Upload
others
View
2
Download
0
Embed Size (px)
Citation preview
A STUDY OF THE COMBINED USE OF
CONCEPTUAL MODELS
Mohammad Ali Jabbari Sabegh
MBA, BSc
Submitted in fulfilment of the requirements for the degree of
Doctor of Philosophy
School of Management
QUT Business School
Queensland University of Technology
2017
A Study of The Combined Use of Conceptual Models i
Supervisory Panel
Principal Supervisor
Professor Jan Recker, PhD
School of Management
QUT Business School
Queensland University of Technology
Associate Supervisor
Professor Peter Green, PhD
School of Accountancy
QUT Business School
Queensland University of Technology
ii
Keywords
Conceptual Models, Combined Use, Model Combinations, Ontology, Completeness, Overlap,
Ontological Analysis, Context, Faithful Use
A Study of The Combined Use of Conceptual Models iii
Abstract
Conceptual models are graphical representations of the relevant features of a system
domain. They play a significant role in the early detection and correction of information
systems development errors and help analysts, developers, and end users to better understand
the domain. However, despite conceptual modelling being an active research area in
information systems, much understanding of conceptual modelling practice remains
outstanding. This thesis addresses such aspect: while practitioners typically use multiple, and
different, types of models in system analysis and design tasks, no intense scrutiny of how and
why they use model combinations has been undertaken.
The objective of this research is to provide an in-depth investigation into the use of
model combinations during information systems analysis and design. A multi-method
research design was used to achieve this objective. The contribution of this thesis is built on
three streams of studies: a systematic literature review, exploratory interviews, and a lab
experiment for theory testing.
First, a systematic literature review was conducted to analyse the extent to which
research over the past two decades has identified research opportunities and provided answers
to Wand and Weber’s (2002) question regarding “How can we model the world to better
facilitate our developing, implementing, using, and maintaining more valuable information
systems?”.
The next two studies focussed on the theory of faithful use of conceptual model
combinations. This theory was developed to predict and explain the use of multiple models in
combination; however, this theory has not yet been tested. In the second study, semi-
structured interviews were conducted to explore and determine the circumstances that lead
information system (IS) professionals to use multiple models.
In the third study, lab experiments were designed and executed to test the hypotheses
derived from the propositions of the theory of faithful use of conceptual model combinations.
The results led to a refined and improved theory that provides explanation and in-depth
understanding of how and why practitioners use model combinations.
iv
Table of Contents
Supervisory Panel ................................................................................................................. i
Keywords ............................................................................................................................ ii
Abstract .............................................................................................................................. iii
Table of Contents ................................................................................................................ iv
List of Figures ................................................................................................................... vii
List of Tables .................................................................................................................... viii
List of Abbreviations ........................................................................................................... x
Statement of Original Authorship ........................................................................................ xi
Acknowledgements ............................................................................................................ xii
Part A: Preface ..................................................................................................................... 1
Chapter 1: Introduction ................................................................................... 1
1.1 Introduction ............................................................................................................... 1
1.2 The Problem Statement .............................................................................................. 3
1.3 Research Questions .................................................................................................... 5
1.4 Scope of Study ........................................................................................................... 7
1.5 Thesis Structure ......................................................................................................... 8 1.5.1 Part A .............................................................................................................. 8 1.5.2 Part B ............................................................................................................... 8 1.5.3 Part C ............................................................................................................. 10
Chapter 2: Research Background .................................................................. 11
2.1 Conceptual Modelling in Information Systems ......................................................... 11
2.2 Research on Single Grammars or Models ................................................................. 13 2.2.1 Conceptual Model Design .............................................................................. 13 2.2.2 Conceptual Model Interpretation .................................................................... 15
2.3 Research on Multiple Grammars or Models .............................................................. 16
2.4 The Theory of Faithful Use of Conceptual Model Combinations .............................. 17
2.5 Conclusion ............................................................................................................... 20
Chapter 3: Research Methods ........................................................................ 22
3.1 Research Method Choices ........................................................................................ 22 3.1.1 Systematic Literature Review ......................................................................... 22 3.1.2 Semi-Structured Interviews ............................................................................ 23 3.1.3 Experiment .................................................................................................... 24
3.2 Reliability and Validity ............................................................................................ 24 3.2.1 Reliability and validity of systematic literature review .................................... 24 3.2.2 Reliability and Validity of Semi-Structured Interviews and the Experiment .... 25
A Study of The Combined Use of Conceptual Models v
Part B: Studies ................................................................................................................... 29
Chapter 4: Conceptual Modelling Research in Information Systems: What We
Know Now and What We Still Do Not Know ...................................................... 29
Abstract ............................................................................................................................. 29
4.1 Introduction ............................................................................................................. 30
4.2 Background ............................................................................................................. 31
4.3 Literature Review Procedures................................................................................... 35
4.4 Analysis of the Literature ......................................................................................... 36 4.4.1 Conceptual Modelling Grammars ................................................................... 38 4.4.2 Conceptual Modelling Methods ...................................................................... 42 4.4.3 Conceptual Modelling Scripts ........................................................................ 45 4.4.4 Conceptual Modelling Context ....................................................................... 48 4.4.5 Identified articles that did not match the research framework by Wand and Weber
(2002) altogether ............................................................................................ 53
4.5 Guiding the next wave of conceptual modelling research: A new framework ............ 54
4.6 Conclusion ............................................................................................................... 59
Chapter 5: Combined Use of Conceptual Models in Practice: An Exploratory Study
…………………………………………………………………………………….61
Abstract ............................................................................................................................. 61
5.1 Introduction ............................................................................................................. 62
5.2 Background ............................................................................................................. 63 5.2.1 Ontological Analysis as a Theoretical Foundation of Conceptual Modelling ... 64 5.2.2 Empirical Studies of Conceptual Modelling Practice ...................................... 66
5.3 Research Method ..................................................................................................... 67 5.3.1 Design ........................................................................................................... 67 5.3.2 Participants and Procedures ............................................................................ 69 5.3.3 Data Analysis ................................................................................................. 70
5.4 Findings ................................................................................................................... 72 5.4.1 Why practitioners use multiple types of models .............................................. 73 5.4.2 How practitioners decide on which models to use ........................................... 78
5.5 Discussion ............................................................................................................... 84 5.5.1 Implications for ontological research on conceptual modelling ....................... 88 5.5.2 Implications for empirical research on conceptual modelling .......................... 90 5.5.3 Implications for the practice of conceptual modelling ..................................... 90 5.5.4 Limitations ..................................................................................................... 91
5.6 Conclusion ............................................................................................................... 92
Chapter 6: An Experimental Study to Test and Extend the Theory of Faithful Use of
Conceptual Model Combinations......................................................................... 94
6.1 Introduction ............................................................................................................. 95
6.2 Use of Multiple conceptual models .......................................................................... 96
6.3 The Theory of Faithful Use of Conceptual Model Combinations .............................. 97
vi
6.4 Research Method ..................................................................................................... 99 6.4.1 Preliminaries .................................................................................................. 99 6.4.2 Experimental Design ...................................................................................... 99 6.4.3 Materials ...................................................................................................... 101 6.4.4 Establishing Levels of Combined Ontological Completeness and Ontological
Overlap ........................................................................................................ 103 6.4.5 Measures ...................................................................................................... 106 6.4.6 Covariates .................................................................................................... 108 6.4.7 Participants .................................................................................................. 109 6.4.8 Procedures ................................................................................................... 109
6.5 Hypothesis Development ....................................................................................... 112 6.5.1 Selection of Model Combination .................................................................. 112 6.5.2 Development of Domain Understanding ....................................................... 114 6.5.3 Perceived usefulness of model combinations ................................................ 116
6.6 Data Analysis ......................................................................................................... 117 6.6.1 Preliminary Tests ......................................................................................... 118 6.6.2 Hypothesis Testing: Model Selection ............................................................. 118 6.6.2 Hypothesis Testing: Domain Understanding ................................................. 124
6.7 Discussion ............................................................................................................. 134 6.7.1 Summary of Findings ................................................................................... 134 6.7.2 Theoretical Implications ............................................................................... 138 6.7.3 Practical Implications ................................................................................... 139 6.7.4 Limitations ................................................................................................... 140
6.8 Conclusion ............................................................................................................. 141
Part C: Retrospective ....................................................................................................... 142
Chapter 7: Discussion ................................................................................... 142
7.1 Synthesis of Findings from Studies ........................................................................ 142
7.2 Implications ........................................................................................................... 147 7.2.1 Implications for Research ............................................................................. 147 7.2.2 Implications for Practice............................................................................... 149
Chapter 8: Conclusions ................................................................................ 151
8.1 Contributions ......................................................................................................... 151 8.1.1 For research ................................................................................................. 151 8.1.2 For practice .................................................................................................. 152
8.2 Limitations............................................................................................................. 153
8.3 Future Research ..................................................................................................... 155
8.4 Conclusion ............................................................................................................. 156
Bibliography ....................................................................................................... 157
Appendices .......................................................................................................... 176
A Study of The Combined Use of Conceptual Models vii
List of Figures
Figure 1.1 Examples of different models developed using various conceptual modelling
grammars .................................................................................................... 2
Figure 1.2 Outline of thesis structure and research objectives ................................... 9
Figure 2.1 Premises of the Theory of Faithful Use of Conceptual Model Combinations
(Recker, 2014) .......................................................................................... 18
Figure 4.1 Examples of different scripts developed using four conceptual modelling
grammars .................................................................................................. 32
Figure 4. 2 Framework for Research on Conceptual Modelling (Wand & Weber, 2002,
p. 364) ...................................................................................................... 33
Figure 4.3 A New Research Framework to support future conceptual modelling research
................................................................................................................. 57 Figure 5.1 Interview sampling and theoretical saturation ........................................ 71
Figure 5.2 Illustration of coding procedure ............................................................. 72
Figure 5.3 Example of diagram use to represent different aspects of a system ......... 76
Figure 5.4 The role of different factors that determine the use of multiple models... 86
Figure 5.5 The role of different factors that determine the selection of different types of
models ...................................................................................................... 87
Figure 6.1 Illustration of Ontological Completeness and Overlap of Two Conceptual
Models (Recker, 2014) .............................................................................. 98
Figure 6.2 Four Conceptual Models for the Library Management System Case
(Whiteley, 2013, pp. 228-263) ................................................................ 102
Figure 6.3 Overview of experimental procedure ................................................... 110
Figure 6.4 Estimated Marginal Means of Performance on Domain Understanding 132
Figure 6.5 Estimated Marginal Means of Perceived Usefulness ............................ 134
viii
List of Tables
Table 1.1 Papers included in this dissertation ............................................................ 5
Table 4. 1 Research Opportunities in Conceptual Modelling (Wand & Weber, 2002, p.
