the cooperation process in learning. So we have chosen to design relatively more complex applications on the "Unix + Ethernet" platform. Voice handling has not been a problem because we have decided to experiment with the system inside the classroom: the participants were co-located and able to speak altogether without any technical support.

The choice of the applications was constrained by several factors:

- the availability of students to experiment in a real situation and that of a domain-expert (a professor);

- the goal to implement and experiment in a non-trivial application: i.e. which offers a rich set of interaction patterns from stand alone individual works just coordinated by time constraint and some shared data, to highly collaboration awareness phases with the use of a unique shared tool;

- the necessity to imply several kind of shared documents in the common information space: from read-only text pages to interactive pictures (buttons embedded in the picture) and even a true multi-user electronic spreads.

4.2 The Kanban Simulation: A General Overview (Vilers 1992)

We have chosen the domain of production management and planning because of its importance and the difficulty to teach and learn, and also because it is intrinsically a cooperative process (between the different plants or the different workshops in order, for example to have a just-in-time, no stock, no delay production). There are several methodologies used in the planning of a manufacturing process. We have implemented two of them: the Manufacturing Resources Planning MRP, and the Kanban approach.

In the educational context, the learning of such a methodology is often achieved by a collective simulation game. Each student or team of students (sub­group) plays a role (he is in charge of a particular workshop, or of the supervision of the whole process).

In the following, we will focus only on one of the Kanban methodologies in a way sufficiently general to understand the general approach.

248 A.C. Derycke, C.Vieville

The Kanban learning environment is a collective game or a multiparty simulation illustrating a Japanese approach to the planning of manufacturing production. In the traditional classroom, without computer systems, it is run by the instructors during lecture periods. We have designed a computer assisted version using the real-time multimedia conferencing system of up to 5 or 6 workstations (students or subgroups). The purpose of the game is to study the decision making process in planning, and the implications and results for the manufacturing process (cartering, delay, stock levels, etc ... ). Participants play different roles (production manager, sales manager, workshops manager ... ).

The group activities in the game pass through several phases:

i) a common task phase to establish the planning of a next week's work for different parts of the plant (it is a mechanical manufacturing with production of gear, wheels ... ). This phase is based on documents containing relevant information such as orders of the previous week, etc... ; these documents are put in the shared information space. The task is highly collaborative, as all the participants work on the same documents, in their respective roles, and have to plan collectively;

ii) the simulation of the manufacturing process based on the collective decision taken earlier; here coordination levels decrease as each "player" is in charge of a different part of the plant. Nevertheless, these parts are related to each other because products are exchanged between different parts (coordination by data exchange), and the planning must be respected. The simulation also mimics unplanned events such as machine breakdowns and even strikes ;

iii) a de-briefing phase, with analysis of the results of the game; here the instructor's role is very important in drawing conclusions and pointing to the results of each team's decisions. All the information collected during the simulation is stored in a common database for review and analysis (the shared information space has also a role of a group-memory).

This collaborative learning activity is an interesting example because:

- it is complex enough to represent many games and simulations closed to those used in the learning of economics and management sciences ; - it needs a variety of tools to achieve rapid results (tools for group coordination, for decision-making, specialized tools such as spreadsheet, etc); - it introduces temporal constraints, because some of the information must be updated rapidly for all participants (i.e. in real time) and because time intervals (each hour of production) and deadlines have to be respected. This implies a need for tools for participants to ensure that their work is coordinated, and they are all in phase.

Real-Time Multimedia Conferencing System 249

4.3 The Kanban Simulation: Some User's Interface Aspects on a Unix Workstation

Figure 3 gives a copy of the screen of one instance of the presentation phase and planning of the work. The public window shows a schematic diagram of the articulation of the different workshops (for example montage is the assembly of all the pieces of a motor reductor, couronne is the machine which produces gear, ... ).

At this time the different participants discuss the organization and give each other the responsibility of a subpart: they put their name on the workshop (Adele will be in charge of the carter workshop and David has not yet chosen ... ).

We can also see the conference manager window which gives some information about the connection-deconnection process, the status of the control (Adele has currently the control symbolised by a pen or a chalk passing) and some menus to choose a style of interaction through the telepointer: just a cursor arrow, a cursor with permanent trace (a pen, a rectangular selection or for designing object in inverse video ... ).

Fig. 3. Screen copy of the presentation phase and planning of the work

250 A.C. Derycke, C.Vi6ville

Figure 4 shows another example of the user's interface during the planning phase: the view of the conference mediator. This is often the instructor's role but it· can be played by anyone of the participants. The learners have a slight different view: the content of the spreadsheet is of course the same but they have different menus to interact with the spreadsheet. This is a true multi-user system because each participant is able to submit concurrently a value for a precise cell: in our example the cell at the intersection of Mercredi (Wednesday, day of the planning) and of Reductor R4, the considered workshop. The mediator has the possibility during a discussion to select one of the propositions (in the example 100 - 200 - 300) or to apply sequentially all the propositions in order to show their impacts on the planning of the data sheet.

Remember that at each time, each user (student or team of students) can open one or more private windows to examine some common information stored in the shared information spaces. This implements some viewpoints feature : a user will see the marketing information, the other will see the information about failure of the last week of production. n must be noted that this first prototype implemented in Smalltalk 80 has an interface different from that of CO­LEARN.

iiii=lg o o 200 o 200 200 o 200 200

c ............ v ... 0 c.......-........ o c-. ....... T .... 200 ~_ 200

==~,.Ig 200

Fig. 4. The user's screen during the planning phase

Real-Time Multimedia Conferencing System 251

Figure 5 shows the simulation phase following the Kanban methodology: in this approach the flow of pieces and stock management are tracked by using some "token", Kanban in Japanese, attached to a batch of pieces. The Kanban are symbolised by different rectangular icons placed in the differents rows relative to each manufacturing machine.

At the left upper corner the users have informations about the progress of the simulation (day and hour of the simulation). They have also some views on the planning of activities and some information about the overall process. In the right upper corner the users have information about some events which are random such as a breakdown and a two hour stop which occurs occasionally. In this phase each user has a different view, the coordination is relatively low.

There is only a synchronisation by the exchange of mechanical pieces. However all the events are discussed by a verbal conversation between the participants and some decisions are commented.

r-, ...... .;0,,;;. ':2 ~t ..................... ....me ." ... ,. ~ ... O~ •• AmII

1~[RlllllIiI. I I: [RIlE I.

I~ t¥+M1."' ..... tiiIIiI .. & ........ c • .,..

I: .-TlI orrAllTll

1= IR1 Imm 1"1" I"" I"" lAP IAI 101 I'" I TO ''''I"" I~ 1R2 I~ Ig I~:~ I- I- I- I- I- I- I- I-

I~~ I! 1-

tii~

Ii! IrRlll I I- I- I- I- I- I- I- I- I- I-

• [RIlE 111111 Il~ I EtlJl £:] I

li~ I- I- I- I- 1- I: c III aoa

I 1110 I l(~~~ I-111+++ ~ 1-

;1:1:: I I~ I- - I- - -I~

§ I- !- I-I-

..;.

Fig. 5: The user's screen during the Kanban simulation

252 A.C. Derycke, C.Vi6viUe

4.4 A First Evaluation

We have conducted a first limited experiment with the simulation game with students who learn manufacturing automation and management at the graduate level. This is not a full evaluation of the collaborative process due to lack of time and machine availability so that the whole process has not been conducted. In a real situation the simulation which runs in the lecture room takes roughly 40 hours for the study of the MRP and Kanban methodologies. Therefore our evaluation is not complete from the pedagogical point of view because we have not been able to determine precisely the effectiveness of the computer assisted approach versus the traditional one.

However, students were very interested and it seems that the simulation is faster than the paper and pencil approach.

The debriefing phase is more intensive for the computer assisted cooperative simulation: in the traditional approach the de-briefing is just an instructor speaking from the notes he has collected during the simulation process.

However in the computer approach the possibility of using collectively some traces is very valuable. For this first experiment we have been more interested in the human factors of our design: usability of the first level of interface, articulation of the information space and of the conversation space.

Because we have no control over the voice turn-taking, everyone can speak when he wants, the evaluation process shows that the rigidity of our first "floor algorithm" implementation, combined with the Telepointer characteristics, entails some ergonomics problems. The gesture of a speaker is not always "seen" because the present speaker is not necessary the one who has the control of the public space and is the holder of the Telepointer (the "chalk"). The lack of concurrency in the use of the public window and of its underlying application software or model constrains the conversation which tends to be serialised such as the written text in this public space. This has consequences on the conversation which tends to be less natural: each user plans his contribution too much which is not always appropriate to the context of the current step.

Two solutions are simultaneously engaged in our current design:

- to give a better integration of voice control (sound) and user's action control. In Fig. 1 this is achieved by the integration of the sound channel in the user workstation with some local control (voice inhibition for example) and by the supervision of the audiobridge by the server which holds the kernel of the conference manager; - to offer a concurrent access to the public space. This could be achieved by a finer granularity of this space : the user will have the control of some objects (parts) of the public space and we are offering multi-telepointers which give feedback on the others users interactions.

The conclusion we have drawn is close to those of others working in CSCW applied to the office. At the more general level we have a research agenda which has some analogies with the work of Greenberg and al. (1992) or of Ensor and al. (1988).

Real-Time Multimedia Conferencing System 253

5 Issues for the Design of a Real· Time Multimedia CoHaborative Learning Environment

5.1 The Multimedia Issue

This issue is twofold. The first problem is with the integration of the voice channel or more

generally with the support of sound not only at the learner location but also at the server location (audio conferencing). It appears that low quality monophonic sound, such as that available on traditional telephone lines, is not enough for group education: . - in the teleteaching mode a medium for high quality voice channel is required to transport all the subtle tone and pitch, of the teacher voice and to afford a good sound reproduction at the remote classroom; - in the real-time multimedia conferencing mode high fidelity sound is necessary to ensure a good speaker recognition in a multiparty conversation.

Even if multimedia is gaining popularity in the personal computer domain with the advent of low cost digital audio boards and their integration in the operating system (MPC extension of Windows 3.1 for example), it is still a problem. There are two reasons : the first one is the incompatibility of this PC audio standard with those of international network standard (for example the CCnT G 722 standard for digital voice with a 7 KHz bandwidth which implies a very expensive audio board) ; the second one is the availability of very flexible and intelligent digital audiobridge allowing the speaker detection, control of sound level..., and controlled by a computer. We have identified this problem as one of the major problems and we have started a research program toward these issues (Derycke and Barme 1993).

The second problem is the handling, for example in the public space, of multimedia documents. In several learning situations it will be interesting to work collectively around a live video sequence with the possibility to superimpose some pointers (overlays of icons, texts ... ). With the progress of the hardware integration of multimedia it is now possible to have live video in one of the windows of the workstation. However the problem is the sharing of the same video access, its distribution and its control. At the campus level, because it is possible to use available technologies, such as analogous video transport network, we have started an implementation of a cooperative video server. The major question is then: what kind of tools should we offer to support collaborative learning organised around a video document and what articulation should be provided between the sound of the video and the sound of the group conversation ?

254 A.c. Derycke, C.Vieville

5.2 The "Specialization" Versus "Collectivization" Issue

In the previous example, the Kanban simulation, the groupware which supports the public window, is especially designed for a well identified educational goal. It takes into account the true multi-user constraints of the collaborative process.

This is a powerful way but it requires : - a knowledge of how to design groupware. Fortunately the use of an open architecture and of an object oriented environment eases this laborious process; - the choice of a pertinent field of application due to the high cost of this specialised development.

At the opposite end there are many tools (commercial applicative or educational software) which are available for the learning process. One idea is the use of such a software in collaborative learning activities. Unfortunately this "collectivization" is not so easy, even with the use of network software such as TIMBUKTU (Clement, Vieville and Vilers 1992) because the software has not been designed for multi-user interaction: it exhibits no control access schema and no multi-user feedback.