373) .......................................................................................................... 34
Table 4.2 Reviewed papers ..................................................................................... 36
Table 4. 3 Research on conceptual modelling based on focus of the study .............. 37
Table 4. 4 Number of papers per type of evidence .................................................. 37
Table 4. 5 Number of papers per type of model ...................................................... 38
Table 4. 6 Papers on conceptual modelling grammars and type of evidence ............ 39
Table 4. 7 Papers on conceptual modelling method and type of evidence ................ 43
Table 4. 8 Research on conceptual modelling scripts and type of evidence ............. 45
Table 4. 9 Papers in conceptual modelling context and type of evidence ................. 49
Table 5.1 Examples of studies on ontological completeness of modelling grammars66
Table 5.2 Demographic information ....................................................................... 70
Table 5.3 Examples of open codes collected ........................................................... 71
Table 5.4 Most widely used models ........................................................................ 73
Table 6.1 Ontological evaluation of conceptual model constructs - Library Management
System and High Peak Bicycles .............................................................. 105
Table 6.2 Overlap Analysis................................................................................... 105
Table 6.3 MOC and MOO for combinations of three and four models - High-Peak
Bicycles Case.......................................................................................... 106
Table 6.4 Measurement instrument for dependent variables .................................. 107
Table 6.5 Distribution of participants in each group .............................................. 111
Table 6.6 Reliability and validity analysis of scaled measures............................... 118
Table 6.7 Test of homogeneity of variances for dependent measures .................... 118
Table 6.8 Frequency of participants and the number of models in combination ..... 119
Table 6.9 Frequency and descriptive statistics for second model selection ............ 119
Table 6.10 Mean and standard deviation of constructs for model selection............ 119
Table 6.11 Correlation matrix for selection of second model................................. 120
Table 6.12 Hierarchical logistic regression for selection of second model ............. 121
Table 6.13 Correlation matrix for the selection of the third model......................... 121
Table 6.14 Hierarchical logistic regression for the third model selection ............... 122
A Study of The Combined Use of Conceptual Models ix
Table 6.15 Number of models selected and number of models they used in combination
to answer the questions ........................................................................... 123
Table 6.16 Frequency of types of models .............................................................. 124
Table 6.17 ANOVA test to compare familiarity difference between groups .......... 124
Table 6.18 Descriptive statistics............................................................................ 125
Table 6.19 Construct correlations ......................................................................... 126
Table 6.20 Results for model comprehension ........................................................ 127
Table 6.21 Results for problem-solving ................................................................ 128
Table 6.22 Simple effect analysis for problem-solving performance ..................... 129
Table 6.23 Results for Cloze test .......................................................................... 130
Table 6.24 Results for perceived usefulness .......................................................... 133
Table 6.25 Summary of hypothesis testing ............................................................ 134
x
List of Abbreviations
AIS Association for Information Systems
BWW Bunge, Wand, and Weber
BPMN Business Process Modelling Notation
DFD Data Flow Diagram
EPC Event-driven Process Chains
ERD Entity Relationship Diagram
IS Information Systems
MOC Maximal Ontological Completeness
MOO Minimal Ontological Overlap
ORM Object Role Modelling
UML Unified Modelling Language
YAWL Yet Another Workflow Language
A Study of The Combined Use of Conceptual Models xi
Statement of Original Authorship
The work contained in this thesis has not been previously submitted to meet
requirements for an award at this or any other higher education institution. To the best of my
knowledge and belief, the thesis contains no material previously published or written by
another person except where due reference is made.
QUT Verified Signature
xii
Acknowledgements
There are several people that I am delighted to express my immense gratitude and
appreciation to for helping me in many ways throughout my PhD journey.
I am incredibly fortunate and truly proud to have Professor Jan Recker and Professor
Peter Green as my supervisors. I am a great admirer of your immense knowledge and
enthusiasm toward science. It inspires me to continue my career as an academic. I would also
like to thank you for your continued support, thoughtful feedback, criticism, guidance,
motivation, and patience during my research. You have been tremendous mentors to me. Your
advice about my research and my career has been priceless. I could not imagine having better
supervisors and mentors for my PhD study.
I would also like to thank other people who helped me during my research. Thanks to
Professor Michael Rosemann, Dr Ross Brown and Professor Marcello La Rosa for their
helpful comments on my research. I was privileged to have Professor Ron Weber as a panel
member of my PhD final seminar. I appreciate your detailed and constructive feedback. A
debt of gratitude also to Dr Artem Polyvyanyy, Professor Alistair Barros and Tim Dreesen for
their support in data collection. I also received helpful advice from participants at the ICIS
Doctoral Consortium and anonymous editors and reviewers of the studies included in this
thesis – thank you all.
I would also like to thank professional editor, Kylie Morris, who provided copyediting
and proofreading services, according to university-endorsed guidelines and the Australian
Standards for editing research theses.
During these years in Australia, I had the opportunity and pleasure to meet, interact, and
work with colleagues, both in the IS School and the Business School that had a positive
influence on my work and my life. There were also friends who made this journey more
enjoyable. I would like thank you for being in my life and for making my work-life easier and
extraordinary.
A big thank you to my family, who has been with me all along. Thanks to my late father
for everything he taught me. Thanks to my mother, my siblings, and the rest of the clan for
A Study of The Combined Use of Conceptual Models xiii
their unconditional love and their continued support in every possible aspect of my life. Your
encouragement and motivation are what has sustained me thus far.
Last, but not least, I would like to express my appreciation to my beautiful partner,
Farzaneh. Thank you for your unwavering love, admirable support, understanding, patience,
and encouragement.
Chapter 1:Introduction 1
Part A: Preface
Chapter 1: Introduction
1.1 INTRODUCTION
Conceptual models are simplified visual representations of the real world and are used
by humans to make sense of the world. The human mind conceptualises subjects by
constructing concepts and mapping them together as diagrams (Ausubel, 1968; Nesbit &
Adesope, 2006). Although the term “conceptual models” was developed approximately 40
years ago (P. P.-S. Chen, 1976), visual representations and mapping concepts as diagrams
have been used for learning and communication since at least the 13th century (Sowa, 2000).
Information systems analysts and designers often develop conceptual models to
represent the relevant features of a real-world domain (Gemino & Wand, 2004; Siau & Tan,
2005). Information systems are socio-technical systems developed based on an understanding
of a certain domain to support organisational activities in that domain (Bera, Burton-Jones, &
Wand, 2014; Wand & Weber, 1995). Conceptual models play a significant role in analysing
and promoting users’ knowledge of the requirements of the application domain for the
analysis and design of information systems (Gemino & Wand, 2004; Siau & Tan, 2005).
A conceptual model is developed using a modelling grammar1 (Wand & Weber, 2002),
which is a set of constructs and rules that demonstrate how to combine the constructs to
model real-world domains. There are many different types of conceptual modelling
grammars, which can be used for different purposes and organisational objectives (Fettke,
2009). For instance, process models are developed to describe organisational activities, and
business operations (Petri, 1962; Scheer & Nüttgens, 2000), while data models represent data
structures of systems (P. P.-S. Chen, 1976). Figure 1.1 represents only a few types of models
developed using different modelling grammars.
1 In practice, conceptual modelling grammars may also be referred to as modelling languages or techniques.
2
Check the order
Issue bill
Check quality
Ship the product
Not
in g
ood
qual
ity o
r dam
aged
BPMN
Check the order
V
Issue billXOR
Check the product
XOR
V
Start shipping
Not good quality or damaged
Good quality
EPC
ERD
Orders
-customerName
-dateCreated-dateShipped
-orderID
Customer
-address-customerName
User
-password-UserID
Administrator
-adminInfo-adminName
-email-contactdetails-shippinginfo
1..1
0..*
UML
Shop
Process order
Shop
Management information
Shop
Maintain rates
Customer
Discount rates
Product
Customer
Staff
2
1
3
D3
D2
D1
P1
T1
T2
T3
P2
P3
P4
DFD Petri Nets ORM YAWL
Figure 1.1 Examples of different models developed using various conceptual modelling grammars
The decision regarding which models to use requires an upfront decision about which
type of model to develop. It also requires substantial investment to develop those models.
Organisations spend a significant amount of time and money on conceptual modelling, such
as purchasing appropriate tools, licensing, training of employees, and developing
organisational conventions to create and implement the models (Recker, Indulska, Rosemann,
& Green, 2010; Rosemann, 2006). However, practitioners believe that the plethora of
modelling grammars and the availability of different methods to use these grammars are
major barriers to making investment decisions, and therefore increase the risk of investment
in appropriate modelling initiatives (Fettke, 2009). Therefore, knowledge regarding which
types of models to use is instrumental in facilitating well-founded decisions in organisations
and well-informed investments in conceptual modelling initiatives.
This knowledge about which types of models to use is also important given the high
failure rate of information systems projects. More than 60% of IT/IS projects fail to satisfy
their users or fail to meet deadlines (Palvia, Jacks, & Brown, 2015). Researchers recognise
that the cost of fixing errors grows exponentially as a function of elapsed time to discovery
(Shanks, Moody, Nuredini, Tobin, & Weber, 2010). IS failures in Australia include the
Chapter 1:Introduction 3
Queensland Health payroll debacle, where reports reveal that approximately $210 million and
18 months were needed to fix the system, while the original system development cost was $65
million (Glass, 2013). Conceptual models play a significant role in facilitating early detection
and correction of system development errors, such as design flaws or misunderstanding
between users and developers (Liddle, 2011; Maes & Poels, 2007; Wand & Weber, 2002).
This is particularly true when dealing with large and complex domains, as using appropriate
models improves understanding of the domain and its requirements, and therefore contributes
to a high-quality information system.
1.2 PROBLEM STATEMENT
System analysts and designers normally use more than one conceptual model in most
system analysis and design tasks (Fettke, 2009; Whittle, Hutchinson, & Rouncefield, 2014).
In fact, practitioners use multiple models, often developed using different grammars, to better
represent and understand the system under study. Although using multiple models assists
practitioners to better understand a domain (da Silva, 2015; J. Kim, Hahn, & Hahn, 2000;
Siau & Lee, 2004), it introduces problems, such as challenges in integrating different models,
inconsistency between models, or disproportionate or incomplete representation (e.g., more
details on some parts than others) (Lange, Chaudron, & Muskens, 2006; Mussbacher et al.,
2014; Petre, 2013).
Given the importance of conceptual modelling in system analysis and design (Bera et
al., 2014; Burton-Jones, Wand, & Weber, 2009; Wand & Weber, 2002), it is expected that IS
literature provides practical and theoretical guidance to explain the use of multiple and
different types of models. However, IS literature is inconsistent with practice. While empirical
evidence suggest that practitioners often use multiple models in their analysis and design
tasks (Fettke, 2009; Hutchinson, Whittle, Rouncefield, & Kristoffersen, 2011), the vast
majority of relevant research on conceptual modelling has focused on the use of single
modelling grammar or model. By contrast, research on the use of multiple models is very
sparse. To address this problem, Recker (2014) proposed a theory of faithful use of
4
conceptual model combinations2 to predict and explain the combined use of models.
However, this theory has not yet been tested.