In the CO-LEARN project, we try to balance the following opposing ideas: we are investigating some generic collaborative learning tools such as multi­users drawing board, or text editors, brain-storming tools or some kind of semantic networks for group users. This follows the approach proposed in the COLAB project (Stefik and al. 1987). This generic software for collaborative learning will be tailored especially for this purpose.

But we also offer some provision to allow for a collaboration between a commercial package, if it is an OLE server (the mechanism of Windows 3.1), and the generic white-board. The exchange is quite similar to those using a clip­board but the update are automatic and it is possible to modify the content of the white-board using an editor on one of the private spaces of each user.

5.3 The "Conversationalist" Issue

We have observed that the voice channel is the most important communication link in the real-time mode. The distributed computer screens are just secondary communication links which represent the objects of the conversation. Forgetting this fundamental aspect can lead to failure in the design of usable CSCW system (Tatar, Foster and Bobrow 1991). But "conversationalists" say that conversation between humans cannot be reduced to the exchange of utterances but gesture, eye movements and gaze are also conversation's components. Even if the use of a video conferencing channel could improve this factor, it is not very usable on the collaborative learning framework because the specialised hardware (codec and camera) is expensive or requires a higher ISDN channel rate.

We think that some indication of the user's position and reaction can be provided by a set of dynamic icons, for example an icon for the representation of the current speaker, either a digital photo or a symbol, and by a richer semantics of the telepointer, for instance by changing its appearance according to the use made by the owner.

Real-Time Multimedia Conferencing System 255

It is important to point out the importance of the conversation and underlying voice channels and audio-conferencing system, because it allows some potential for a double level language: a level of coordination, rather technical and constrained, and a more cultural level (Robinson 1991).

6 Conclusion

We are involved in the CO-LEARN project which intends to offer a rich environment for collaborative learning activities for the third generation of open and distance learning institutions. It means a shift from the classical "media-centered" paradigm to a conversation or dialogue paradigm. This shift allows the re-introduction of the social resources: educational agents and other learners. The CO-LEARN system which is under development could be seen as a social resource server, a kind of virtual learning resource centre with opportunities to meet some-one. Perhaps in the near future, in respect to the ROOMS interface metaphor, we will offer some halls or galleries to favour a non-expected encounter such as in the CRUISER system (Root 1988).

Moreover, we expect that Computer Supported Collaborative Learning will have a positive effect on the abilities for the learners to work together. This reflexive characteristic of some processes is clear in the Kanban methodology for the manufacturing process which can be seen in itself as a cooperative process.

Acknowledgements

The authors want to thank all the members of the "Nouveaux Outils pour la Communication Educative" team (L. Barme, H. Cabre, D. Clement, P. Croisy, J.N. Gers, F. Hoogstoel, P. Vilers) and the other partners of the DELTA CO­LEARN project, especially A.R. Kaye (Open University, UK.) for numerous discussions.

This work is partially funded by a EC DELTA project, and by "Ganymede" which is the regional inter-laboratories research program on Advanced Communication funded by the Nord-Pas-de-Calais Regional Council and the French State.

References

Bannon, 1.1., Schmidt, K. (1991) CSCW : four characters in search of a context. Bowers, J.M. and Benford, S.D. (eds.) Studies in Computer Supported Cooperative Work, 3-16, Amsterdam, Elsevier Science Publishers.

Bjorn-Andersen, N. and Ginnerup, L. (1991) The support of cognitive capacity in future organisations : towards enhanced communicative competence and cooperative problem-structuring capability. In: Rasmussen, Brehmer, Leplat (eds), London, John Wiley and Sons.

256 A.C. Derycke, C.Vieville

Clement, D., Vieville, C., Vilers, P. (1992) An experiment of cooperative learning with Hypercard. ICCAL'92, Wolfville (Canada), Tomek, I. (ed.), 150-160, Lecture Note in Computer Sciences, Berlin, Springer.

CO-LEARN(June 1992) Representation models for collaborative educational situations and collaborative learning activities, public deliverable n° 4, Bruxelles, DELTA head Unit.

DELTA (March 1992) Research and technology development of telematic systems for flexible and distance learning. Delta Central Offices, Bruxelles.

Derycke, A.C., Loosfelt, P., Cornilliet, G. (1986) Cheap Local computer network: an ultimate solution for the classroom? Proceeding of Eurit'86 conference, Moonen, J. and Ploomp, T. (eds.), London, Pergamon Press, 207-210.

Derycke, A.C., Vieville, C., Poisson, D., Stach, c., N'Guyen, H. (1988) Le Nanoreseau: utilisations pedagogiques d'un reseau local, 7-20, Paris, Techniques et Sciences Informatiques.

Derycke, A.C., Vieville, C., Vilers, P. (1990) Cooperation and Communication in Open Learning: the CoCoNUT project. Proceeding of the fifth World Conference on Computer and Education lAP, Sidney, Australia, 957-965, Amsterdam, North­Holland.

Derycke, A.C. (1992) Toward a hypermedium for collaborative learning? In: Kaye, A.R. (ed.) Collaborative learning through computer conferencing. The Najaden papers. NATO ASI Series F, Vol. 90, 211-224, Berlin: Springer.

Derycke, A.C., Kaye, A.R. (1993) Participative modelling and design of collaborative distance learning tools in the CO-LEARN project. Proceeding of Teleteaching'93, IFIP, Trondhein, Norway, 20-25 August, G. Davies and B. Samways (eds.), Amsterdam, North-Holland, 191-200.

Derycke, A.C., Barme, L.B. (1993) Some issues on sound in CSCW: an often neglected factor. Submitted to third European Conference on CSCW, Milan, Mimeo.

Ensor, J.R., Ahyo, S.R., Hom, D.N., Lucco, S.E. (March 1988) The rapport multimedia conferencing system: a software overview. Proceeding of the IEEE Conference on Computer Workstation, Santa Clara (USA), 52-58.

Greenberg, S., Roseman, M., Webster, D., Bonnet, R. (1992) Human and technical factors of distributed group drawing tools. Interacting with Computer, Butterworth.

Hai"mmai"nen, H., Sulonen, R., Berard, C. (1987) PAGES: Intelligent mail and distribution. Bracchi, G., and Tsichristzis, D. (eds). Office Systems: methods and tools, 45-57, Amsterdam, Elsevier Sciences Publishers.

Henderson, D.A., Card, S.K. (1986) Rooms: the use of multiple virtual workspaces to reduce space contention in a window-based graphical user interface. ACM transactions of Graphics, Vol. 5, n° 3, 211-243.

Johnson, D.W., Johnson, R.T. (1991) Learning together and alone: cooperative, competitive and individualistic learning, Boston, Allyn and Bacon.

Lauwers, J.C., Joseph, T.A., Lantz, K.A., Romanow, A.L. (April 1990) Replicated architectures for shared window systems : a critique. Proceeding of the ACM Conference on Office Information System, 249-260, Cambridge, MA.

Malone, T.W., Grant, K.R., Lai, K.Y., Rao, R., Rosenblitt, D.A. (1987) Semi-structured messages are surprisingly useful for Computer-Supported Coordination. ACM transactions on Office Information System, Vol. 5, 115-131.

Mugny, W. (1989) Psychologie sociale du developpement cognitif, Berne, Peter Lang.

Real-Time Multimedia Conferencing System 257

Pankoke-Babatz, U. (1989) Computer-based Group Communication, the AMIGO Activity Model, Chichester, Ellis Horwood.

Robinson, M. (1991) Computer Supported Co-operative work: Cases and concepts. Proceeding of Groupware 1991, The potential of team and organisational Computing, 59-75, Hendrik (eds.), Utrecht, SERC.

Root, R.W. (1988) Design of a multi-media vehicle for social browsing. In Proceeding of the CSCW'88 ACM Conference, Portland, OR, 25-38.

Stefik, M., Foster, G., Bobrow, D., Kahn, K., Lanning, S., Suchman, L. (January 1987) Beyond the chalkboard : Computer support for collaboration and problem solving in Meeting. Communication of ACM, Vol. 30, n° 1,32-47.

Tatar, D., Foster, G., Bobrow, D.G. (1991) Design for conversation: lessons from cognoter. Greenberg, S. (eds.), Computer Supported Cooperative Work and Groupware, 55-79, Academic Press.

Vilers, P. (February 1992) Etude d'un systeme de Teleconference temps reel: application a un jeu de roles en gestion de production. Ph.D dissertation, 220 p, Universite des Sciences et Technologies de Lille, France, in French.

Vygotsky, L.S. (1962) Thought and language, Boston, MIT Press. Winograd, T. (1986) A Language Action perspectives on the design of cooperative

work. Proceeding of the CSCW'86, ACM Conference, Austin, TX, 203-220.

20 Distributed Multimedia Environment for Distance Learning

Encarna Pastorl , Gonzalo Sanchez2 and Javier Alvarez2

1 Department of Telematic Systems Engineering (DIT-UPM), Madrid University of Technology, ETSI Telecomunicaci6n, Ciudad Universitaria, E-28040 Madrid, Spain

2 Software de Base, S.A. Torres Quevedo, sin. E-28760 Tres Cantos, Madrid, Spain

Abstract. The paper presents the main findings and thoughts of the first phase of an R&D project (EDUBA) that aims to develop a distributed collaborative multimedia educational environment over broadband ISDN. The adequacy of the unstructured communication paradigm as an approach to the system design is discussed as well as the need of an open distributed architecture which considers all aspects of distribution, processing, specification and communications as facets of a single problem. A prototype simulating the human interface is briefly described in order to highlight the importance of an integrated view of media and tools.

Keywords. Computer supported cooperative work (CSCW), multimedia, broadband communication, open distributed processing, informal communication.

1 Introduction

Research on learning from the constructive and interactive perspectives (post­Piaget) shows that cooperation among actors of the educational process is one of the most important aspects in knowledge acquisition (Mugny 1989).

The scope of our work is to design and develop a distributed educational environment for distance education scenarios where cooperation among teachers, learners, tutors and experts will be supported in a very flexible way. It is focused on the development of multimedia CSCW applications for education over the Spanish broadband network (RECmA) as well as in pedagogical experiments with learners in actual settings.

Advances in computer technology, closely followed by network technologies and telematic applications have played an important role in the increasing use of computer-mediated communication, allowing people to access a wide range of services. Standardization on Open Systems Interconnection and Integrated Services Digital Networks (ISDN) both narrowband and broadband have made the access to these services less difficult. Moreover, enhanced display and multimedia technology have opened up new communication channels in addition to more conventional media.

CSCW, still a young research field, came into focus in this scenario as a powerful mix of disciplines and technologies.

Distributed Multimedia Environment for Distance Learning 259

Paul Wilson (Wilson 1991) states that CSCW demands availability of communication networks and high bandwidth; in-depth understanding of human individuals and group behavior; and specialised application software. A multiplicity of paradigms, models, theories, tools and applications already exist. Taking the words of Power and Carminati (1993) we might characterize the current state of the CSCW field as one of brainstorming.

From the user's perspective, carrying out cooperative work using CSCW tools can be an extremely complex and exasperating experience. Most of the systems and applications currently available have been designed to support very specialised areas of functionalities and remain isolated from the usual work environment as well as from other applications. A comprehensive user support requires a plural approach, integration of current technologies and a careful study of the cooperative work context, ensuring that the new technology does not disrupt the collaborative activity that is already in progress (Bannon and Schmidt 1991).

On the other hand, from the view of the application designers, no general architectural framework has been defined for building distributed CSCW systems, and thus much design effort is duplicated. In particular, it is important that the development of such systems occurs in an open manner which will allow cooperation between systems. The architectural framework must be a reference point for designers to include shared functionalities and take advantage of shared resources.

These two aspects will be addressed in this paper. The following section gives an overview of the project we are currently involved

in as the context for our observations. Subsequent sections discuss the communication paradigm that underlies our approach to the integration of technologies and comprehensive user support. A brief description of the EDUBA user interface (mock-up prototype) as an example of integrated environment is also given. Finally, some considerations are made on the relevance of emerging open distributed processing standards (ODP) for the provision of architectural frameworks to application designers.

2 The Context of our Research

The General Directorate of Telecommunications (DGTEL) jointly with the National Council for Scientific Research and Technological Development and the Ministry of Industry launched a research and development programme on broadband communication networks in early 1992 (PLANBA). It embraces projects from end­user oriented applications to Asynchronous Transfer Mode (ATM) switching technology and advanced networking.

The programme envisaged several applications fields where people and organizations could really take advantage of the broadband communication facilities which are to be offered in the (not so far) future.

One of these application fields was distance education and thus, the EDUBA project was born with the global aims of developing cooperative multimedia applications for distance education scenarios and carrying out pedagogical experiments in actual settings.

260 E. Pastor, G. Sanchez, J. Alvarez

The main concern of EDUBA is to design and develop a distributed educational environment where the cooperation among teachers, learners, tutors and experts be supported in a very flexible way. The environment will provide a set of tools to hold real-time conversations and share information through video, audio and multimedia documents.

The Spanish broadband network RECmA will be the target communication platform, although the prototypes will be implemented in an incremental way fIrst over LAN-Ethernet and then over LAN-ATM platforms to facilitate early educational experiments, so that the evaluation may have a feedback on the design and architectural choices.

Two organizations from industry (Software de Base and Istel) and one university department (DIT-UPM) are participating in EDUBA and, of importance to be pointed out here, there is also an involvement from three educational organizations (GATE-UPM, CAM-IMAF, MEC-PNTIC) which will emphasize the defInition of educational scenarios, defIne the user requirements and conduct the experiments in their respective educational settings.

Then, a positive aspect we expect to exploit in the project is the multidisciplinary profIle of the participants, absolutely needed for success in the CSCW system design adventure.

The project started on the last term of 1m and has a duration for three years. The fIrst phase (6 months) has already been completed. This phase was dedicated mainly to the description of educational scenarios, analysis of users requirements, conceptual design and defInition of a general system architecture. Moreover, a prototype simulating the user interface was developed to validate the conceptual design. The involvement of the educational organizations (in fact the users) was crucial.

Our main fIndings and thoughts from the fIrst phase are presented in this paper. The experience collected by the authors' participation in the CO-LEARN Project (DELTA D2(05) has represented a very valuable input to the EDUBA work. CO­LEARN is thoroughly described in Derycke and Vieville (1993).

3 Cooperative Learning and Communication Support

Cooperation in education could be described as the specific cooperative relationships characterized by common goals and responsibilities shared by a group of people (learners, teachers, tutors, ... ). The global purpose of teachingllearning is reached mainly by interactions among members of the group.

In an environment of distance education using computers and communication networks. interactions for cooperation among the actors of the educational process have to be accomplished by means of communication (information interchange and sharing).

Thus. from our point of view. the choice of a model for communication support influences heavily the character of the cooperation and determines the limits of the interactions.

Several communication paradigms have emerged in the development of cooperative systems. The provision of suitable communication paradigms is an on­going debate within the CSCW community and perhaps one of the mos~

Distributed Multimedia Environment for Distance Learning 261

problematic areas (Rodden and Blair 1991). The following sections give a brief overview of them.

3.1 Structured Communication

One of the most traditional paradigms is the structured communication paradigm, in which people have to organize the communication in a structured and coordinated fashion and a group cooperates primarily by exchanging messages. This paradigm can be found in the work of Winograd and Flores (1986) based on the speech act theory, as well as in a number of research contributions within the CSCW field. Some examples are: Bowers et al. (1988), De Cindio et aI. (1986), Pankoke-Babatz et aI. (1989) and Kreifelts and Woetzel (1986).

Systems based on the structured communication paradigm often assume an asynchronous mode of cooperation and a model of control providing support. Human to human interactions are hardly possible because the technology/system always plays a mediator role, which means regulations and assumptions imposed by the system.

3.2 Communication Through Intelligent Agents

Communication could also be seen from the perspective of the intelligent agent paradigm, based on dialogues between an individual and an intelligent agent. Proposals in this research line are found in Cuena and Garcia-Serrano (1993), Cerri (1993), Guin (1993) and Danielsen and Pastor (1988).

The support provided by the agents in the system and their mediator role is stronger than in the structured communication paradigm. The interacting (human) group is compelled to communicate through agents which, taking different roles, control the rules of cooperation. The social aspect of communication is missed, perhaps one of the most important aspects of collaboration (Kraut et al. 1990).

3.3 Unstructured Communication

The unstructured communication paradigm is perhaps the least formally constructed. Moran and Anderson (1990) discuss it as the informal interaction paradigm for CSCW system design. It is based on multimedia and communication networks technologies, allowing direct interactions between people with the technology itself becoming relatively invisible.

From our point of view, the interest of this paradigm for learning relies on the fact that it allows the design of a contextual tailorable environment and helps the actors of the educational process in organizing the collaborative tasks without the limitations imposed by the system.

Moreover, some experiments on multimedia communication in classrooms (Rosenberg et al. 1992) show that teaching is not only a formal transmission of information but a more informal process in which direct interaction leads to a transformation in both teacher and student.

262 E. Pastor, G. Sanchez, J. Alvarez

The enhancement of the communication tools allowing live interactions by using video and voice channels should accelerate understanding and discussions. Teachers and students would interface to generic communication media, as generic as the telephone This is a new technological scenario for communication and cooperation at distance that will need further investigations on its impact on distance education. It will be addressed in the next phases of EDUBA.

4 Integrated Support: A View of the EDUBA Interface

Part of our design strategy is to adopt the unstructured communication paradigm in our project, where actors of the educational process are engaged in informal, unstructured communication. The technologies employed are broadband networks providing audio and video channels between people located at different places.

Such an approach means that the system should be designed as a set of independent and powerful tools that the actors in the educational process use according to their needs for collaboration. According to Moran and Anderson (1990) these tools should not only serve as support but also as enhancement of the work and the creative development of activities.

On the other hand, this set of tools have to be designed in an integrated environment. One of the aspects of integration that we address here is the user interface that should offer a coherent and consistent view of the system to its users.

4.1 The Room Metaphor

The interface aims to be as intuitive as possible. It is a result of the fIrst phase of the project where the functional and interface requirements of the collaborative tools were defined. It was simulated on a PC under the MS-WINDOWS environment using a video camera.

All its elements try to represent objects usually found in an educational setting. Thus, the fIrst view that appears to the students/teachers whenever they enter in the system is a group of doors (Fig. 1) that let them go into any of the rooms where they will participate in different activities.

Each of the rooms is dedicated to a concrete activity, from work in groups to self­study in the library, and the participants will fInd inside specifIc tools to follow the chosen activity.

The set of tools available in each room is different depending on the role of the participant either teacher/tutor or student. The appropriate environment and tools are determined whenever a participant enters in the system and provides herlhis identifIcation.

The view in Fig. 1 is common to every participant. The fIrst three doors give access to rooms with real-time cooperative tools where groups will be engaged in direct communication with different turn-taking control policies depending on the activity to be carried out: lecture, discussion groups, case studies, tutoring, brainstorming, panel, etc. The room of Tutoring provides also the facility of an asynchronous conference system to support message-based interactions.

Distributed Multimedia Environment for Distance Learning 263

-----l~I------

Fig. 1. A view of the entrance hall

The Library room allows to access to a broad range of learning material that could be either of public or restricted access. In the former case, typical tools are available to consult on-line databases, software libraries, document stervers, etc. In the latter case, a multimedia database managed by the educational organization will permit to browse and retrieve multimedia documents and courseware specially designed for the group of students.

The last room is a workshop where students, teachers and tutors will have access to any other tool normally found in a workstation for personal work: editors, interpersonal communication system (electronic mail), etc.

The suite of rooms as metaphor for the user environment has proved to be very useful and intuitive for users. Alexander et al. (1993) describes a similar user environment for interactive learning, ILSE, where students exchange messages and have discussions in a meeting room. Meeting rooms are also linked transparently via communications. The difference in the EDUBA prototype is that such links for the real-time rooms include live audio and video channels so the students are in active contact with every other member of the group.

4.2 The Teacher's View of a Lecture

Figure 2 shows an example of the Classroom room from the point of view of the teacher when an activity of lecture has already started.

The Classroom activity allows a teacher/expert to give a lecture/conference to remotely located students using multimedia material previously elaborated with the help of the Workshop tools. All the participants are attending in real-time.

In Fig. 2 three different windows appear: the shared window, the video window and the control window.

264 E. Pastor, G. Sanchez, J. Alvarez

J 0 URN A L > ~ ..

PROGRAMMING ~+++=-I-¥-Rr+-t'i MUlTlf,UDIA

·;IfIli.Wo,Gl>..U;" · ~,...;I, It~ ~JO ''' i: - ,.. t.· III . .m.(Q\!t~ ::i.

~~rS~~ '.: ....... ·L· ...

!~:,c:.t·l!·I" (-':'~'{0jf-.ul:l"'(

Fig. 2. A teacher view of a lecture

;,i

,';,'

:111" ..... _ .

I _~. .'. t ... '

~.",:" · ~",,I " ~

The shared window follows the WYSIWIS (What You See Is What I See) concept. The teacher, who has initially the control over the shared window, will put some (multimedia) document or slide on this window and it will be shared for all the remote participants. All the modifications to the document are transmitted to the participants in real-time. The documents can include text, graphics, motion video and hand drawn figures.

The upper left window (video window) shows one of the participants. The teacher will be able to choose the remote location from where s(he) wants to receive the real-time video image, while remote attendants to the lecture will always receive images from the teacher location. However, in some situations the participants in the lecture are allowed to switch the video window to local mode so that they can also see their own image. The video window could also be used for presenting objects. The main tools in the control window specific for the teacher role are briefly described in Fig. 3.

The rest of the tools in the control window allow the teacher to navigate through pages/links of the multimedia document to be shown on the shared window, to control the sound levels, to start and to close the lecture/session.

4.3 The Student's View

The tools in the control window of students are quite different as they can not take as many actions as the teacher. Figure 4 presents the aspect of an on-going lecture from the point of view of the student attending to the lecture.

The attendants may request the control over the shared window when they like to have the opportunity to ask some questions and draw/write on the shared window. They also may take notes about the lesson or consult/edit documents.

Distributed Multimedia Environment for Distance Learning 265

It gives the control over the shared window to a selected person who has reviousl re uested it

A personal notebook will be available to take notes about the lesson or consult/edit some document It interchanges the positions of the shared window and video window

A drawing tool will be avai lable to write and draw on the shared window

The movements of the local pointer are made visible on the shared window

the shared window is

Fig. 3. Available tools in the teacher control window

~~ ... DIFERENTES MODELOS DE AGRUPACIO . DE RASCACIELOS

Fig. 4. A student view of a lecture

266 E. Pastor, G. Sanchez, J. Alvarez

5 Architectural Framework

Taking the view of the application developers, the development of distributed multimedia applications and tools is not easy. The designers and developers have to cope with the complexity of using heterogeneous platforms, closed worlds from different vendors, where application interoperability and portability is very difficult, almost impossible.

On the other hand, the developers have to be familiar with different technologies, from communication networks, protocols, operating systems, data bases, ... , to audio and video devices, synchronization, management and control, etc.

This technological scenario leads to the need of a common architectural framework as the basis for a platform to support the development of distributed multimedia applications (Pastor and Jager 1993). Such a framework might facilitate the portability, interoperability and openness of applications. It would hide irrelevant infrastructure details managing transparently networks and protocols. Finally, to a greater extent, it would reduce the development complexity and avoid duplicated effort.

Several research contributions has been made along this line. See, for example, Herbert (1989), Magee et al. (1990).

The standardization bodies are als~ paying attention to the issue because they are aware of the growing need for an architectural model covering the aspects of cooperation between systems and applications. The architectural model will be a reference point for designers. Thus, efforts towards a reference model on Open Distributed Processing (ODP) are progressing (ISO 1991) to provide the basis for the common architectural framework discussed above.

Both the CSCW and ODP fields share mutual interest and have similar aims and goals developed from different perspectives. While ODP standardization considers portability, interworking and distribution problems of distributed systems, CSCW has a focal point on the use of computers to support group work. Navarro et al. (1991) analyses this important relationship.