In addition to the lack of research in IS literature to explain how or why practitioners
use multiple models, the ever-changing world also stresses the importance of reviewing and
revising our own research efforts in conceptual modelling (Recker, 2015). Similar to other
fields, conceptual modelling research is rightfully under constant scrutiny in terms of its
validity, applicability, relevance, and utility in the ever-changing world. New requirements
have emerged that go beyond the boundaries of traditional information systems. For example,
open information environments (Parsons & Wand, 2014) extend information systems beyond
organisational boundaries. Open information environments provide opportunities for users to
use and generate new information where organisations may have no control over structure and
content. In these environments, new sources and new uses of data may emerge (Lukyanenko
et al., 2017). In addition, the rapidly changing IS field creates new demands (Storey & Song,
2017) and novel forms of representations that challenge assumptions of traditional conceptual
modelling (Lukyanenko & Parsons, 2013; Recker, 2015).
Therefore, this research is set out in a multi-method research program to make
contributions to the body of knowledge on conceptual modelling in a several ways: first,
through a systematic literature review, this thesis shapes the current understanding of
conceptual modelling and stimulates a forward-looking wave of conceptual modelling
research. Second, through an exploratory study and a theory-testing study, this research
provides empirical insights to explain the use of model combinations and refines and
improves the theory of faithful use of conceptual model combinations. These contributions are
positioned in three individual but related studies (see Table 1.1): one systematic literature
review, with the primary focus on the existing studies on conceptual modelling and proposing
a research framework that proposes opportunities for future studies in the field; one
qualitative study, with the primary focus on exploring the use of multiple conceptual models
in practice; and one laboratory experiment, with the primary focus on testing and extending
the theory of faithful use of conceptual model combinations.
2 See also Section 2.4 for more details on the theory
Chapter 1:Introduction 5
Table 1.1 Papers included in this dissertation
Relevant
Chapters Study Type
4
Conceptual Modelling Research in Information Systems: What we know now
and what we still do not know. To be submitted to Information Systems
Research
A short version of the paper was published as: Jabbari Sabegh, M. A., Lukyanenko, R., Recker, J. C., Samuel, B., & Castellanos, A. (2017).
Conceptual modeling research in information systems: What we now know and
what we still do not know. Paper presented at the Symposium on Research in
Systems Analysis and Design (AIS SIGSAND 2017).
Systematic
Literature review
5
Combined Use of Conceptual Models in Practice: An Exploratory Study
Published as: Jabbari Sabegh, M. A., & Recker, J. (2017). Combined Use of
Conceptual Models in Practice: An Exploratory Study. Journal of Database
Management, 28(2), 56-88
Qualitative
6
An Experimental Study to Test and Extend the Theory of Faithful Use of
Conceptual Model Combinations. To be submitted to MIS Quarterly
An initial version of the paper was published as: Jabbari Sabegh, M. A., Recker, J., & Green, P. (2016). Designing Experiments to Test the Theory of
Combined Ontological Coverage. Paper presented at the International
Conference on Information Systems, Dublin, Ireland
Quantitative
1.3 RESEARCH QUESTIONS
The three studies in this thesis aim to answer three research questions. The first research
question is a general question regarding the current state of knowledge about conceptual
modelling. Conceptual modelling is regarded as a core activity of information systems
analysis and design. Research in this field is an active research area, with the aim of
improving both the process and the outcome of conceptual modelling. In 2002, Wand and
Weber proposed a seminal research agenda to motivate research addressing the question of
“How can we model the world to better facilitate our developing, implementing, using, and
maintaining more valuable information systems?” (Wand & Weber, 2002, p. 363). After 15
years of raising this question, the first research question in this study is:
Research Question 1: What is the current state of knowledge of conceptual modelling
in IS?
This research question aligns with Study 1 (Chapter 4). To answer this question, an
understanding of the current state of knowledge in conceptual modelling and addressing new
perspectives of conceptual modelling is required. This involves a systematic literature review
of existing studies about conceptual modelling in IS to identify the aspects of conceptual
6
modelling addressed by existing studies, the extent of the contributions of these studies, and
the aspects of conceptual models that have not yet been addressed. In doing so, this thesis
contributes to the body of knowledge in conceptual modelling by synthesising the current
understanding and assumptions about conceptual modelling and providing substantive
directions for future research.
The second and third research questions focus on the combined use of conceptual
models. Empirical evidence suggests that practitioners use model combinations in their
systems analysis and design tasks. While Recker (2014) provided a theoretical explanation for
the combined use of conceptual models, relatively little is currently known about the nature of
the use of multiple models in practice. Therefore, the second research question is:
Research Question 2: Why do systems analysts and designers use multiple conceptual
models?
The second research question aligns with Study 2 (Chapter 5). To answer this question,
an exploratory study was required to understand the current practices in the actual use of
model combination. By answering this question, this thesis contributes to theory in a number
of ways: First, it makes an important contribution to the literature on conceptual modelling.
To the best of my knowledge, this study presents the first empirical investigation into the
actual use of model combinations in practice. Second, the study explores factors that impact
the use of conceptual models and suggests how existing theories might be improved to
provide better explanations about the use of models in practice.
The third research question of this thesis follows the findings from the exploratory
study, with the primary focus on the propositions of the theory of faithful use of conceptual
model combinations. This theory provides three propositions to evaluate the combined use of
conceptual models in terms of user’s selection, understanding, and perceived usefulness of
model combinations. Therefore, the third research question is:
Research Question 3: What characteristics of models affect users’ selection,
understanding, and perceived usefulness of model combinations?
This research question aligns with a theory-testing study (Study 3, Chapter 6). A
controlled investigation was required to investigate the exact impact of proposed factors on
users’ selection of models, generation of domain understanding from multiple models, and
Chapter 1:Introduction 7
perceived usefulness of model combinations. Therefore, a laboratory experiment was set to
control extraneous variables and manipulate the causes. Study 3 makes a number of
significant contributions. First, to the best of my knowledge, it is the first dedicated empirical
study set to evaluate the propositions of the theory of faithful use of conceptual model
combinations. The results of this study provide empirical evidence regarding the propositions
of the theory and extend the boundaries of the theory. Second, the findings of this study
provide practical guidance for both model users and designers to use and develop appropriate
combinations of conceptual models.
1.4 SCOPE OF STUDY
Research on conceptual modelling appears prolifically in academic journals. To scope
this research, Study 1 focuses on the papers published in the Association for Information
Systems (AIS) basket of eight journals (F. Liu & Myers, 2011), plus the Journal of Database
Management. These journals were selected because the AIS basket of eight journals are
considered by the academic community as appropriately representative for the mainstream
research in the field by the academic community and the Journal of Database Management
has traditionally been one of the leading substantive journals publishing studies about
conceptual modelling. However, the literature review is a continuing process, and the
literature review sections of Studies 2 and 3 are independent and not limited to the papers
reviewed in Study 1.
System analysis and design tasks typically have many different stakeholders, such as
top management, project management, system analysts, system designers, and system end-
users. Conceptual modelling during the systems analysis and design process has two phases:
model creation, which applies to those who develop models; and model interpretation, which
applies to those who read the models to understand the domain represented by models
(Gemino & Wand, 2004). These two phases are not often mutually exclusive. Stakeholders
can be a model user or both a model creator and a user. In either case, stakeholders read the
models and apply them in support of their tasks. A model creator may read models for the
purpose of redesigning and modification. The scope of Studies 2 and 3 was to investigate the
combined use of models at the individual level and from a model user’s perspective, as
opposed to model creators.
8
Information systems analysts and designers may use various types of conceptual
models. At least 40 different modelling grammars (languages) are cited in a current
practitioners’ survey (Whittle et al., 2014). This thesis focusses on Unified Modelling
Language (UML) and users of UML models. Previous studies (Davies, Green, Rosemann,
Indulska, & Gallo, 2006; Fettke, 2009) indicate that UML is among the most common
conceptual modelling techniques used in systems analysis and design. Object-oriented
modelling is often promoted as an ideal option to solving problems in software developments,
and UML has achieved significant popularity as the standard for object-oriented system
development (Batra, 2009; Dobing & Parsons, 2008). In addition, UML offers a variety of
conceptual modelling constructs to support an object-oriented approach to system analysis
and design (Gemino & Parker, 2009). Moreover, a large number of practitioner articles and a
substantial body of research on UML has emerged (Evermann & Wand, 2001; Petre, 2013),
such as evaluation of the effectiveness of various UML models (Siau, 2010), and analysis of
the language’s complexity (Siau, Erickson, & Lee, 2005).
1.5 THESIS STRUCTURE
This thesis is organised into three parts. Part A provides the preface and an overview of
the studies. Part B presents the main body of this thesis, which includes three studies. Part C
provides the summary of the findings from the three studies.
1.5.1 Part A
Part A is structured into three chapters.
- Chapter 1 provides an introduction to this thesis and introduces the motivations
of the study, followed by the research questions and scope of the research.
- Chapter 2 provides the background of the research on conceptual modelling in
information systems and then discusses the theoretical foundation of the thesis.
- Chapter 3 outlines the research methods implemented in the three studies. The
reliability and validity of the selected research methods are then discussed.
1.5.2 Part B
Part B presents the three studies that were conducted to answer the three main research
questions of this thesis. Figure 1.2 illustrates the three studies as the main body of the thesis,
Chapter 1:Introduction 9
corresponding to the research questions. In Part B of this thesis, the term “we” is used
because the three studies were conducted in collaboration with the co-authors of the studies.
Thesis title:
A Study of the Combined Use of Conceptual Models
What is the current state of knowledge
of conceptual modelling in IS?
Semi-structured interviews
What characteristics of models affect
users’ selection, understanding, and
perceived usefulness of model
combinations?
Study 1:
Why do systems analysts and designers
use multiple conceptual models?
Experiment
Research Question 1 Research Question 2 Research Question 3
Literature review
Study 2: Study 3:
Provided an overview of the
current state of knowledge in
conceptual modelling
Motivated further research on
conceptual modelling
Explored how practitioners
reportedly use conceptual
models
Extended implications of
ontological theories in
conceptual modelling
Tested and extended the theory
of faithful use of conceptual
model combinations
Developed an understanding on
the use of model combinations
Provided results from an extensive investigation into the combined use of conceptual models , and it
attempts to validate and extend the theory of faithful use of conceptual model combinations.
Contribution of Study 1: Contribution of Study 2: Contribution of Study 3:
Contribution of the research:
Figure 1.2 Outline of thesis structure and research objectives
- Study 1 (Chapter 4) reviewed existing research on conceptual modelling in IS.
In doing so, the research on conceptual modelling published in high-quality IS
journals was reviewed. The objective of this research was to provide an
understanding of what is already known and what is unknown in conceptual
modelling. This paper highlights the research areas in conceptual modelling that
appear to be neglected and provides a new framework that draws the attention of
conceptual modelling research to future research opportunities. However, the
literature review is a continuing part of the research program in all stages of the
research to express the link of the current research with the previous work and
contributions of current research to the knowledge in the area (Neuman, 2006).
10
- Study 2 (Chapter 5) conducted an exploratory study to understand why
practitioners use different types of conceptual models in combination. From the
findings of Study 1, it became clear that there is a lack of knowledge in the IS
literature to explain the use of multiple models in system analysis and design
tasks. However, in practice, practitioners use more than one model in their
design and analysis tasks (Dobing & Parsons, 2008; Fettke, 2009). On the other
hand, there are arguments that challenge the use of models in requirements
gathering in system analysis and design (B. Ramesh, Cao, & Baskerville, 2010).