However, the provision of appropriate facilities for collaboration support may require a careful reexamination of some of the decisions embedded in the emerging ODP standard. In particular, as Blair and Roden (1991) analyses, research on support for multimedia cooperation is still in its infancy and it is poorly addressed in the current standard. This issue as well as the support for high speed communication is object of research in EDUBA.

6 Conclusion

Many ingenious systems and prototypes for collaborative work have been generated in research laboratories and tried out in pilot experiments. The technical report of RARE (Adie 1993) is a survey of the state-of-the-art in multimedia networking. It identifies more than 50 current research projects and describes around 30 products already commercially available or in the public software domain.

However, a closer look at that survey reveals that only a few prototypes! products support cooperation. Most of them have been developed in order to

Distributed Multimedia Environment for Distance Learning 267

experiment a new technology and are too technology-centered. Systems are designed and developed first, and possible uses and users are tried to fmd afterwards. As Power and Carminati (1993) point out, when launching a new technology such an approach may be inevitable: people did not realize that they needed Walkman cassette players until Sony produced them. It is true, and we share interest about research on multimedia technologies and broadband communication. Thus we have affinity with the technology-centered approach. However we also believe that our concern must be not only with the enabling technologies but also with the use of technologies.

The objective to reach is to combine features and potentialities of the computer and broadband network technologies with user needs in the specific educational context.

In this line and as result of the work in the first phase of EOUBA, some considerations may be emphasized:

- A multidisciplinary team is needed for success in the design of the collaborative learning system and to satisfy the users' expectations. User involvement from the very early stages of the design activity is crucial.

- Mock-up prototypes facilitate early educational experiments so that the evaluation may have an important feedback on the design and architectural choices.

- The communication support model influences the character of cooperation. Control and support should be enabling rather than constraining. The informal interaction paradigm is, from our point of view, appropriate for learning purposes, even more when the communication media is enhanced by using live video and voice channels.

- The design of the system as a set of independent tools may lead to a user controlled, flexible and tailorable environment that users could manage according to their needs

- The development of the system is dependent on the facilities provided by existing distributed platforms, but the properties of the collaborative systems question established principles of such platforms. Openness, reliability and easy integration are the major conditions which determine the success of the system. OOP as an emerging standard provides potentially the basis to fulfil these conditions and may be used by developers as a referential architectural framework. However the OOP proposal in its current state needs further investigation on the support for multimedia collaboration.

Acknowledgements

Special thanks to Felisa Verdejo who encouraged us to write this paper. Javier Alvarez and Miguel Arjona were the developers of the EOUBA prototype and they presented a nice demo in Segovia. We would also like to thank the other members of the project for their contributions to the work. Our acknowledgements to the partners of the CO-LEARN project for most of the inspiring ideas. The EOUBA project is partially funded by the PLANBA Programme.

268 E. Pastor, G. Sanchez, J. Alvarez

References

Adie, C. (ed) (1993) A survey of Distributed Multimedia. Research, Standards and Products. RARE Technical Report 5, 1st. edition. RARE Secretariat. Amsterdam

Alexander, G., Lefrere, P., Matheson, S. (1993) Towards Collaborative Learning at a Distance. In this volume

Bowers, J., Churcher, J., Roberts, T. (1988) Structuring Computer-Mediated Communication in COSMOS. In: Speth, R. (ed.) Research into Networks and Distributed Applications. North-Holland, Amsterdam

Bannon, L.J., Schmidt, K. (1991) Four Characters in Search of a Context. In: Bowers, J.M., Benford, S.D. (ed) Studies in Computer Supported Cooperative Work. North­Holland

Blair, G.S., Rodden, T. (1991) The Impact of CSCW on Open Distributed Processing. Proceedings of the International Workshop on Open Distributed Processing. Berlin

Cerri, S.A. (1993) The "Natural Laboratory". Methodology Supporting Computer Mediated Generic Dialogues. In this volume

Cuena, J., Garcia-Serrano, A. (1993) Intelligent Computer Support. In: Power, R. (ed.) Cooperation Among Organizations. The Potential of Computer Supported Cooperative Work. Springer-Verlag

Danielsen, T., Pastor, E. (1988) Cooperating Intelligent Agents. In: Speth, R. (ed.) Research into Networks and Distributed Applications. North-Holland, Amsterdam

De Cindio, F., De Michelis, G., Simone, c., Vasallo, R., Zanaboni, A. (1986) CHAOS as a Coordination Technology. Proceedings of the Conference on CSCW. Austin, Texas

Derycke, A.,C. Vie ville, C. (1993) Real-Time Multimedia Conferencing System and Collaborative Learning. In this volume

Guin, D. (1993) A Model of Interaction between an Artificial Agent and a Human Agent. In this volume

Herbert, A. (1989) The Advanced Networked Systems Architecture. In: Distributed Systems. Mullender, S. (ed.) Addison-Wesley

ISO/IEC JTCI SC21 WG7 (1991) Basic Reference Model of Open Distributed Processing. Part 1: Overview. Working draft. Dec. 1991

Kraut, R.E., Egido, C., Galegher, J. (1990) Patterns of Contact and Communication in Scientific Research Collaboration. In: Intellectual Teamwork. Social and Technological Foundations of Cooperative Work. Galegher, J., Kraut, R.E., Egido, C. (eds.) Lawrence Erlbaum Associates. Hillsdale, New Jersey

Kreifelts, T., Woetzel, G. (1986) Distribution and Error Handling in an Office Procedure System. IFIP WG8.4 Working Conference on Methods and Tools for Office Systems. Pisa

Magee, J., Kramer, J., Sloman, M., Dulay, N. (1990) An Overview of the REX Software Architecture. 2nd. IEEE Workshop on Future Trends of Distributed Computing

Moran, T.P., Anderson, R.J. (1990) The Workaday World as a Paradigm for CSCW Design. Proceedings of the Conference on CSCW. Los Angeles

Mugny, W. (1989) Psychologie Sociale du Developpement Cognitif. Peter Lang. Berne. Navarro, L., Prinz, W., Rodden, T. (1991) Open CSCW and ODP. Proceedings of the

International Workshop on Open Distributed Processing. Berlin Pankoke-Babatz, U. (ed.) (1989) Computer Based Group Communication. The AMIGO

Activity Model. Ellis Horwood, London Pastor, E. , Jager, J. (1993) Architectural framework for CSCW. In: Power, R. (ed.)

Cooperation Among Organizations. The Potential of Computer Supported Cooperative Work. Springer-Verlag

Distributed Multimedia Environment for Distance Learning 269

Power, R. Carminati, L. (1993) Computer Supported Cooperative Work. In: Power, R. (ed.) Cooperation Among Organizations. The potencial of Computer Supported Cooperative Work. Springer-Verlag

Rodden, T., Blair, G. (1991) CSCW and Distributed Systems: The Problem of Control. In: Bannon, L., Robinson, M., Schmidt, K. (ed) Proceedings of the Second European Conference on Computer Supported Cooperative Work. Kluber A.P., Amsterdam

Rosenberg, 1., Kraut, R.E., Gomez, L., Buzzard, A. (1992) Multimedia Communication for Users. IEEE Communication Magazine, May 1992

Sinderen, M., Schot, J. (1991) An Engineering Approach to ODP Systems Design. Proceedings of the International Workshop on Open Distributed Processing. Berlin

Wilson, P.A. (1991) Computer Supported Cooperative Work: Origins, Concepts and Research Initiatives. Computer Networks and ISDN Systems, no. 23. North-Holland

Winograd, T., Flores, F. (1986) Understanding Computers and Cognition: A New Foundation for Design. Ablex Publishing Corp. New Jersey

Zorkoczy, P. (1993) Educational Scenarios for Telecommunication Applications. In this volume

21 Production of Flexible and Modularized Course Material in COSYS 1

Mette Ringsted, Finn Grj1jnbrek, Peter Busch, Hanne Shapiro

Danish Technological Institute, Teknologiparken, DK-8000 Aarhus C, Denmark

Abstract. This paper discusses aspects of the production of flexible learning material and co-authoring issues in a DELTA pilot project COSYS, where five end user testbeds are set up across Europe to explore new approaches to the production of flexible (and multimedia) learning materia1.

Keywords. COSYS, co-authoring, production environment, flexible material, modularized course production

1 Background

Jobs associated with new technology are rapidly making many traditional skills obsolete and are fundamentally changing the way work is organized and organizations are managed, leading to new skills requirements, new working patterns and new organizational forms.

Interdisciplinarity and the demand for cross-organizational skills put new requirements on the further educational environment: publishers, producers and providers - public and private actors.

There is a growing need for a different production and delivery infrastructure in terms of:

• Cross-organizational course development with publisher/producers and providers • Cross-institutional provision, tailoring and support of courses • Cross-institutional marketing and administration taking place in a common

framework guided by a higher degree of customer and market orientation with emphasis on flexibility, adaptability and speed (just-in-time production and delivery).

Adequate infrastructures and systems that bring together geographically dispersed major European actors in a cohesive manner are a prerequisite for further support of:

• Recurrent training schemes and are therefore a prerequisite in the restructuring of production and service sectors and a sine qua non to ensure future competitiveness for European industry.

I Copyrights belong to the authors of the paper.

Production of Flexible and Modularized Course Material in COSYS 271

• A commercial set-up for European course production encompassing the whole production and delivery cycle.

1.1 An Example

Let us for a moment look at a typical training organization in Europe offering continuing training to industry.

An enterprise has identified a training need in a given technological area. The company requirements generate questions to the training organization that will often be:

• We need this training within the next few months, what can you offer us? • Is there any material already available, and is it possible to tailor it, so

it matches our specific company needs? • Will it be possible to run the course as on-the-job-training, and how can we

integrate our own personnel in the process? • Will it be cheap enough for us even if the group of students only consists of

five individuals? • Do we have to develop the whole course from scratch, or is it possible to

combine different sources of material with something new?

With the current course production and delivery structures available, the problem is difficult to solve for most training institutions in a rational and a cost-effective way, especially if it requires input and expertise from different experts.

On the other hand, as it is known also from the recent IRDAC report on skills shortages in Europe, training needs in many companies are typically defined by:

• Being cross-disciplinary in nature • Encompassing different levels of employees in the company • Having both a general managerial and a more specific technological component • Needing to be solved here and now without costing a fortune.

So there is a problem!

Both from a user's and a producer's point of view, traditional ways of producing course materials are neither really efficient nor cost-effective. There is a long life cycle for learning material produced in relatively large batches. Material may be outdated before it is sold out, or it is hard to update, and will therefore often not correspond to rapidly changing and diversified qualification needs.

Anybody who has worked in a distributed course team, be it in an organization across departments, nationally, or internationally will have most certainly experienced the frustrations of handling the entire production and delivery process. It may be related to issues such as version control, which tools to use, layout, pedagogical design, and overall managerial issues linked to the course production phase.

A more extensive use of distance and flexible learning will undoubtedly require a more extensive course materials supply and new types of information about available course materials. It will also require a number of value added services

272 M. Ringsted et al.

facilitating the easy maintenance, access, redesign, delivery, billing and copyrights handling of educational material.

It was against this background and these experiences that the concepts behind the COSYS system were conceived.

2 The Aim of the COSYS Project

The aim of the COSYS project is therefore to pilot a distributed course production and delivery system which will facilitate:

• Exchange and handling of information concerning training demands and supply.

• Tailoring, translation and version control of existing material. • Co-authoring with:

- On-line access to raw material from the material database. - On-line help facilities and help desk functions. - Communication facilities such as e-mail, computer conferencing

and groupware. • Use of different output formats depending upon technological platform among

users, concerning text, examples could be ASCn, RTF, MS-Word, etc .. • Use of different distribution channels depending upon technological platform

among users, e.g. paper, diskettes, CD-ROM, On-Line • Management of administration, security, billing and copyrights. • Generation of training materials for a given target group by accessing,

browsing, picking and mixing material from a database.

The COSYS project will explore the implementation of:

• Bandwidth on demand, i.e. adequate communication solutions in the pilot environments (dial-up lines, X.25, ISDN).