The different viewpoints highlight that the use of multiple models is a poorly
understood phenomenon. Therefore, semi-structured interviews were conducted
to explore why IS professionals use conceptual models in practice.
- Study 3 (Chapter 6) was a theory-testing study. The findings of Study 1
highlighted the lack of knowledge in IS literature describing the use of multiple
models. Study 2 indicated that similar to predictions of the theory of faithful use
of conceptual model combinations, ontological factors have potential impacts on
using different types of conceptual models. The theory of faithful use of
conceptual model combinations provides explanations about how ontological
factors may affect model selection, domain understanding, and perceived
usefulness of model combinations. However, there are no empirical insights to
explain the phenomena. Study 3 aimed at investigating the use of multiple
models. An experimental study was conducted to examine the propositions of
the theory of faithful use of conceptual model combinations.
1.5.3 Part C
Part C of this thesis provides a summary of the findings from three studies.
- Chapter 7 synthesises the findings, and discusses the implications that stem
from the combination of the three individual studies.
- Chapter 8 discusses the substantive and theoretical contributions of this
research. The limitations that pertain to this thesis and the outlines of future
research opportunities are then presented.
Chapter 2:Research Background 11
Chapter 2: Research Background
2.1 CONCEPTUAL MODELLING IN INFORMATION SYSTEMS
Research on conceptual modelling has gained widespread attention as the basis for
information systems analysis and design (Moody, 2005). Information systems are developed
to support organisational activities in a certain domain based on an understanding of that
domain (Bera et al., 2014). Such an understanding is supported by developing a conceptual
model independent of the technology used to construct the system (P. P.-S. Chen, 1976;
Denning, 2003).
Conceptual models are often graphical representations of human perceptions of the
semantics of a real world domain. These models are normally developed in such a manner
that they can be represented in an information system (Kent, 1978; Wand & Weber, 2002).
Conceptual modelling involves capturing various aspects of the real world domain to promote
an understanding of the conceptual aspects of a domain to be supported by the information
system prior to designing and programming the system (Bera et al., 2014; Burton-Jones et al.,
2009). However, it also extends to further application areas, such as supporting
communication between users and analysts, or analyst and developers; providing input for the
design process; and documenting the original requirements for future reference (Siau, 2004;
Wand & Weber, 2002).
The field of conceptual modelling was founded approximately 40 years ago (Storey,
Trujillo, & Liddle, 2015). Since then, research on conceptual modelling has actively
investigated conceptual modelling related phenomena and evaluated capabilities of conceptual
models (Burton-Jones et al., 2009; Maes & Poels, 2007). Researchers argue that using high-
quality conceptual models in the early stages of information systems analysis and design
processes can enable early detection and correction of errors (Wand & Weber, 2002). The
relatively high failure rate of information systems highlights the importance of conceptual
modelling in systems analysis and design (Dwivedi et al., 2015).
Conceptual modelling consists of a grammar (i.e., a set of constructs and rules that show
how to combine the constructs to model a real-world domain), a method (i.e., provide
12
procedures that explain how to use a grammar), a script (i.e., the product of the modelling
process), and a context (i.e., the setting in which conceptual modelling occurs, and scripts are
used) (Wand & Weber, 2002). For example, entity-relationship modelling grammar (P. P.-S.
Chen, 1976) provides constructs such as entity, relationship, and attribute. Rules in the entity-
relationship modelling grammar specify that an entity can have an attribute and two entities
can be associated only through a relationship.
Research on conceptual modelling has focused on evaluating capabilities of existing
conceptual modelling grammars and methods and developing new ones. Different types of
conceptual modelling grammars and methods have been developed for multiple reasons. First,
different types of models are used for different purposes. For example, data models represent
the data structures of systems (P. P.-S. Chen, 1976), process models represent the operations
and dynamics of systems (Petri, 1962), and enterprise models represent the structure of an
organisation (Checkland & Scholes, 2000). Second, different models have been developed
that claim to be superior to previous ones. For example, different types of process modelling
grammars, such as Business Process Modelling Notation (BPMN) grammar (OMG, 2011a),
Event-driven Process Chains (EPCs) (Scheer & Nüttgens, 2000), Yet Another Workflow
Language (YAWL) (van der Aalst & ter Hofstede, 2005), or UML-Activity (OMG, 2011b).
Thirdly, different approaches to system analysis and design emerged. For example, object-
oriented modelling to describe a system in the forms of objects (Booch, 1986) or object-
process modelling to describe both the structure and the operations of the system by
representing blocks object and process (Dori & Goodman, 1996), just to name a few. Still, the
ever-changing world requires constant improvement of modelling methods and innovative
approaches as conceptual modelling moves beyond traditional boundaries (Lukyanenko &
Parsons, 2013; Lukyanenko, Parsons, & Samuel, 2015; Recker, 2015), for example, flexible
database technologies (i.e., noSQL databases) (Kaur & Rani, 2013) and instance-based
conceptual modelling (Lukyanenko et al., 2015).
Traditional research on conceptual modelling can be categorised into two streams of
research viz. research on conceptual model design and research on conceptual model
interpretation (Gemino & Wand, 2004). Research on conceptual model design focuses on the
capabilities of modelling grammars, how to use grammars and the context in which
conceptual modelling occurs from a model designer (model creator) perspective. On the other
Chapter 2:Research Background 13
hand, research on conceptual model interpretation focuses on the capabilities of developed
models and how a model user (model viewer) reads and understands a developed model.
These two types of research about conceptual modelling are not mutually exclusive,
because, first, the capability of a model (script) is constrained by the capabilities of the
modelling grammar used to develop that model. Second, researchers normally evaluate the
capabilities of modelling grammars based on developed scripts (i.e., how understandable is
the model?) (Aguirre-Urreta & Marakas, 2008; Burton-Jones et al., 2009; Gemino & Wand,
2004; Parsons & Cole, 2005; Siau, 2004; Wand & Weber, 2002). The dominant objective of
research on conceptual modelling is to improve the conceptual modelling process and
outcomes (Siau & Rossi, 2011). Improving the quality of conceptual models is also likely to
improve the quality of developed systems (Moody, 2005). The next section briefly reviews
the research on single modelling grammars and models.
2.2 RESEARCH ON SINGLE GRAMMARS OR MODELS
2.2.1 Conceptual Model Design
Research on conceptual model design has widely focused on the evaluation of
capabilities of grammars and methods (e.g., Bera, 2012; Bera, Burton-Jones, & Wand, 2011)
and developing new ones (e.g., Loucopoulos & Kadir, 2008; Lukyanenko et al., 2017).
Researchers proposed several guidelines to evaluate conceptual modelling grammars and
scripts (e.g., Burton-Jones et al., 2009; Gemino & Wand, 2003). For many years, the leading
approach in this area has relied on Bunge, Wand, and Weber (BWW) representation theory
(Wand & Weber, 1990a, 1993, 1995).
The BWW representation theory developed from an adaptation of an ontology proposed
by Bunge (1977). In general, ontology explains the nature of the real-world, the interaction
between what exists in reality, and describes the properties and structures of real-world things
(Bunge, 1977). Since information systems are representations of real-world domains that are
essentially created by humans, the representation theory based on ontology may lead
modellers to conceive of conceptual structures to represent a real-world domain (Wand &
Weber, 1993, 1995).
The BWW representation theory suggests that conceptual modelling grammars must be
able to represent all features of the real-world domain that the information systems users are
14
interested in. Otherwise, the created models based on the grammars would be incomplete, and
the designers/analysts would have to somehow augment the models (Weber, 1997). The
theory proposes four possible ontological deficits: construct deficit, construct overload,
contract redundancy, and construct excess.
Construct deficit: an ontological construct does not map into any grammatical
construct (a 1:0 mapping).
Construct overload: one grammatical construct maps into several ontological
constructs (a m:1 mapping).
Construct redundancy: two or more grammatical constructs map into one
ontological construct (a 1: m mapping).
Construct excess: a grammatical construct does not map into any ontological
construct (a 0:1 mapping).
Research on conceptual modelling based on BWW theory has mostly focused on
ontological completeness (i.e., the existence of construct deficit) and ontological clarity (i.e.,
the existence of construct overload, redundancy, and excess) (Wand & Weber, 1995) of
modelling grammars and scripts.
Three dominant findings from ontological studies are highlighted below:
Any single modelling grammar, and therefore any single model, has some level
of ontological deficit (e.g., Green & Rosemann, 1999; Irwin & Turk, 2005;
Opdahl & Henderson-Sellers, 2002; Recker et al., 2010; Wand & Weber, 1989,
1993; Weber, 1996; Weber & Zhang, 1996).
Following ontological guidelines improves the conceptual modelling process
and outcome (e.g., Bowen, O'Farrell, & Rohde, 2006; Milton, Rajapakse, &
Weber, 2012; Saghafi & Wand, 2014; Shanks, Tansley, & Weber, 2004).
The existence of ontological deficits affects the use of conceptual modelling
grammars and models (e.g., Recker, Rosemann, Green, & Indulska, 2011).
Research into conceptual modelling grammars is not limited to ontological evaluations.
Researchers have gone beyond the ontological capabilities of grammars and proposed other
quality measurements and guidelines to develop high-quality conceptual models. For
Chapter 2:Research Background 15
example, Clarke, Burton-Jones, and Weber (2016) proposed that ontological evaluation of
modelling grammars should be coupled with logical evaluation. Other examples of quality
measures are syntactic quality (i.e., how well all statements in the model correspond to the
syntax of the grammar), and semantic quality (i.e., how well all statements in the model
correspond to the domain), pragmatic quality (i.e., how well is the correspondence between
the model and the audience’s interpretation of the model defined), and perceived semantic
quality (e.g., Krogstie, 2001; Krogstie, Sindre, & Jørgensen, 2006; Recker, 2007). Similarly,
Mendling, Reijers, and van der Aalst (2010) proposed seven guidelines to develop high-
quality conceptual models, or activity labelling (Mendling, Reijers, & Recker, 2010).
In developing conceptual models, researchers have also investigated the contextual
factors affecting the modelling process. For example, studies uncovered factors such as
organisational conventions (Delfmann, Herwig, Lis, & Stein, 2015; Recker et al., 2010;
Rosemann, 2006; Samuel, Watkins III, Ehle, & Khatri, 2015), and tool supports and tool
functionality (Fettke, 2009; La Rosa et al., 2011; Recker, 2012) as influential factors on
conceptual modelling processes. However, the findings uncovered that the availability of tools
does not always lead to producing high-quality conceptual models (Poppe, Brown, Recker,
Johnson, & Vanderfeesten, 2017; Recker, Mendling, & Hahn, 2013).
2.2.2 Conceptual Model Interpretation
Research on conceptual model interpretation has focused on both model related factors
and contextual factors, such as cognitive capabilities and individual difference factors. Similar
to research on modelling grammars, research on model interpretation has evaluated the impact
of ontological capabilities of models on model understandability. Examples of this research
include the impact of ontological clarity on model understandability (e.g., Allen & March,
2006; Bodart, Patel, Sim, & Weber, 2001; Bowen et al., 2006; Shanks, Tansley, Nuredini,
Tobin, & Weber, 2008). The findings of these studies are equivocal. Some findings have
revealed that ontological clarity has not always led to better understanding or does not have a
clear effect on model user’s performance (e.g., Bowen, O'Farrell, & Rohde, 2009; Fickinger
& Recker, 2013).
Examples of research beyond the ontological capabilities of models are the effects of
perceptual discriminability and semiotic clarity of different process models on model
16
understandability (Figl, Mendling, & Strembeck, 2013), the effect of perceptual
discriminability of gateway constructs on a user’s abilities to interpret process models
(Recker, 2013), the effects of personal, model, and content-related factors (Mendling &
Strembeck, 2008), or the effects of the interactivity of model elements and presence of a
different control flow (Figl & Laue, 2015).
In addition to the capabilities of a model stemming from the capabilities of a grammar,
researchers have identified additional aids, such as visual cues (J. Kim et al., 2000), text
(Gemino & Parker, 2009), customisation (Samuel et al., 2015), colours (Masri, Parker, &
Gemino, 2008; Reijers, Freytag, Mendling, & Eckleder, 2011), and routing symbols (Figl,
Recker, & Mendling, 2013) that can influence model interpretation.
Researchers have also found contextual factors and individual differences to be
influential factors on a user’s ability to understand a model (Figl, 2017). Examples of this
stream of research are the effects of cognitive style and task characteristics, cognitive abilities,
and learning style (Figl & Recker, 2014; Recker, Reijers, & van de Wouw, 2014), and IS and
domain knowledge (Bera et al., 2014; Khatri & Vessey, 2016; Khatri, Vessey, Ramesh, Clay,
& Park, 2006) on an individuals’ ability to interpret a model.
2.3 RESEARCH ON MULTIPLE GRAMMARS OR MODELS
Unlike research on single grammars or models, research on multiple grammars or
models is sparse. Conceptual modelling can take many forms. With the availability of
different modelling grammars and methods, empirical evidence suggests that practitioners
normally use multiples in combination in their systems analysis and design tasks (Davies et
al., 2006; Fettke, 2009; Green, Rosemann, Indulska, & Recker, 2011; Petre, 2013; Recker,
2012; Whittle et al., 2014).
Ontological theories predict how multiple grammars can be used in combination
(Green, 1996; Weber, 1997). Based on an extension of the BWW representation theory, Green
(1996) argued that the combination of multiple modelling grammars should achieve Maximal
Ontological Completeness (MOC) (i.e., a combination of grammars that maximises
grammatical constructs that cover all phenomena to be modelled) and Minimal Ontological
Overlap (MOO) (i.e., a combination of grammars that minimises overlapping grammatical
constructs to represent real-world phenomena). A few studies then investigated which
Chapter 2:Research Background 17
combinations of grammars model designers might select. For example, Green, Rosemann,
Indulska, and Manning (2007) analysed multiple grammars to define systems interoperability
and examine which combinations of grammars could theoretically achieve the best coverage.
zur Muehlen and Indulska (2010) investigated the appropriate combination of business rules
and business process modelling grammars, and Green et al. (2011) analysed how users
employed multiple modelling grammars in combination to model a system in a CASE tool.
Few studies have investigated the use of multiple models. The existing studies in this
area found that understanding of multiple models is facilitated by incorporating visual cues
and contextual information (J. Kim et al., 2000). Use case and class diagrams have very little
overlap in the information captured, and both are perceived as necessary for requirement
analysis (Siau & Lee, 2004), and using use cases in combination with use case diagrams
results in better performance (Gemino & Parker, 2009). These studies indicate that the
combined use of models improves domain understanding. However, these findings did not
indicate which attributes and factors define the best combination of models. Recker (2014)
proposed a theory of faithful use of conceptual model combinations to explain the factors that
may affect the combined use of conceptual models. Recker’s (2014) theory is explained in
more detail in the following section.
2.4 THE THEORY OF FAITHFUL USE OF CONCEPTUAL MODEL
COMBINATIONS
Practitioners normally use several types of conceptual models or grammars during
information systems analysis and design processes (Dobing & Parsons, 2008; Fettke, 2009;
Recker, 2012). Empirical evidence has suggested that practitioners do not substitute models,
they use combinations of models developed using various grammars (Green et al., 2011;
Petre, 2013; Recker, 2012). However, research on conceptual modelling has focused almost
exclusively on single grammars or models and largely neglected this use of combined models.
Recker (2014) developed the theory of faithful use of conceptual model combinations,
which suggests propositions to explain and predict the use of multiple conceptual models in
combination. Recker (2014) indicated that model use occurs as part of a particular task at
hand. Model users are those individuals who read and apply models in support of the task at
hand (Gemino & Wand, 2004). Model users also refer to model readers or model interpreters
as opposed to those who develop models. The use of models involves reading models to
18
develop an understanding about the presented domain in support of the task at hand.
Therefore, the faithful use of models by an individual in a particular task will be evaluated in
light of the extent to which the individual was able to accomplish the task (Recker, 2014).
Consequently, Recker (2014) indicated that the faithful use of conceptual models can be
evaluated in the three steps of engaging in a particular task:
1) Prior to engaging in a model-based task: can be captured as the perceived
performance model-users expected by selecting a set of conceptual models for a
specific task from different types of models.
2) During the engagement in the model-based task: evaluate the performance based
on the extent of domain understanding learned from using conceptual models.
3) After engaging in the model-based task: evaluate the beliefs about the performance
based on the extent of domain understanding learned from using conceptual models.
The primary conjecture of the theory of faithful use of conceptual model combinations
is that the user’s selection of a set of models, the domain understanding they gain from
multiple models, and the perceived usefulness of multiple models will be dependent on the
models’ combined ontological completeness and ontological overlap. Combined ontological
completeness refers to the level of representational coverage a set of models provides about a
focal real-world phenomenon. Ontological overlap refers to the set of redundant
representation across a set of models representing real-world phenomena (Recker, 2014).
Figure 2.1 illustrates the two premises of the theory of faithful use of conceptual model
combinations.
Figure 2.1 Premises of the Theory of Faithful Use of Conceptual Model Combinations (Recker, 2014)
Completeness of
domain representation
of model A
Completeness of
domain representation
of model B
Resulting
ontological overlap
between models A
and B
Remaining construct
deficit of combined
representation
Chapter 2:Research Background 19
The basic rationale behind this conjecture is that, first, users will select multiple models
to maximise their coverage of some relevant real-world phenomena (Green, 1996). This is
because ontological evaluations of modelling grammars show that any single modelling
grammar suffers from construct deficit (Green et al., 2011; Irwin & Turk, 2005; Siau, 2010).
In other words, no one grammar can be used to model all potentially relevant real-world
phenomena. Moreover, often, not all constructs provided via a grammar are used in any given
model (Siau et al., 2005; zur Muehlen & Recker, 2008). Therefore, users will use multiple
models in combination to overcome ontological deficits in a single model and to maximise
combined ontological completeness through multiple models.
Second, users will also seek to minimise ontological overlap in the representation to
keep the coverage as simple and clear as possible (Wand & Weber, 1990a, 1990b, 1993).
According to Green (1996), the primary intention to select multiple grammars is to reach
maximum ontological completeness. However, some studies suggest that a simpler
representation may be more useful than a more complete representation (Siau & Lee, 2004;
zur Muehlen, Recker, & Indulska, 2007).
On the basis of these two main factors, the theory suggests three propositions that detail
the faithful use of multiple models:
Model selection: model users will select a combination of conceptual models to
maximise combined ontological completeness; however, they will select additional
models until the model combination reaches a particular level of overlap that users
perceive to be bearable.
Domain understanding: the level of understanding can be increased by presenting
more elements of a domain that are represented in conceptual models (ontological
completeness); however, this level of understanding will be moderated by the level
of ontological overlap among models.
Perceived usefulness: Perceptions of the usefulness of model combinations will
increase when the ontological completeness of the achieved representation
increases; however, a combination of models with high ontological overlap, that is,
high redundant constructs, will be evaluated as less useful.
20
2.5 CONCLUSION
Three main points are evident from the research on conceptual modelling in information
systems and the theory of faithful use of conceptual model combinations.
First, apart from a few studies (Dobing & Parsons, 2008; Fettke, 2009; Gemino &
Parker, 2009; Green et al., 2007; Green et al., 2011; J. Kim et al., 2000; Siau & Lee, 2004),
most studies on conceptual modelling have focused on using single grammars or models.
Research on conceptual modelling is motivated by evaluating single grammars (Green et al.,
2007; Krogstie, 2001; Maes & Poels, 2007; Opdahl & Henderson-Sellers, 2002; Wand &
Weber, 1989), different conceptual modelling methods (Siau & Rossi, 2011), as well as the
performance of users of a single model (Allen & March, 2006; Bowen et al., 2006; Chan,
Wei, & Siau, 1993). Most of the studies on conceptual modelling have used Wand and
Webers’ (1990a, 1993, 1995) representation theory adapted from Bunge’s (1977) ontology as
their theoretical foundation to evaluate a single model or grammar.
Second, evaluation of conceptual modelling grammars revealed that almost any single
model has some level of construct deficits (Green & Rosemann, 1999; Irwin & Turk, 2005;
Opdahl & Henderson-Sellers, 2002; Recker et al., 2010; Wand & Weber, 1989, 1993; Weber,
1996; Weber & Zhang, 1996). Therefore, no single model would provide a complete
representation of a real-world domain. This is because the representational completeness of
any single model is limited to the extent of the level of construct deficit in any single
grammar. Moreover, the actual level of completeness of any single model is often less than
the level of completeness of the grammar used to develop the model, as most models do not
contain all grammatical constructs - only a small subset (zur Muehlen & Recker, 2008; zur
Muehlen et al., 2007).
Third, Recker’s (2014) theory proposed that the combined use of conceptual models
depends on two ontological factors of combined ontological completeness and ontological
overlap. Based on these factors, the theory then proposes how users may select models, how
domain understanding can be generated from the model combinations, and how the perceived
usefulness of model combinations may vary.
This thesis focusses on the use of multiple models and how conceptual models in
combination are interpreted and understood by users, as opposed to how people create or
Chapter 2:Research Background 21
develop models (see Gemino & Wand, 2004). Previous studies have highlighted the use of
model combinations (Fettke, 2009). Recker’s (2014) theory of faithful use of conceptual
model combinations provides a foundation and insights to predict and explain the use of
models in combination. The theoretical foundation of the current study is therefore built on
the principles of the theory of faithful use of conceptual model combinations, with the
objectives to test and extend the propositions of the theory.
22
Chapter 3: Research Methods
3.1 RESEARCH METHOD CHOICES
This section explains the research methodologies adopted to collect, analyse, and report
data on the use of conceptual model combinations. This research employed a multi-methods
design. The success of each stage of the research depends on the selection of appropriate
research methods based on the objectives of each stage and the state of knowledge in the area
being studied (Mingers, 2001). This research aims to answer three main questions.