• Printing/publishing on demand, i.e. customized output formats, structure, layout, quality and number of copies and delivery media.

• Training on demand, i.e. the integration of information and training systems facilitating the production of course material to certain target groups within a very limited time.

3 The COSYS System

3.1 The Concept of the COSYS System

The concept of the COSYS system is on-line databases containing course material accessible via telecommunication lines throughout Europe. The system will support a set of functions covering:

• Collaborative authoring of course material • Archiving of course material

Production of Flexible and Modularized Course Material in COSYS 273

• Ordering of course material • Delivery and Distribution of course material • Management and Administration

The system will support flexible delivery on-demand of user tailored course materials both printed on paper and as interactive, hypermedia documents.

Through remote access and telecommunication services, the system will be able to support the development, production and delivery of course materials throughout Europe.

Common production tools such as commercially available word processors, desktop publishing and hypermedia software will be supported as both input and output interfaces to the course database.

3.2 The Application Domain

The pilot application proposed will demonstrate and evaluate five different scenarios all based on existing training activities. They will include time saving procedures for the production and distribution of services and new approaches to course production. There are five pilot applications:

1. EGIN, a cooperative cross-European network of SMEs and branch organizations as a distributed production and delivery unit in the graphical branch. 2. COS TEL, a network of large-scale traditional public and private providers of open learning courses based on self study material, video and audio - cooperating in a COMEIT II pilot project - co-authoring expert material - locally/nationally tailored and implemented with on-line teleconsultancy services. 3. Dee, a Multinational European producer of copiers, plotters and laser printing systems testing and evaluating the practical use of the COSYS model and prototype in the process of producing, reviewing, translating and maintaining user manuals. 4. Philips, in-house industrial training represented by a multinational company testing the course production and delivery system integrated with information systems to support on-demand training and product information services in a distributed organization. S. Industriens Forlag, a publishing company represented by a commercial national publisher testing production and delivery of training materials to course providers and trainers.

A survey in these areas with regard to State of the Art and user requirement will form the basis upon which a generic model will develop which will be a central part of the prototype development of the system.

3.3 The Target Audience

The multimedia course production and delivery system will be seen as an integrated part of the user environment, where the interaction between publishers, authors,

274 M. Ringsted et al.

producers, customers (providers and teachers) and the use and re-use of course (modules) will determine the structure and the interfaces in COSYS.

The five pilot test environments all represent a technology level past first generation of computing and flexible learning production and provision. It is only because the test environments are already working within this field, that it is possible to have five different test sites with such a limited budget.

The five test environments together represent a very diversified field within the area. As such, the test sites will be real-life working environments representing the following types of actors:

• Publishers including publishing organizations with staff dealing with copy rights and price structures, electronically based services, management and co-authoring of materials, stylesheets for DTP and metaphors for hypermedia.

• Technical and administrative support for production, archiving and delivery of flexible learning materials and help desk services.

• Producers of printed materials (printing on demand), video, audio, CBT and CD-ROM.

• Distributors of flexible learning materials. Distribution of printed materials, video- and audio cassettes, CD-ROM, floppy disks with CBT materials and electronical distribution of learning materials through the pick and mix system.

• Authors including designers, writers and editors of flexible learning materials. • Providers of flexible learning materials, libraries of flexible learning modules for

pick and mix of learning materials. • Trainers who can tailor training materials through the pick and mix system

targeting the students.

3.4 Services - The Flexible COSYS System

The COSYS environment with the pilot application cover a wide range of aspects of distributed course production and delivery ranging from company internal services to a publishing house offering publishing on-demand.

To cover the different basic requirements of the pilot application, the COSYS system will be modularized and flexible, giving the pilot application the possibility of picking and mixing from a wide range of services to make a customized version of a course development, production, and delivery system.

As an example, the pilot application have different needs and approaches concerning copyrights and royalties. They vary from company internal security needs for managing copyrights of distributed material to the extent of a publishing house wishing only to present highly compressed picture material on the users' screen to ensure that no violation of copyrights due to Print-Screen Commands will be made during pick and mix from the COSYS system.

To obtain this high-scale flexibility for the COSYS system, the functionality of the system is divided into a set of services, making it possible for the user to choose a customized version of the COSYS system to fit exactly his needs.

• Some of the services in the system are internal, whereas others can be accessed externally by the end user.

Production of Flexible and Modularized Course Material in COSYS 275

• Some of the services are provided as integral parts of the system, whereas others may be added later by the end users.

• Some of the services are provided by software tools or communication services, whereas others are provided as human actions and communication.

3.5 Workbenches - The Modularized COSYS System

The services are integrated and organized on four workbenches: • Management and publishing workbench

The management and publishing workbench will provide services for administration, management and control of the development, production and distribution of course materials.

• Collaborative development workbench The collaborative development workbench will provide services for the creation of courses and training materials on paper and as multimedia.

• Production and distribution workbench The production and distribution workbench will provide services for ordering, retrieval, master creation, duplication and distribution of course materials.

• Systems maintenance and user support workbench

The system maintenance and user support workbench will provide services for maintaining the technical systems and services, and provide users in the environment with adequate support.

4 The Co-authoring Environment in COSYS

4.1 Co-authoring in COSYS

The Co-authoring definition in COSYS covers a number of activities. The overall environment can be characterised as distributed on-line production of flexible (multimedia) course material. In the different test environments the three issues: distributed, on-line and multimedia are optional, but basically the general demands for the test beds are that they should be able to support these features.

Setting the stage of co-authoring in this environment raises· the demands for specifications of functions including the co-authoring process and functions undertaken elsewhere. Also there is a need to specify the roles in the processes, because they are not necessarily transferable from other (and more traditional) production environments. Even if the approach to the definition of these roles and functions is very structured at a general level in the project, the actual organization of these roles varies from testbed to testbed in the way different persons will undertake the roles and execute the functions.

276 M. Ringsted et al.

4.2 Functional Description

The definition of the co-authoring environment encompasses the following functions:

• Project initiation The trigger is usually the requirement from a client for a new publication or course, or the identification of a potential market. The project initiation support facilities should encompass a short guide to the format of the specific material to be produced: Course aim, objectives, training plan, training aids and materials, training resources, course duration, target trainee profile, trainer/instructor profile, copyrights and acknowledgements to mention some.

• Instructional design This is the process where the master plan for the training materials is made. The instructional designer gives guidelines to and discusses developments with content creators. This aspect of co-authoring either takes place in a collaborative fashion, with the instructional designer working with the content creator or on a separate basis - share tasks are allocated by editors or project leaders. Both these models need to be supported by the COSYS system facilities.

• Production Planning The aspect of production planning covers the identification and allocation of the components of a course/publication to different content creators. As such, it is only a subset of management and administration of a production which is included in the definition of co-authoring here. From the COSYS point of view the tools to support co-authoring should be those developed to support the general management and administration, rather than tools specific to co­authoring.

• Content creation Content creation is the core activity of co-authoring, the process where materials are actually developed, the authors are writing and the media producers create graphics, audio etc. This process follows the guidelines set out by project management, instructional designers and editor.

In the COSYS project the primary requirements for content creation are first to support text and hypertext authoring and then support the authoring of other media. Furthermore, the collaboration can be described more like serial collaboration than actual CSCW.

The test environments want to be able to support their authors on common word processors by providing them with stylesheets and help facilities rather than requiring them to use applications with which they are unfamiliar, however optimal they may be from other perspectives (database, open information interchange). On the other hand, in the test environments there is interest in seeing the potential of more sophisticated tools for coauthoring, so these should be supported on the workbench.

Production of Flexible and Modularized Course Material in COSYS 277

• Project management The project is managed according to the projectplan, made by the project manager as a result of discussions with the involved partners. Project management allocates the needed staff to the project.

• Editing Editing of materials is divided into pre-and post-editing. Pre-editing is setting up guidelines (outlining and annotation) to the authors. Post-editing is reviewing the developed materials, i.e. quality coritrol concerning the content and the guidelines.

• Content composition This activity encompasses the extraction of content from a database and presenting it to the author in a suitable form, and extracting content from the authoring tool and putting it in the right place in the database. It is not necessarily the purpose of content composition in co-authoring to

carry out composition for final publication, although in the case of content with a simple presentation this may be done simultaneously. The use of stylesheets (which integrate composition and content creation) is again a requirement, with an alternative being an SGML sensitive editor.

• Layout mastering The layout mastering is the provision of formats for content in a form that will be suitable for customers of the COSYS system. There are two aspects to this: - Final layout for the complete document. - Access via the pick and mix system. A need has been identified for stylesheets/formats which can give an indication of the presentation of the full document and support the generation of order forms etc.

• Browsing, picking and mixing This covers giving the customers the ability to browse only down to a defined level of detail: Picking and mixing takes place by ordering components from the content provider.

Authors on the other hand, need to browse and have the facility to modify materials. Their need can be met by loading content into stylesheets. Material will be transferred to local sites in some testbeds whereas in others it will be held in central databases for online querying.

These elements serve as the basic co-authoring workbench building blocks in the COSYS pick and mix system.

4.3 Synthesis of Requirements for Co-authoring Support

An analysis of the specifications indicates that the requirements can be consolidated to:

278 M. Ringsted et al.

1. Help files to support authors in their use of tools and their interpretation of specifications. 2. Authoring stylesheets or context sensitive editors for content creation and content editing. 3. Presentation stylesheets for potential customers downloading elements of course material. 4. Applications which can extract content from stylesheets for automatic placement in databases. 5. Applications which can extract content from databases for automatic placement in a content viewer, made available to potential customers. 6. An application that customers can use to browse content on remote or local databases.

4.4 The Roles in the Process of Co-authoring in COSYS

The definitions given below refer to roles and not necessarily to separate individuals in all cases. In many cases the roles of editor and instructional designer, or author and instructional designer can be carried out by the same person. Some of the roles will be included in the co-authoring process and some of them will cooperate with the roles in the co-authoring process. The specific co-authoring environment in COSYS puts the focus on the new roles needed to support this specific co-authoring process, even if they are not specificly part of the actual authoring task as such.

4.4.1 Co-authoring Roles

• Training advisor The training advisor has the first contact with the initiator. He/she will discuss and analyse the objectives and the framework of the possible project.

• Instructional designer The instructional designer designs the 'blueprint' for a course or set of training materials for a given target audience, from an initial specification in tenns of a perfonnance problem, a training need analysis, or a content or skills descriptions, considering the relevant economics, technological, pedagogical, cultural, socio­psychological and resource constraints; and the fonnative and summative evaluation of the courses/training materials.

• AuthorlI'echnical Writer (content creation) The author creates the written material and the scripts for audio-visual (A V) and computer-based training (CBT) material based on inputs from the instructional designer and content expert roles.

• Co-author The co-authors are the group of authors who develop training material stored in the database.

Production of Flexible and Modularized Course Material in COSYS 279

• Content experts The content experts are persons with expert knowledge and/or skills in a specific subject or task.

4.4.2 Other Roles

• Initiator The role (or need) that initiates the production of training or educational materials for a specific purpose and target audience and decides on the final acceptance of the materials that are developed.

• Manager The production manager is the role that undertakes the complete planning, coordination, budgeting and scheduling of the creation, development, testing, production and maintenance of training or educational material.

The delivery manager is the role that perf ormes the management on the customer side, including organizing support for products and communication of user needs back to the development team.

• Graphics Designer The role that creates and implements coherent design principles for text and associated media materials, in collaboration with the instructional design, authoring, and media production roles. Production of any graphic materials according to these design principles.

• Editor The role that converts written material, moving from the authoring aspect to the formatting in a quality suitable for final production and duplication.

• Media Producer The role that converts scenarios or scripts provided by an author (and graphic designer) into mediated material, i.e. DTP on paper-based materials, AV material, CBT material, simulation software, multimedia material.

• Programmer The role that develops the software for computer-based training applications, interface drivers etc.

• Database Administrator The role that takes the responsibility for the physical and logical consistence and maintenance of the database, i.e. designing database structure, granting user access, handling the version control, back up and clean up, etc.

• Translator The role that undertakes the translation of course material from one language into another.

280 M. Ringsted et al.

• User The user of the COSYS system includes the end users (the customers) who can use the system directly or through an intermediary person (e.g. training advisor or sales person).

• Selection and ordering support The Selection and ordering support supports the user, while he/she is selecting and ordering (i.e. pick and mix) materials from the database.

• Help desk The help desk function supports the customer/users in using the COSYS System. This is on-line or phone/fax support which ensures short response time.

• Reproducer The material reproducer receives master copies of materials and produces the needed amount of copies.

4.5 Production Environments

The development of training and educational materials in an environment like COSYS applies new roles and functions to the process. The production of materials moves to a professional publishing environment, and the focus of development moves from the content creation to the professional management of the production process.

New traditions and types of jobs will evolve, and a lot of skills not found within the environment of developing traditional training materials will be required.

High-level project management skills are required to manage all resources in the process of co-authoring materials. The editor needs overall knowledge of the database content and the ability to structure material in such a way that it is suitable for re-use in other configurations. Authors need skills in computer mediated communication (e-mail, computer conferencing, groupware) and a basic knowledge in using stylesheets with their word processor. A user support/help desk function will be necessary, and computer literates are needed for running the system both in terms of database management, groupware, user interface design and multimedia authoring.

New roles and new functions require new skills, and new skills require retraining and education of the people involved in this new production form.

There are basically two types of production environments in COSYS:

• Professional publishing environment • Network cooperation environment

These two environments will have to organize the work according to different conditions:

• Distributed/central authoring • Existing profile of personnel • Business environment/project network

Production of Flexible and Modularized Course Material in COSYS 281

• Existing technology platfonn and knowledge within the usergroup • Workflow

Basic questions in this process of organizing work will be:

• Which new roles will be necessary and how can they be organized in an adequate work process?

• What are the existing skills of the staff and authors in the environment, and what kind and level of retraining and education is necessary?

• How much will technology be able to support or, alternatively, automate different

functions in the production process?

These are some of the key issues that will be focussed on the next two years in the test of a generic COSYS system.

The testbeds start now and finish next year when hopefully we can make some conclusions and point out specific points for further development and investigation.

References

DELTA 2011, COSYS Technical Annex: Design and implementation of a computer based course production and delivery system COSYS

DELTA 2011, COSYS Deliverables: DIWPOI-I: Report on market needs DIWPOI-2: Report on technical possibilities and constraints DIWP02.I-l: Description of a generic scenario based on the application pilots DIWP02.2-1: Specification of user requirements and services DIWP03: Description of the architectural model for distributed course production delivery DIWP04.I-I: Implementation of a generic multimedia database -phase I DIWP05-1: Implementation of management and administrative tools DIWP06-1: Co-authoring workbench - phase I DIWP07 -I: Implementation of pick & mix system - phase I DIWP08-1: Communication network and services - phase 1, distributors and users DIWPI 0-1: COSYS Evaluation Handbook. An external perspective.

Conferences:

Telecommunication in Education and Organization, Aarhus, Denmark August 1992 Shapiro, H. and Busch, P. (1992) Production of course materials in a Multimedia­

database and on-line services

22 OSCAR: A System for Collaborative Distributed Authoring of Multimedia Training Materials

Antonio Ulloa, Antonio De Girolamo, Stephen Delaney

Tecnopolis CSATA Novus Ortus, Str. provo per Casamassima KIn 3, 1-70010 Valenzano (BA), Italy

Abstract. The paper gives an overview of the OSCAR system supporting collaborative and distributed authoring of multimedia training materials. OSCAR refers to an application scenario featuring the cooperation of multiple actors and the distribution of the authoring resources such as authors, authoring tools, information base and equipment. Increasing needs of uniformity, quality, productivity and reliability of the courseware development process along with increasing opportunities for large scale training projects are making collaborative and distributed development of training materials one of the most promising sectors of the educational technology market. OSCAR, based on an open system architecture, provides multimedia communication facilities, a common information repository and high level services (co-authoring services) supporting multimedia authoring, collaboration, coordination, co-decision and re-use in courseware projects. In what follows the functionalities and the architecture of OSCAR are described with a special emphasis on the coordination facilities offered by the system. The OSCAR system is being developed within the CEC DELTA Programme.

Keywords. Co-authoring, workspaces, coordination, collaboration, co-decision, re-use, open system architecture, joint courseware development.

1 Introduction

Over the past years the needs of uniformity, quality and productivity of the course development process are increasing. At the same time, the opportunities for collaborative course development projects in which multiple organizations are involved with different roles are also increasing. This is particularly pushed by the current process leading to a single European market, that is deeply affecting training and education policies. To make the trend more evident it is enough to recall some initiatives and frameworks of joint course development at European level: the CEC COMETT and Euroform Programmes, the ESA CBT Programme, the network of the European Open Universities, the big European companies involved in transnational training projects. And in general, collaborative modes of course development have been adopted by courseware producers and publishers who have set up a systematic production process including, in most cases, a team approach and/or a distributed development process. Courseware design and production tools available today on the market still fail to address all the problems

OSCAR: A System for Collaborative Distributed Authoring 283

of courseware development in a collaborative and distributed way. One problem is that courseware development projects for all but the most simple courseware. involve many different people, representing several different skills:

- Subject matter experts organizing the material to be presented. - Courseware designers laying out the instructional strategy to be followed. and

the details of the student interactions in that strategy. - Graphic designer creating interactive screens. - Programmers implementing in a programming or authoring language the

resulting instructional strategy and student interactions. - Evaluators assuring the resulting courseware is both free of errors and meets the

instructional objectives.

The tools we have today neither recognize this diversity of skills. nor allow the people possessing them to work together effectively. Secondly. the existing authoring tools are primarily aimed at the software implementation phase of the development process: they don't help the authors in the early stages of the development process. Thirdly. the authoring tools available today on the market do not support the concept of courseware life cycle. including maintenance. re-use, modification and update to correct or make more current the content of a course. or the adaptation of a course from one spoken language to another. or from one industry to another, or from one kind of equipment to another. In order to overcome some of these problems. the OSCAR system offers advanced facilities specifically devoted to joint courseware development projects.

2 The Reference Scenario

OSCAR specifically addresses the joint courseware development process and refers to a scenario featuring the cooperation of multiple actors and the distribution of the elements involved in courseware development projects (such as authors. authoring tools. information base and equipment). The following figure outlines the reference scenario. identifying the main technological elements and processing environments (Fig. 1).

The authoring workstation is the execution environment of an authoring application and represents the physical workplace of an individual author. The authoring workstation is essentially based on a multimedia personal computer.

The authoring workstation is a part of the workgroup environment where an aggregate of authors work on strictly related activities. sharing common information. physical devices. applications and communication facilities. Within a workgroup environment. local authoring workstations are networked by means of a Local Area Network (LAN). Remote authoring workstations can be also connected to the workgroup environment via Integrated Services Digital Network (ISDN). The workgroup server is the natural location for common information and communication facilities. The workgroup environment plays a central role in OSCAR as most of the cooperative services are designed to serve the needs of a workgroup involved in a courseware project.

284 A. Ulloa, A. De Girolamo, S. Delaney

..................................

I . I I I I I I I

1":" ALIIIIGrinp ~ wworkstlilDn -

. . . . . . . . . . . . Fig. 1. The reference scenario

• • • • • • • • • • Joint cw Project Envitonment • •

The workgroup environment is the fundamental element of a joint courseware project environment which consists of a collection of workgroups cooperating in a given project over several sites, even remotely located. The workgroups perform parallel development of pieces of courseware or execute different activities of the courseware development process. In both cases, they need to exchange information and integrate their results. The interconnection of remote workgroups is realised through geographical links based on ISDN and satellite.

The joint courseware project environment concides with the workgroup environment when a project does not need to be distributed over different sites and over different workgroups.

3 OSCAR Architecture

The services provided by the OSCAR architecture have been grouped in layers to represent better their organisation and relationship.

The OSCAR layers are shown in Fig. 2. The layer representation of the OSCAR architecture helps one to understand the

scope of each service and the relationship with other services. Within a layer the following concepts are defined:

- Space: It offers a collection of services which contribute to the same general objective.

- Component: It is a key OSCAR object, representing a well-defined set of functionalities that can be used as a single block.

OSCAR: A System for Collaborative Distributed Authoring 285

Fig. 2. The OSCAR architecture

Each Space is made up of several components. For instance the Object Oriented Data Base Management System is a component of the Common Information Space.

The OSCAR architecture can be viewed as divided in two logical parts: the infrastructure and the co-authoring services. The OSCAR infrastructure includes: Hardware Platforms; Operating Systems; Communication Space (CS) which supports multimedia information exchange and allows the access to local and/or remote data bases of existing multimedia training materials; Common Information Space (CIS) which holds the information objects generated during the courseware development ensuring integration, distribution, sharing and re-usability over different authoring activities and actors.

Co-authoring Services have been logically grouped in a co-authoring Space (CAS) and are based on the above infrastructure layers for supporting the following functions: Authoring of multimedia courseware, Coordination of the interdependencies between authoring activities performed by multiple actors, Collaboration between multiple actors in a given authoring activity through the sharing of information objects and tools, Co-decision as a joint decision making process about issues relevant to the courseware design and production, and Re-use of existing multimedia training materials and/or half-fabricates for new courses.

286 A. Ulloa, A. De Girolamo, S. Delaney

The OSCAR Author Desktop (OAD) provides user interface functions, allowing applications to interact with people in a consistent and intuitive way. OAD offers tailored views of OSCAR which can be considered as a sort of virtual workspaces for the different actors involved in a courseware development project. Each workspace provides guidelines, tools, functionalities and information relevant to specific role(s) within the courseware development process. According to the profiles of the user (which role he/she plays) and to the courseware development methodology adopted, the OAD configures itself providing a methodological guide to the user and allowing a task oriented activation of the system facilities. Phases, activities and tasks of the courseware development methodology are used as meaningful handlers to activate the related supporting tools: each authoring task is mapped in a proper tool activation to allow the author to carry out the task itself.

3.1 OSCAR Hardware/Software Platform

The OSCAR hardware/software platform is based on a client/server model, which refers to industry standard systems. Such a platform allows the coexistence of the MS-DOS and UNIX worlds in order to combine the versatility of personal computers with the networking and transaction capabilities of workstations (OSCAR 1993b).

An Object Oriented Database Management System (OODBMS) runs on the server under UNIX to hold all the multimedia objects produced during the courseware development process. The client applications run under MS­WINDOWS on PCs connected to the host system by means of an Ethernet LAN.

The selected OODBMS is available for both a standalone and a networked environment. On the latter configuration, it offers a multi-client/multi-server architecture and features distribution and transaction management for concurrent access to objects among members of a work group.

4 Co-authoring Services

OSCAR provides four classes of co-authoring services (OSCAR 1992c): authoring, collaboration, coordination, co-decision and re-use. All of them are incorporated within the Co-authoring Space which ensures integration between different services and uniform access to the lower infrastructure OSCAR services.

The Authoring services are realised through a set of software tools supporting design and production of multimedia courseware. In particular, OSCAR provides tools to specify the instructional elements of a courseware, tools helping authors in laying out effective instructional strategies (Van Marcke 1990) and tools to implement the actual instructional materials. In order to exploit existing tools and make the OSCAR system more flexible, external authoring tools, whether available on the market or coming out of other research projects, can be incorporated in OSCAR and tailored to specific training applications. The authoring tools running in OSCAR are made interoperable between themselves: for instance, the elements of an instructional strategy defined with the courseware design tool can be semantically linked to the materials developed with the courseware production tool.

OSCAR: A System for Collaborative Distributed Authoring 287

The Coordination services support the organisation and management of courseware projects (i.e., courseware project management, quality assurance, work flow control, project documentation management) through asynchronous cooperation by various actors. The Coordination services allow a courseware authoring team to decide on the authoring methodology which best fits their needs and then to configure the way the OSCAR System works according to the chosen approach. Once the approach is defined and selected, the Coordination services support adherence to the chosen methodology.