The first research question of this thesis requires understanding of the current stage of
knowledge in the field; thus, a systematic literature review was carried out. An exploratory
study was conducted to answer the second research question. To answer the third research
question, an explanatory, theory-testing, approach to the research was undertaken.
IS researchers have used various types of research methods that can, at one level, be
broadly categorised into qualitative or quantitative methods (Lee & Hubona, 2009;
Venkatesh, Brown, & Bala, 2013). Both qualitative and quantitative methods can possibly be
used to answer similar research questions. However, qualitative methods have typically been
used to investigate little-known phenomena, and explore and develop new theories
(Eisenhardt, 1989; Walsham, 2006). On the other hand, quantitative methods are typically
explanatory studies and have been used to test theories.
3.1.1 Systematic Literature Review
In order to advance the field of conceptual modelling research in IS, it is useful to
explore the history of the field and to determine its current state by identifying, analysing, and
interpreting the relevant research to date that is related to conceptual modelling. To achieve
this objective, a rigorous and systematic approach was undertaken (Bandara, Furtmuller,
Gorbacheva, Miskon, & Beekhuyzen, 2015; Kitchenham et al., 2009; Webster & Watson,
2002) and a descriptive literature review was conducted as described by Paré, Trudel, Jaana,
and Kitsiou (2015) and Rowe (2014).
Chapter 3:Research Methods 23
In contrast to what is occurring in practice, there is a lack of knowledge in IS literature
to explain how and why practitioners use different types of conceptual models. Research on
conceptual modelling continues to be an active research area in information systems research
(Burton-Jones et al., 2009). Many aspects of conceptual modelling have been addressed
through ongoing research studies appearing in top IS journals. However, it appears that little
research has been undertaken into some other aspects of conceptual modelling. A systematic
literature review was therefore conducted to determine the current state of the knowledge in
the field and propose a framework to motivate further research on little-known aspects of
conceptual modelling.
3.1.2 Semi-Structured Interviews
To achieve the second objective of this study, that is, to explore how and why
practitioners use multiple models in combination, an exploratory, qualitative research design
was used. Qualitative research enabled the completion of a study in an area in which few
previous studies exist (Benbasat, Goldstein, & Mead, 1987; Myers & Newman, 2007; Yin,
2009). Specifically, this was achieved through the use of semi-structured interviews.
Interviews are a dominant data collection method in qualitative research techniques
(Yin, 2009), such as case study research or grounded theory studies (Myers & Newman,
2007). Semi-structured interviews can be used with any philosophical perspective, whether
positivist, interpretive, or critical (Myers & Newman, 2007), deductive or inductive (Shanks,
2002), descriptive, explanatory, or exploratory (Dubé & Paré, 2003). To ensure reliability and
validity, the guidelines of leading positivist qualitative researchers were applied (Shanks,
2002; Yin, 2009). An inductive approach was also used to explore theoretical concepts and
patterns from qualitative data (Bhattacherjee, 2012).
The notion that multiple grammars are used in practice has previously been explored,
firstly, by Weber (1997) and Green (1996). However, there is a little knowledge about the use
of conceptual models in combination. Recker (2014) formulated a theory of faithful use of
conceptual model combinations. However, what is notably absent is empirical evidence about
how practitioners work with multiple models. The existing theories predicted the use of
multiple models; however, use of multiple models is a poorly understood phenomenon. For
example, in addition to the characteristics of models, contextual factors and different
24
individuals’ involvement in model-based activities may affect how and why practitioners use
model combinations.
3.1.3 Experiment
To achieve the third objective of this study, that is, to test which characteristics of
models affect users’ selection, understanding, and perceived usefulness of model
combinations, an experimental study was adopted. Given the nature of this objective, the
experimental study was used to obtain preliminary evidence to explain and predict the spirit
of the use of model combinations, rather than only describing the phenomena. Experiments
have commonly been used in information systems research to build knowledge on system
analysis, design, development, and management (Palvia et al., 2004). The third objective of
this study is to test and extend the theory of faithful use of conceptual model combinations.
While this is the first study that aims to test the theory, the primary objective was to ensure
strong internal validity (Calder, Phillips, & Tybout, 1981).
Experimental research is one of the most rigorous of all research designs (Bhattacherjee,
2012). The main strength of experimental research is its internal validity due to its ability to
link cause and effect through treatment manipulation. An experimental design also provides
unique abilities in controlling external factors that might have spurious effects (Shadish,
Cook, & Campbell, 2002).
3.2 RELIABILITY AND VALIDITY
The reliability and validity of the research methods adopted for this thesis are discussed
in two parts. First, the reliability and validity considerations for the systematic literature
review as discussed, this is followed by a discussion of the reliability and validity of the semi-
structured interviews and the experiment.
3.2.1 Reliability and validity of systematic literature review
To ensure the reliability and the reproducibility of the literature review process, the
steps and approach used is clearly documented in the process of the systematic literature
review (Paré et al., 2015). Moreover, to ensure the reliability of the analysis, the multi-coder
approach was used to interpret the existing literature. Regarding the validity of the review
process, top IS journals were the focus, using multiple databases (Webster & Watson, 2002).
Chapter 3:Research Methods 25
3.2.2 Reliability and Validity of Semi-Structured Interviews and the Experiment
Reliability and validity are important cornerstones to evaluate the quality of a research
design. Validation principles have been applied and suggested for both qualitative (Maxwell,
1992; Mertens, 2014) and quantitative research (Straub, Boudreau, & Gefen, 2004).
Researchers have reached general agreement with regards to validation principles and
processes of quantitative studies; however, when it comes to applying validation principles in
qualitative studies, researchers do not have any such agreement (Venkatesh et al., 2013).
However, the validity of both qualitative and quantitative studies are discussed in this
research based on the main validation principles; that is, internal validity, external validity,
construct validity, and reliability (Lee, 1991; Shanks, 2002; Venkatesh, Brown, & Sullivan,
2016; Yin, 2009). In general, reliability is discussed as stability and consistency of the
research data, and validity as the concerns with regards to whether the collected data in
different situations really measures what the researcher sets out to measure and leads to the
same conclusions (Shanks, 2002).
Internal Validity
Internal validity refers to the validity of the research design in quantitative studies,
which can refer to descriptive validity and credibility of the design of qualitative study
(Lincoln, Lynham, & Guba, 2011; Venkatesh et al., 2013). Internal validity concerns the
validity of assumptions about causal relationships between independent and dependent
variables (in a quantitative study) (Straub et al., 2004) and accuracy of reports of events,
objects, behaviours, and credibility of the explanations (in a qualitative study) (Maxwell,
1992).
Given the exploratory nature of the semi-structured interviews in this study, internal
validity was not considered a threat (Yin, 2009). However, to ensure the internal validity of
the semi-structured interviews for explanatory purposes, previous literature was used to
develop clear explanations for the definitions of terms to avoid confusion in understanding the
phenomena (e.g., a precise definition of a conceptual model). In addition, all interviews were
audio recorded and transcribed to be sure that all of the information that participants intended
to deliver in the interviews was collected. Finally, six pilot studies were conducted and
previous literature was used to develop a relevant protocol based on the initial propositions of
26
the research. Some of the guiding questions were adopted from Recker et al. (2010), Green et
al. (2011), Davies et al. (2006), and Fettke (2009).
The experimental research design is the most rigorous design to collect and analyse data
that highly relies on theory to design treatments, specify hypotheses, and analyse data. In
addition, a high level of control ensures that the results of the experiment are attributable to
the defined constructs and not some extraneous variables (Shadish et al., 2002). The
experimental design in the current research mainly followed the previous experimental studies
on conceptual modelling (e.g., Bera et al., 2014; Burton-Jones & Meso, 2008; Figl & Recker,
2014; Gemino & Wand, 2005; Khatri et al., 2006).
External Validity
External validity also refers to the validity of the research design in quantitative studies,
which can refer to transferability of the design of the qualitative study (Lincoln et al., 2011).
External validity concerns the generalisability of the findings over different settings and
contexts, such as variations in persons, measurement variables, and treatment variables
(Venkatesh et al., 2013).
External validity is ensured by including an inductive approach in the design phase of
semi-structured interviews (Shanks, 2002; Yin, 2009). Semi-structured interviews aim to
extend existing theories. They also aim to explore and develop a complementary knowledge
to explain the use of combinations of models in practice. In this thesis, the theory of faithful
use of conceptual model combinations was used as a foundation from which to derive the
interview protocol. More specifically, the results of ontological analysis of modelling
grammars motivated the design of the guiding questions. Moreover, to increase the
generalisability of this research, practitioners were interviewed from different countries,
different organisations, with different experience, and with different roles. The diversity of the
respondents also allowed for replication logic in the data analysis.
Generalisability is a problem in most experimental studies, which raises questions about
their relevance to practice (Moody, 2000). In this research, as with many experimental
studies, undergraduate students were used as participants. In the context of this study, students
were regarded as reasonable proxies of junior working experts. For instance, Arisholm and
Sjøberg (2004) found little difference between students and junior professionals in
Chapter 3:Research Methods 27
understanding UML diagrams and maintaining Java programs. The design of tasks and
models in the experimental study is considered to be very simple compared to those used in
practice. However, the primary strength of experimental research is internal validity, which
ensures the validity of the results of empirical evaluation of theoretical propositions (Calder et
al., 1981). Moreover, very little is actually known about how combinations of conceptual
models are used in practice. Therefore, the primary aim of this thesis was to ensure internal
validity and control potential extraneous factors that may influence the dependent variables
(Parsons & Cole, 2005).
Construct Validity
Construct validity refers to measurement validity in quantitative research and can also
be referred to as analytical validity of qualitative research (Venkatesh et al., 2013). Construct
validity concerns the validity of designed variables and measures to be able to collect relevant
information to study the phenomena of interest and whether the findings of the study fit the
data from which they are derived (Straub et al., 2004).
To ensure construct validity, multiple stakeholders were interviewed from different
backgrounds (Yin, 2009). Moreover, the constructs and factors in the research design were
derived from the literature review based on relevant theories. The exploratory nature of the
semi-structured interviews also contributed to construct validity by providing opportunities
for new or refined constructs to emerge. To ensure the validity of the findings, the established
methodologies of qualitative analysis were followed (Silverman, 2006; Urquhart, 2013).
Moreover, the collected data have been maintained in a database to provide opportunities to
reconstruct the data collection and analysis procedures.
To meet construct validity in the experiment, the previous studies, measurements, and
instruments were used to operationalise the constructs of the current study (Burton-Jones &
Meso, 2008; Figl & Recker, 2014; Gemino & Wand, 2005). To do so, the definitions of the
constructs used in this thesis are close to those of the constructs used by other authors (e.g.,
Bera et al., 2014; Burton-Jones & Meso, 2008; Cierniak, Scheiter, & Gerjets, 2009; Figl &
Recker, 2014; Recker, 2010a; Recker & Rosemann, 2010a, 2010b).