The Collaboration services support group communication and sharing of information. Such functionalities are aimed at supporting synchronous collaboration between people working in a given activity of the courseware development process. For instance, a facility provided by the Collaboration Services is a tool that allows two authors to cooperate in real time in co-editing courseware specifications or in working on the same courseware material, each one bringing different contributions and skills.

The Co-decision services support remote group decision making among people that are not together either in the same place or at the same time in a courseware development project, by providing means to facilitate and organise both formal and informal asynchronous discussions in a work group via mechanisms that support their definition and flow control.

The Re-use services support the reusability of existing training materials and/or half-fabricates through mechanisms of retrievability and adaptability of courseware components and/or half-fabricates of the courseware development process. The Re­use services refer to three classes of tools: CIS (Common Information Space) Browser, Conversion Tool Kit and a Module to support Common Views of Collaborators during Reuse. The first is meant to facilitate retrievability, while the second and the third tackle problems related to adaptability.

5 Coordination Services

Coordination of a process (OSCAR 1993, Charette 1986) made up by many different concurrent activities means management of the interdependencies that exist between the many activities performed by multiple actors dealing with information objects that are in most of the cases common to two or more activities (OSCAR 1993a).

In order to analyse interdependencies in joint courseware development, a model of collaborative authoring has been first defined using the AMIGO Activity Model (Pantoke-Babatz 1989) and then four modules have been identified to group all of the functions needed to coordinate a courseware authoring project (OSCAR 1993):

- The Authoring Methodology Module This module provides the person that is in charge of it with functionalities to decide what is the authoring methodology to adopt for the project, to edit and customise it with respect to: the characteristics of the project itself, the steps and the persons involved in each step. In this module, a Methodology is defined as a set of steps that have to be carried out in order to complete the project. The steps have been hierarchically divided into two classes: Phases and Activities, where a Phase is made of Activities. Within each Activity one or more person playing a

288 A. Ulloa, A. De Girolamo, S. Delaney

Role in it, produces one or more Activity Deliverable (an Activity Deliverable is a structured project output foreseen by the methodology and that is produced at the end of a given step).

- The Context Sensitive Module The basic function performed by the C.S.M. (OSCAR 1992c e 1993a) is to read the data that is entered when choosing and configuring the methodology. This capability mainly allows the C.S.M.: to control and coordinate the workflow of the project and to deduce the state of the project with respect to activities completed, Activity Deliverables produced, people involved and so on. This module is mainly seen as a coordinator that on the basis of the state of the project sends warnings to the persons that can start their work on the project. The project responsible decides before starting the project whether the coordination will be strong, weak or none. When a user logs in the system and chooses to work on a given project, the C.S.M. is automatically activated and it will configure his/her desktop with respect to the things he/she must do according to his/her role. This means that the author's desktop will show the activities that have been assigned to himlher (i.e. some will be enabled and some others will not depending on the state of the project and on the role the author covers). A given author is allowed, in the context of a given project, to work only on the activities that have been assigned to himlher during the methodology setup phase. Some activities will probably have time constraints in the sense that they cannot be started until another activity has issued its output. In this case three different scenarios could be presented on the author desktop with respect to three possible levels of coordination.

- The Quality Standards Module The Quality Standards Module within the OSCAR system provides as services a reference library where authors can browse information concerning quality standards for courseware authoring; and functionalities by which a set of quality standards can be associated with a courseware authoring project

- The Structured Documentation Library The Structured Document Library is a reference library through which courseware authors will have access to structured documents produced by their own and by other authoring projects. The structured documents produced by a project will describe both management and design decisions taken during the course of the project, as well as the rationale behind those decisions. They will also describe the outputs from each phase and activity of the project.

6 Conclusion

OSCAR is an authoring system being developed within the CEC DELTA programme. The main features of the project are:

- Models of collaborative and distributed development of courseware which, focusing on communication paths within the work group involved in courseware

OSCAR: A System for Collaborative Distributed Authoring 289

project, permit the specification of regulations and procedures for the coordination of the individual tasks.

- Open system architecture referring to de-jure or de-facto standards and able to incorporate external tools.

- Multimedia communication facilities supporting the distribution of the authoring process and applications such as electronic mail, file transfer tailored to the distributed and collaborative authoring.

- Data base of training materials and half-fabricates of the courseware development process, acting as a common repository where information shared by various applications are consistently managed, concurrently made available to and updated by different actors.

- Authoring tools covering the phases of design and production of the courseware life-cycle.

- Software tools, supplying high level services supporting collaboration, coordination, co-decision and re-use in courseware projects.

- Methodology and practical guidelines for joint courseware development at European level.

On the basis of the results achieved, the OSCAR Project will bring significant benefits to the European training industry through an increase in the market competitiveness of the courseware producers, in terms of better economies of scale, improvements in courseware quality and effectiveness.

In particular, the contribution of OSCAR is towards an improvement in the effectiveness and flexibility of authoring systems and, in general, of IT&T supports to the courseware producers

Finally, OSCAR is one of the key projects of the current phase of DELTA, which is expected to bring contributions to the Common Training Architecture (CTA) definition. CTA is the DELTA forum through which recommendations of standards will be communicated to the IT and telecommunication industry and its relevant standard bodies (CTA 1992).

References

Charette R.N. (1986) Software Engineering Environments. Intertexts Publishing Inc. CTA Consortium: Education and Training Enterprise Models (1992). Delta Deliverable

D202302 Dark S.L. (1986) The Implications of Quality Systems in the Development of

Courseware For Computer Based Training. MSc Thesis (unpublished) Lancaster University

OSCAR Consortium (1992a) Refined model for collaborative authoring and reuse. Delta Deliverable D2006 01

OSCAR Consortium (1992b) Training needs and authoring requirements of aerospace industries. Delta Deliverable D2006 02

OSCAR Consortium (1992c) Preliminary logical architecture. Delta Deliverable D2006 03

OSCAR Consortium (1993a) Final overall architecture specifications. Delta Deliverable D200604

OSCAR Consortium (1993b) Technological choices for prototyping. Delta Deliverable D2006 05: 1993

290 A. Ulloa, A. De Girolamo, S. Delaney

OSCAR Consortium (1993c) Specifications of Authoring Space. Delta Deliverable D2006 06: 1993

OSCAR Consortium: Specifications of the Coordination Services. (1993) Delta Deliverable D2006 08

Pankoke-Babatz, U., Danielson, T., Patel, A., Pays, P.A., Prinz, W., Speth, R. (1989a) Computer Based Group Communication, the Amigo Activity model: Ellis Horwood

Russell D.M., Moran T.P., Jordan D.S. (1989b) The Instructional-Design Enviroment. In: Psotka, J., Massey, L.D., Mutter, S.A. (eds.) Intelligent Tutoring Systems: Lessons Learned. Hillsdale, New Jersey: Lawrence Eribaum Associates

Van Marcke K. (1990) A Generic Tutoring Environment, in: proceedings of the 9th European Conference on Artificial Intelligence (ECAI), Stockholm, Sweden

List of Contributors

Gary Alexander 65 Javier Alvarez 258 Beatriz Barros 11 Freimut Bodendorf 97 Andre Boder 202 Peter Busch 11, 270 Stefano A. Cerri 1,119,181 Benita Compostela 11 Jose Cuena 142 Thore Danielsen 21 Stephen Delaney 282 Alain C. Derycke 236 Henk Ellerman 78 Ana Garcia-Serrano 142 Christine Gardiol 202 Antonio De Girolamo 282 Finn Groenbaek 270 Dominique Guin 170 Willibrord Huisman 78 Anthony R.Kaye IX Paul Lefrere 65 Christian Lemaitre 126 Cristina Loyo 126

J. Arnaldo Martins 109 Steve Matheson 65 David McConnell 46 Encarna Pastor 258 Morten Flate Paulsen 11, 33 Mirabelle Quesnel 11 Mette Ringsted 270 T. Rodden 219 Gonzalo Sanchez 258 Victor German Sanchez 126 Ad Schellekens 78 Ralph Seitz 97 Hanne Shapiro 270 Cristina Sim6n 60 Carla Simone 156 I. Sommerville 219 Antonio Sousa Pereira 109 Michael B. Twidale 219 Antonio Ulloa 282 J.A. Veiga Pires 109 M. Felisa Verdejo 1,142 Claude Vi6ville 236 Peter Zorkoczy 210

Subject Index

abduction 181 abstraction 181 action 142; 170; 181; 210 actors

administrators 156 designers 11; 33 program planners 11; 33 teachers 11

agents 97-108; 142; 170-180; 181; 210 computer, artificial 126; 170 heterogeneous 126 human 126; 170 loosely coupled 126 software 97

analysis 60-64 applications to learning 119-125 applications in

academic consulting 142 aerospace 282 course production 270 diagnosis 181 document production 142 management 46 mathematics 170

linear algebra 78-96 medicine 21; 109-116; 202 product information 270 renewable energy technology 65-77 SME in the graphical branch 270 software engineering 219 telecommunications 210 university 78; 97

apprenticeship 33 architecture

cognitive 142 training system 210

articulation conception-situations 170

artificial intelligence 142 distributed 126-141

assignment group 78 transcript bases 33 tutor corrected 78

assignments 78-96 attitudes 98 audio conferencing 236-257 authoring, co-authoring 97-108; 142;

270-280; 282-290 behavioural level 170-180 belief 170 brainstorming 33; 219-235; 258 broadband communication 258-269 bulletin board 97-108

CD-ROM 65-77 CDTDL (see: collaborative dialogue

technologies in distance learning) CMC (see: computer mediated

communication) co-decision 282-290 cognitive

architecture 142 maps/design 219-235

collaboration 170; 282-290 tools 109

collaborative design 21-32 dialogue technologies

in distance learning 119; 181 dialogues 156-169; 181-201; 219 learning 65-77; 97-108; 219-235

at a distance 65; 98 work 109-116 working 60-64

commitment 156 common lexicon

definition of 119 communication

broadband 258-269 computer mediated 21-32; 33-45; 78 informal 258-269 protocol of 170-180 relevant 126 repetitive, standardized 156 structured 258 styles 60 thread 65

computer conferencing 33-45; 65-77; 97-108; 202-209 asynchronous 109; 236 multimedia 236 real time 109; 236

computer mediated communication 11-20; 65-77; 78; 181; 219-235 dialogues 181

computer supported collaborative learning 119-125; 181; 236

obstacles to 119; 210 cooperative work 21; 119-125; 126-141; 142-155; 156-169; 181-201; 219-235; 236; 258-269

computer teleconferencing (see: computer conferencing)

conceptions 170-180 conceptual field 170 conferencing, audio 236 consensus formation 142