28
Reliability
Reliability refers to the quality of measurement and is a precondition for validity in
quantitative research. In qualitative research, consistency and dependability of data and
analysis are two terms that are conceptually similar to reliability in quantitative
research(Venkatesh et al., 2013). Reliability concerns the repeatability and consistency of
measurement variables (Lincoln et al., 2011; Straub et al., 2004).
To build reliability of collected data in the semi-structured interviews, the interview
protocol was developed to strengthen the reliability of the collected data. The protocol was
consistently used during interviews to collect data. Subsequent and consistent use of the
designed interview protocol for all interviews assisted in providing consistency to the
structure and conduct of data collection. This consistency contributed to the reliability of the
gathered data. Moreover, all interviews were audio-recorded and transcribed by the
researcher. All collected data from the interviews are stored in an interview database.
To test the reliability of measurement variables of the experimental study, the most
widely used statistical measure, “Cronbach’s α”, was used to test for internal consistency
(Straub et al., 2004). Moreover, previously validated measures from the literature were used
to ensure the reliability. Moreover, multi-coder procedures were followed to code the
collected data and to establish the reliability of interpreting mappings and ontological
analysis, similar to procedures described in previous studies (Recker et al., 2010).
Chapter 4:Conceptual Modelling Research in Information Systems: What We Know Now and What We Still Do Not Know 29
Part B: Studies
Chapter 4: 3 Conceptual Modelling Research in
Information Systems: What We
Know Now and What We Still Do
Not Know4
ABSTRACT
Conceptual modelling tasks involve building a representation of a system domain to
better understand the system requirements. For many years, researchers have evaluated
different aspects of conceptual modelling to determine ways to improve information systems
analysis and design. Much of this research was guided by a seminal research agenda
developed by Wand and Weber (2002), which illustrated twenty-two research opportunities
and categorised them into four main areas: conceptual modelling grammars, methods, scripts,
and contexts. We examine the extent to which research over the past two decades has
addressed the identified research opportunities and answered the question of “How can we
model the world to better facilitate our developing, implementing, using, and maintaining
more valuable information systems?”. Our findings from a review of the literature show a
dialectic: several of the opportunities noted in 2002 have been addressed substantially; other
have been entirely neglected. We also found several path-breaking studies that did address
3 This study was conducted in collaboration with co-authors. The conceptualisation of the work and its
realisation was primarily the responsibility of the principal author. The principal author wrote the main parts of
the manuscript, compiled the literature sources, and analysed and interpreted the data. The co-authors assisted
with data interpretation, checked the references, and edited and wrote parts of the manuscript.
4 Conceptual Modelling Research in Information Systems: What we know now and what we still do not know.
To be submitted to Information Systems Research.
A short version of the paper was published as: Jabbari Sabegh, M. A., Lukyanenko, R., Recker, J. C., Samuel,
B., & Castellanos, A. (2017). Conceptual modeling research in information systems: What we now know and
what we still do not know. Paper presented at the Symposium on Research in Systems Analysis and Design (AIS
SIGSAND 2017).
30
problems not noted by the initial framework. To stimulate a new, relevant wave of conceptual
modelling research, we provide a new framework that draws the attention of conceptual
modelling research to the interplay between digital representations and outcomes.
Keywords: Conceptual modelling, research opportunities, literature review, research
agenda
4.1 INTRODUCTION
The importance of conceptual modelling to the analysis and design of information
systems has long been noted (Burton-Jones & Meso, 2006). High-quality conceptual models
play a significant role in the early detection and correction of systems development errors
(Wand & Weber, 2002), which is important given the high failure rates in information
systems projects (Gemino & Wand, 2004).
Conceptual modelling, while a core activity since the early days of the information
systems discipline (Bubenko, 1979; Lyytinen, 1987; Olle, Sol, & Verrijn-Stuart, 1982), has
long been regarded as a niche topic of interest, largely only to the community of scholars
interested in systems analysis and design. Moreover, much of the research was long
characterised as theoretical, of little practical relevance, and with limited empirical support
(Kung & Sølvberg, 1986; Moody, 2005; Siau & Rossi, 2011; Wand & Weber, 1989).
Researchers have suggested research frameworks and research agendas to promote
further research on conceptual modelling. Examples of these works include: ontological
foundations of conceptual modelling (Guizzardi, 2005; Hirschheim, Klein, & Lyytinen, 1994;
Wand & Weber, 1990b), frameworks to evaluate the quality of conceptual models (Lindland,
Sindre, & Solvberg, 1994; Mendling, Reijers, & Cardoso, 2007; Moody, 2005; Recker, 2007),
current issues and challenges in conceptual modelling (Indulska, Recker, Rosemann, &
Green, 2009; Recker, 2015), and the framework that defines the constructs of conceptual
modelling (Wand & Weber, 2002). Among these research frameworks, the current study
focuses on the research framework and agenda proposed by Wand and Weber (2002) to
answer the question regarding how to create high-quality conceptual models to better
facilitate developing, implementing, using, and maintaining more valuable information
systems. Wand and Weber’s (2002) framework provides a more comprehensive research
agenda concerning the foundations, quality, understandability, usability, and context of
Chapter 4:Conceptual Modelling Research in Information Systems: What We Know Now and What We Still Do Not Know 31
conceptual modelling. Since then, a large share of studies on conceptual modelling guided by
this research agenda have been published in the field’s main journals, and conceptual
modelling is now widely regarded as a key topic in information systems research (Allen &
March, 2012; Siau & Rossi, 2011).
Fifteen years after its publication, we believe it is time to ask whether the extant
research addressed the opportunities highlighted by Wand and Weber (2002), and which
aspects of conceptual modelling require further investigation. In this paper, we therefore
pursue a two-fold objective; first, we examine the published research on conceptual modelling
since the publication of the Wand and Weber (2002) paper. Our objective is to synthesise
relevant studies on conceptual modelling that in our view contribute to and shape
understanding of the conceptual modelling discipline, and then identify the remaining gaps in
the field that require further investigation. Second, we also ask whether the framework by
Wand and Weber (2002) remains relevant to this day, or whether a new, innovative agenda
should be set. In addressing both objectives, our paper will provide a complete perspective of
conceptual modelling research, as well as substantive directions for future research.
The rest of the paper proceeds as follows. We begin by explaining conceptual modelling
and the research framework presented by Wand and Weber (2002). The methodology that we
followed to identify relevant studies and the procedure of reviewing them will then be
explained in the research methodology section. In section 4.4, we analyse the literature by
explaining research opportunities for conceptual modelling grammars, methods, scripts and
context and reviewing the studies that have been undertaken about each of them. We then
outline the research areas in which there is a lack of study and propose a revised research
agenda. Finally, we conclude with the findings of this study.
4.2 BACKGROUND
Conceptual models are typically graphical representations of a real-world domain used
to assist analysts to develop an understanding of the system domain; facilitate communication
among different stakeholders to better understand system requirements; document the system
for maintenance, operation, and training purposes; and provide input for the design process
(Fettke, 2009; Wand & Weber, 2002). Conceptual models are often employed to represent
either dynamic aspects of a real-world domain (e.g., processes and events), or static aspects of
32
a domain (e.g., things and properties of things), or both (Burton-Jones & Weber, 2014).
Figure 4.1 provides some examples of different conceptual models developed using, from left
to right, the BPMN grammar, EPC grammar, ERD grammar, and UML grammar.
Check the order
Issue bill
Check quality
Ship the product
No
t i
n g
oo
d q
ua
lity
or d
am
ag
ed
Check the order
V
Issue billXOR
Check the product
XOR
V
Start shipping
Not good quality or damaged
Good quality
Orders
-customerName
-dateCreated-dateShipped
-orderID
Customer
-address-customerName
User
-password-UserID
Administrator
-adminInfo-adminName
-email-contactdetails-shippinginfo
1..1
0..*
Figure 4.1 Examples of different scripts developed using four conceptual modelling grammars
Both the conceptual modelling process and the models as the products of the process of
conceptual modelling are fundamental means in developing and analysing information
systems (Wyssusek, 2006). As a result, research in this area has focused on different aspects
of conceptual modelling to improve the effectiveness and efficiency of the conceptual
modelling process and products generated through it. For example, part of this research
tradition has focused on the development of new grammars and methods (Oei, Van Hemmen,
Falkenberg, & Brinkkemper, 1992; Wand, Woo, & Wand, 2008), while others have studied
individual factors (Recker, 2010a; Topi & Ramesh, 2002) or matters related to cognition and
perception (Evermann, 2005; Veres & Mansson, 2005).
In a seminal paper that structured and guided research on conceptual modelling, Wand
and Weber (2002) delineated four main concepts relevant to conceptual modelling research
(Figure 4.2):
Chapter 4:Conceptual Modelling Research in Information Systems: What We Know Now and What We Still Do Not Know 33
Figure 4. 2 Framework for Research on Conceptual Modelling (Wand & Weber, 2002, p. 364)
Conceptual modelling grammars are sets of constructs and rules that demonstrate
how constructs should be combined together to model a real-world domain. For
instance, the entity-relationship modelling grammar provides constructs such as:
“entity” and “relationship”. The grammar’s rules indicate that two entities can be
associated only via a relationship.
Conceptual modelling methods explain and rationalize how a modelling grammar
can be used. Methods explain, for example, how domain observations can be
mapped to the grammar’s constructs or which instances of phenomena are relevant
to the model (Wand & Weber, 2002, p. 364).
Conceptual modelling scripts are the products of using conceptual modelling
grammars. Conceptual modelling scripts are created by using modelling grammars
based on various conceptual modelling methods. For example, process models are
created with process modelling grammars to graphically represent relevant aspects
of business processes.
Conceptual modelling context refers to the setting in which conceptual models are
created and in which they are employed for specific task goals. Wand and Weber
(2002) proposed three contextual factors: individual difference factors, which
include the matters relating to modellers and model users such as training,
experience, or cognitive ability; task factors, which describe different tasks in
which scripts and grammars are employed; and social agenda factors, which
Conceptual Modelling-Method
Conceptual-
Modelling
Grammar
Conceptual-
Modelling
Scripts
Conceptual-Modelling
Context
Individual Difference Factors
Task Factors
Social Agenda Factors
34
describe the broader setting and organisational change in which grammars are
employed.
Using these four main areas, Wand and Weber (2002) proposed several research
opportunities for each of the four main areas of conceptual modelling research. Table 4.1
briefly summarises these research opportunities.
Table 4. 1 Research Opportunities in Conceptual Modelling (Wand & Weber, 2002, p. 373)
Research
Framework Code Research Opportunities
Conceptual
Modelling
Grammars
G1 Evaluating ontologies based on empirical testing of their predictions.
G2 Evaluating grammars for ontological expressiveness.
G3 Assigning ontological meaning to constructs of design grammars and generating
ontologically motivated modelling rules.
G4 Resolving outstanding ontological problems that impact conceptual modelling (e.g.,
nature of the part-of relationship).
G5 Empirically testing predicted strengths and weaknesses in new and existing grammars
based on their ontological expressiveness.
G6 Determining which combinations of grammars best support users who undertake conceptual modelling work.