294 Subject Index

constraints 11 context

cognitive 156 organizational 156

conversation 156 for action 156 for possibility 156 space 236-257

cooperation in small groups 78 loose form of 156 network 126-141

cooperative distributed problem solving 126 learning 97-108 open systems 126 technology 21-32 work 109-116

coordination 282-290 in distance learning 142-155 modeling 142 of action 210

copresence 156 course

delivery 270 development 270

joint 282-290 production

distributed 270 modulized 270

critical factors 60-64 CSCL (see: computer supported

collaborative learning) CSCW (see: computer supported

cooperative work) OAI (see: artificial intelligence, distributed) debate 33; 202

rhetoric of 202 decision-making 210 design 60-64

cognitive 219 collaborative 21-32 iterative 219·235 organizational 97 pedagogical 97 technical 97

diagnosis 170; 181-201 dialogue 126-141; 142; 170-180

collaborative 156-169 management 156; 181 types 181

didactic functions 78-97 discourse management 156 distance

education 33-45; 258 learning 21-32; 65-77; 97-108; 156; 210-218; 236-257

distributed classroom 210 open processing 258

echoes 78 editors

formulae 78-96

multimedia 109 education 210-218

distance (see: distance education) electronic mail 97-108 empirical studies 97-108 environment

forCMC 11 interactive 202 interactive learning 170-180 learning 78-96 production 270-281 study 78-96 tutor's 78

evaluation 60-64 Carroll theory 97 Fishbein theory 97 formative 219-235 of the prototype 78

evolutionary 170; 181; 202; 219 development 78-96

experiences 60-64 participants' 46-59

experiments in learning 119-125 expert system 126 explanation

in coordination 142 in learning 142

facilitation intellectual 11 organizational 11 social 11 technique 11

facilitator 126-141; 142 flexible

learning 181 material 270-281

games 33 aoss 97-108 gender differences in groups 46-59 generalization 181 goal 126; 142 group

dynamics 46-59 learning 46

group support 142 active 142 negotiation level 142 object-level 142 passive 142

groupware 21-32; 236-257 help systems 181 heterogeneity 126 heuristic 202-209 ideas, refinement of 219 implementation 60-64 integration 21-32 interaction 170; 202-209

didactic 170-180 informal 258

interactive learning environment 170-180 interface

four windows 202 graphic 109; 219; 236; 258

interoperability semantic 126 syntactic 126

interviews 33; 97 ISDN 109-116; 236

broadband 258 compression algorithm 109 narrow band 109

joint courseware development 282-290 knowledge

acquisition 119; 181-201; 202-209 as professional experience 202 based methodology 142-155 based systems 126; 142-155 bases 126; 202-209 communication 181 level 142; 170-180 meta 170-180 selection 202

language 46-59 differences in group work 46

learner behaviour 170 learning

as a side effect 119; 181 contract 33 environment 78; 97-108 groups 46-59 machine 170 open 202 technology supported 210-218

libraries 97-108 online 97

local government 21-32 material

flexible 270 printed 78

meeting room 65 message 170-180

analysis MOSCA protocol 170

automatic classification 142 content 156 semistructured 156-169 structured 126; 142

metacognition 170 metaphor 78; 258 methods and methodologies

gradualist development 219 inverted dialogues 181 iterative design 181; 202; 219 Kanban simulation 236 knowledge based 142 natural laboratory 181 participatory design 219 PENTADE 170 rapid prototyping 219 refinement 219 reification of interactions 202

model construction 181-201 domain specific, generic 181 epistemic 170 refinement 181

Subject Index 295

transposition 170 models 119

classification 119 conceptual 78; 142 conversation 119 industrial 78; 210 of diagnosis 119 of interaction 170 shared 170 user 156-169; 181-201

modulized course production 270-281 monitoring 78-96; 97; 156; 181 motivation 60 multiad 170 multimedia 236; 258-269; 270

editors 109 library 65

multiparty simulation 236-257 negotiation 156 network

agent 126 application 126 communication 126 computer 126 cooperation 126-141 expert system's 126 facilitator's 126 local area 236

online databases 11; 33 environment 46 journals 11; 33 work 46-59

open distributed processing 258-269 open learning 202-209 open systems 126-141; 142

architecture 282-290 cooperative 126-141 interconnection

layers 126 standards 126

OSI (see: open systems interconnection) pane

document 65 outline 65

pedagogical design 97 framework 11-20 scenarios 109; 210 styles 11-20 techniques 11-20; 33-45

DELPHI 33 many-to-many 97 Nominal Group 33 one-alone 97 one-to-many 97 one-to-one 97 Project Group 33

phase space 170 pragmatics of discourse

speech act 126-141 turn taking 46; 170; 181

professional development 202-209

296 Subject Index

prototypes and systems AMIGO 21-32; 236; 282 CHAOS 156 Designer's Note Pad 219 EDUBA258 ISLE 65 NOBILE 181 STUDIENET 78 Tele-Amphi 236

R&D projects CO-LEARN 236; 258 COSYS 270-281 CTA210 DISCOURSE 181 ECOLE 210 EDUBA258 JITOL202 NAT*LAB 181 OSCAR 181; 282 Portable PC for students 97 Tele-COO78 Tele-Educatie 78 TeleCommunity 109

rapid prototyping 219-235 rationality. principle of 170 re-use 282-290 real time

multimedia conferencing 236 teleassistance 109; 236 teleteaching 109; 236

reflection 170 reification 202 representation. changes of 170 resources

external 11 forCMC 11 internal 11

role 11; 21; 142; 156; 170-180; 181; 282 classification 170 co-authoring 270 MOSCA protocol 170 play 33; 97

scenario 119; 142; 181; 210-218 construction 210 coordination oriented 210 learning oriented 210 utilization 210

sector specific solutions 21 selection

of relevant knowledge 202 shared

initiative 170 space 236-257 window 258

sharing

of conceptions 170 of information 156

simulation 33; 181 multiparty 236

standardization and scenarios 210 standards

emerging 119 for open systems 126 ISO-OSI126

study environment 78 guidance 78-96 network mediated support 78

tasks generic 142 global 142 group 142

teaching style 170-180 teachware 97-108 teamwork 236-257 telecommunications 210-218 teleconferencing 109 teleconsultation 109 telemedicine 21-32 telephone 78-96 telepresence 65-77 templates

empaty65 theory construction 119-125

bottom up approach 142 thread 65 time 46-59

response 109 transmission 109 use of in online work 46

training 109; 210-218; 270 scenarios for 210

training system architecture 210-218 tutor support 78 usability 60 user models 156-169; 181-201 viewpoint

content centered 119 content free 119 methodological 119

virtual space 156 team 210

work collaborative 60-64 online 46-59 team 98 tele 98

workspaces 282-290 WYSIWYS 236-257; 258

NATO ASI Series F

Including Special Programmes on Sensory Systems for Robotic Control (ROB) and on Advanced Educational Technology (AET)

Vol. 94: Logic and Algebra of Specification. Edited by F. L. Bauer. W. Brauer. and H. Schwichtenberg. VII. 442 pages. 1993.

Vol. 95: Comprehensive Systems Design: A New Educational Technology. Edited by C. M. Reigeluth. B. H. Banathy. and J. R. Olson. IX. 437 pages. 1993. (AET)

Vol. 96: New Directions in Educational Technology. Edited by E. Scanlon and T. O·Shea. VIII. 251 pages. 1992. (AE?)

Vol. 97: Advanced Models of Cognition for Medical Training and Practice. Edited by D. A. Evans and V. L. Patel. XI. 372 pages. 1992. (AE7)

Vol. 98: Medical Images: Formation. Handling and Evaluation. Edited by A. E. Todd-Pokropekand M. A. Viergever. IX. 700 pages. 1992.

Vol. 99: Multisensor Fusion for Computer Vision. Edited by J. K. Aggarwal. XI. 456 pages. 1993. (ROB)

Vol. 100: Communication from an Artificial Intelligence Perspective. Theoretical and Applied Issues. Edited by A. Ortony. J. Slack and O. Stock. XII. 260 pages. 1992.

Vol. 101: Recent Developments in Decision Support Systems. Edited by C. W. Holsapple and A. B. Whinston. XI. 618 pages. 1993.

Vol. 102: Robots and Biological Systems: Towards a New Bionics? Edited by P. Dario. G. Sandini and P. Aebischer. XII. 786 pages. 1993.

Vol. 103: Parallel Computing on Distributed Memory Multiprocessors. Edited by F. OzgOner and F. Er~1. VIII. 332 pages. 1993.

Vol. 104: Instructional Models in Computer-Based Learning Environments. Edited by S. Dijkstra. H. P. M. Krammer and J. J. G. van Merrienboer. X. 510 pages. 1993. (AET)

Vol. 105: Designing Environments for Constructive Learning. Edited by T. M. Duffy. J. Lowyck and D. H. Jonassen. VIII. 374 pages. 1993. (AE7)

Vol. 106: Software for Parallel Computation. Edited by J. S. Kowalik and L. Grandinetti. IX. 363 pages. 1993.

Vol. 107: Advanced Educational Technologies for Mathematics and Science. Edited by D. L. Ferguson. XII. 749 pages. 1993. (AET)

Vol. 108: Concurrent Engineering: Tools and Technologies for Mechanical System Design. Edited by E. J. Haug. XIII. 998 pages. 1993.

Vol. 109: Advanced Educational Technology in Technology Education. Edited by A. Gordon. M. Hacker and M. de Vries. VIII. 253 pages. 1993. (AET)

Vol. 110: Verification and Validation of Complex Systems: Human Factors Issues. Edited by J. A. Wise. V. D. Hopkin and P. Stager. XIII. 704 pages. 1993.

Vol. 111: Cognitive Models and Intelligent Environments for Learning Programming. Edited by E. Lemut. B. du Boulay and G. Dettori. VIII. 305 pages. 1993. (AET)

Vol. 112: Item Banking: Interactive Testing and Self-Assessment. Edited by D. A. Leclercq and J. E. Bruno. VIII. 261 pages. 1993. (AET)

Vol. 113: Interactive Learning Technology for the Deaf. Edited by B. A. G. Elsendoorn and F. Coninx. XIII. 285 pages. 1993. (AET)

Vol. 114: Intelligent Systems: Safety. Reliability and Maintainability Issues. Edited by O. Kaynak. G. Honderd and E. Grant. XI. 340 pages. 1993.

Vol. 115: Learning Electricity and Electronics with Advanced Educational Technology. Edited by M. Caillot. VII. 329 pages. 1993. (AET)

NATO ASI Series F

Including Special Programmes on Sensory Systems for Robotic Control (ROB) and on Advanced Educational Technology (AET)

Vol. 116: Control Technology in Elementary Education. Edited by B. Denis. IX, 311 pages. 1993 (AET)

Vol. 118: Program Design Calculi. Edited by M. Broy. VIII, 409 pages. 1993.

Vol. 119: Automating Instructional Design, Development, and Delivery. Edited by. R. D. Tennyson. VIII, 266 pages. 1994 (AET)

Vol. 120: Reliability and Safety Assessment of Dynamic Process Systems. Edited by T. Aldemir, N. o. Siu, A. Mosleh, P. C. Cacciabue and B. G. GOktepe. X, 242 pages. 1994.

Vol. 121: Learning from Computers: Mathematics Education and Technology. Edited by C. Keitel and K. Ruthven. XIII, 332 pages. 1993. (AET)

Vol. 122: Simulation-Based Experiential Learning. Edited by D. M. Towne, T. deJong and H. Spada. XIV, 274 pages. 1993. (AET)

Vol. 123: User-Centred Requirements for Software Engineering Environments. Edited by D. J. Gilmore, R. L. Winder and F. Detienne. VII, 377 pages. 1994.

Vol. 124: Fundamentals in Handwriting Recognition. Edited by S. Impedovo. IX, 496 pages. 1994.

Vol. 125: Student Modelling: The Key to Individualized Knowledge-Based Instruction. Edited by J. E. Greer and G. I. McCalla. X, 383 pages. 1994. (AET)

Vol. 126: Shape in Picture. Mathematical Description of Shape in Grey-level Images. Edited by Y.-L. 0, A. Toet, D. Foster, H. J. A. M. Heijmans and P. Meer. XI, 676 pages. 1994.

Vol. 127: Real Time Computing. Edited by W. A. Halang and A. D. Stoyenko. XXII, 762 pages. 1994.

Vol. 129: Human-Machine Communication for Educational Systems Design. Edited by M. D. Brouwer­Janse and T. L. Harrington. X, 342 pages. 1994.

Vol. 130: Advances in Object-Griented Database Systems. Edited by A. Dogac, T. 6zsu, A. Biliris and T. Sellis. XI, 515 pages. 1994.

Vol. 131: Constraint Programming. Edited by B. Mayoh, E. Tyugu and J. Penjam. VII, 452 pages. 1994.

Vol. 132: Mathematical Modelling Courses for Engineering Education. Edited by Y. Ersoy and A. O. Moscardini. X, 246 pages. 1994.

Vol. 133: Collaborative Dialogue Technologies in Distance Learning. Edited by M. F. Verdejo and S. A. Cerri. XIV, 296 pages. 1994.

Vol. 134: Computer Integrated Production Systems and Organizations. The Human-Centred Approach. Edited by F. Schmid, S. Evans, A. W. S. Ainger and R. J. Grieve. X, 347 pages. 1994.