G7 Empirically testing the predicted implications of construct deficit and overload in
grammars.
Conceptual
Modelling
Methods
M1 Evaluating how well different methods allow users to elicit and model critical domain
knowledge.
M2 Developing procedures to assist users of a grammar in identifying and classifying
phenomena according to the grammar's constructs.
M3 Determining the beliefs and values that underlie different methods and evaluating the
consequences of these beliefs and values for practice.
Conceptual Modelling
Scripts
S1 Evaluating competing scripts generated via the same grammar to describe some
phenomenon.
S2 Evaluating competing scripts generated via different grammars to describe the same
phenomenon.
S3 Evaluating different combinations of scripts to determine which combination best
supports the task at hand.
S4 Developing theory to predict and understand how humans use scripts to accomplish
various tasks.
Conceptual
Modelling
Context
Individual
differences
Task
Social
agenda
CI1 Development of knowledge-based tools to support conceptual modelling.
CI2 Predicting which cognitive and personality variables bear on a user's ability to undertake
conceptual modelling work.
CI3 Predicting and testing empirically which social skills affect the outcomes of conceptual modelling tasks.
CT Evaluating the strengths and weaknesses of conceptual modelling grammars, methods,
and scripts in the context of different tasks.
CS1 Understanding which values and beliefs underlie conceptual modelling work in practice.
CS2 Determining the costs and benefits of adopting different values and beliefs when
undertaking conceptual modelling work.
CS3 Articulating detailed conceptual modelling procedures that are congruent with different
beliefs and values.
CS4 Understanding how existing conceptual modelling grammars and methods facilitate
conceptual modelling work under different values and beliefs.
Chapter 4:Conceptual Modelling Research in Information Systems: What We Know Now and What We Still Do Not Know 35
Since then, it would appear that their call for research was heeded – as just one
indicator, at the time of writing this manuscript, their research agenda has been cited over 780
times on Google Scholar.
4.3 LITERATURE REVIEW PROCEDURES
We wanted to ascertain whether the past fifteen years of research on conceptual
modelling has provided appropriate, sufficient, and exhaustive answers to the questions Wand
and Weber posed in 2002. To that end, our literature review method drew on several
approaches (Bandara et al., 2015; Boell & Cecez-Kecmanovic, 2015; Paré et al., 2015; Rowe,
2014; Vessey, Ramesh, & Glass, 2002; Webster & Watson, 2002). It involved four steps:
First, we selected journals for the scope of the review. Given the vastness of the
research on system analysis and design and conceptual modelling, we chose to limit our
sample of studies to those published in the AIS basket of eight journals (F. Liu & Myers,
2011) plus the Journal of Database Management. Papers published in the top IS journals
were considered appropriately representative for mainstream high-quality research in the
information systems field. To broaden the validity of our review results and to accommodate
the view of conceptual modelling as a potential niche topic, we also included studies
published in the Journal of Database Management. This journal has traditionally been one of
the leading substantive journals publishing studies on conceptual modelling.
Second, we performed a full-text search in all papers in the selected journals using
keywords such as “conceptual modelling”, “conceptual model*”, “conceptual modelling
grammar”, “ontology” and meaningful variations of these terms. With these search terms, we
retrieved 3546 papers by October 2016 (See Appendix A). We then examined the title and
abstract of each of the papers to exclude all papers not dealing with conceptual modelling
research. Through this process, the total number of papers related to conceptual modelling
research reduced the total to 133 potentially relevant papers. The vast majority of excluded
papers discussed the term “conceptual model” to refer to a theory or research framework
rather than a description of information systems for purposes of analysis and design. Two
authors then reviewed the 133 potentially relevant papers to confirm relevance to our study.
We excluded papers that defined models as computer executable procedures, programs, or
codes (e.g., Krishnan, Mukhopadhyay, & Kriebel, 2004), or used concepts of conceptual
36
modelling to define other aspects of information systems, such as the effective use of
information systems (e.g., Burton-Jones & Grange, 2013), which we also did not regard as
relevant. Through this process, we arrived at a final number of 105 relevant articles. Table 4.2
summarises the distribution of these papers.
Table 4.2 Reviewed papers
Journals Number of relevant papers
MIS Quarterly 6
Information Systems Research 9
Journal of the Association for Information Systems 17
European Journal of Information Systems 15
Journal of Management Information Systems 1
Information Systems Journal 5
Journal of Information Technology 1
Journal of Strategic Information Systems 2
Journal of Database Management 49
Total 105
Third, we developed and applied a coding scheme. This coding scheme (summarised in
Appendix A) allowed us to categorise papers alongside multiple broad dimensions: the focus
and goal of the paper (Vessey et al., 2002), prominent conceptual modelling area addressed
(building on the classifications used in Wand & Weber, 2002), research method used (W.
Chen & Hirschheim, 2004), and evidence (Atkins et al., 2004), if any.
Fourth, to ensure a reliable application of the coding scheme, one author coded all 105
papers, while a second author independently coded a random subset of 30 papers. Their inter-
coder agreement was 62%. The two authors then discussed disagreements, updated coding
criteria and instructions, and then independently revised the coding over two more rounds
until 100% agreement was reached. The first author then revised the coding of the remaining
75 articles.
4.4 ANALYSIS OF THE LITERATURE
In presenting the findings from our analysis, we discuss the identified literature on the
four main conceptual modelling categories (grammars, methods, scripts, and context) in turn.
We start by providing three broad overviews: Table 4.3 summarises the distribution of papers
by their focus on conceptual modelling areas (i.e., grammars, methods, scripts, and context).
Chapter 4:Conceptual Modelling Research in Information Systems: What We Know Now and What We Still Do Not Know 37
Table 4. 3 Research on conceptual modelling based on focus of the study
Focus Basket of 8 JDM All Papers
No. % No. % No. %
1 Grammar 5 8.93 7 18.92 12 11.43
2 Method 7 12.50 10 27.03 17 16.19
3 Script 3 5.36 6 16.22 9 8.57
4 Context 6 10.71 1 2.70 7 6.67
5 Grammar and method 2 3.57 1 2.70 3 2.86
6 Grammar and script 6 10.71 1 2.70 7 6.67
7 Grammar and context 2 3.57 2 5.41 4 3.81
8 Method and script 7 12.50 3 8.11 10 9.52
9 Method and context 4 7.14 3 8.11 7 6.67
10 Script and context 2 3.57 1 2.70 3 2.86
11 Grammar, method, and script 0 0.00 1 2.70 1 0.95
12 Grammar, method, and context 0 0.00 0 0.00 0 0.00
13 Method, script, and context 0 0.00 0 0.00 0 0.00
14 Grammar, script, and context 3 5.36 1 2.70 4 3.81
15 Grammar, method, script, and context 0 0.00 0 0.00 0 0.00
16 Other 9 16.07 12 32.43 21 20.00
Total 56 - 49 - 105 100
Table 4.4 summarises the papers by type of empirical evidence collected, if any. Table
4.5 summarises the most popular conceptual modelling grammars examined in the literature.
Table 4. 4 Number of papers per type of evidence
Grammar Method Scripts Context Other
Research Method # % # % # % # % # % # %
Empirical 47 44.76 18 58.06 23 60.53 27 79.41 17 68.00 3 14.29
Case study 5 4.76 2 6.45 2 5.26 1 2.941 2 8.00 1 4.76
Action research 1 0.95 0 0.00 1 2.63 0 0 0 0.00 1 4.76
Survey 3 2.86 3 9.68 0 0.00 0 0 1 4.00 0 0.00
Experiment 31 29.52 13 41.94 15 39.47 24 70.59 9 36.00 1 4.76
Interview 4 3.81 1 3.23 2 5.26 0 0 4 16.00 0 0.00
Design science 4 3.81 0 0.00 3 7.89 1 2.941 1 4.00 0 0.00
Leveraging
previous data 0 0.00 0 0.00 0 0.00 1 2.941 0 0.00 0 0.00
Prototype 1 0.95 0.00 1 2.63 0 0 1 4.00 0 0.00
None-Empirical 58 55.24 13 41.94 15 39.47 7 20.59 8 32.00 18 85.71
Literature review 5 4.76 0 0.00 0 0.00 2 5.882 2 8.00 1 4.76
Commentary 7 6.67 3 9.68 0 0.00 0 0 2 8.00 1 4.76
Theoretical 46 43.81 10 32.26 15 39.47 5 14.71 4 16.00 16 76.19
38
Table 4. 5 Number of papers per type of model
Grammar # Grammar #
UML 7 ISO TC87 1
UML Class 12 Merise 1
UML Use Case 5 ebXML 1
UML Activity 5 BPML 1
UML State Machine 2 WSCL 1
UML Sequence 2 WS-BPEL 1
UML Collaboration 1 DEMO 1
ER 22 ProH 1
BPMN 7 REA 1
EPC 6 ORM 1
EER 4 IFO 1
Petrinets 4 FDM 1
Workflow 3 SDM 1
ANSI Flowchart 2 NIAM 1
DFD 1 OMT 1
YAWL 1 OML 1
MibML 1 DFRD 1
As presented in these tables, 46 percent of reviewed studies were published in the
Journal of Database Management, followed by the Journal of the Association for Information
Systems with 16% of the reviewed papers. The notable theme of the reviewed articles is that
more than 37% of the research effort concentrated on more than one area of conceptual
modelling (e.g., grammar and script or method and script). UML and Entity-Relationship
(ER) grammars were the most popular grammars investigated in the reviewed papers.
4.4.1 Conceptual Modelling Grammars
In all, 31 studies of the reviewed articles primarily investigated conceptual modelling
grammars. The types of papers included 20 empirical studies, two commentary papers, and
nine theoretical studies. The choice of research methodologies for the empirical studies
included 13 experimental studies, three industrial surveys, two interviews, one case study, and
one case study with experiment.
Identified research that addressed opportunities proposed by Wand and Weber (2002)
Unlike the early research on conceptual modelling grammars (Oei et al., 1992), the
research on conceptual modelling grammars since 2002 apparently shifted its focus from
developing new grammars to evaluating and improving existing grammars. The common
Chapter 4:Conceptual Modelling Research in Information Systems: What We Know Now and What We Still Do Not Know 39
theme emerging from these studies is that studies on conceptual modelling grammars mostly
used Wand and Weber’s (1990a, 1993, 1995) representation theory (Wand & Weber, 1990a,
1993, 1995) to evaluate conceptual modelling grammars and the implications of ontological
deficits of the modelling grammars, with the exception of one study, which used an
alternative ontological benchmark to evaluate modelling grammars (Milton & Kazmierczak,
2004). Table 4.6 provides a summary of these studies based on their main focus on conceptual
modelling grammars.
One of the central findings from this body of research on the evaluation of modelling
grammars is that any one grammar suffers from some level of construct deficit (Irwin & Turk,
2005; Recker et al., 2010; Recker, Rosemann, Indulska, & Green, 2009; Hong Zhang,
Kishore, & Ramesh, 2007). For example, Irwin and Turk (2005) indicated that use case
modelling grammars are ontologically incomplete due to construct deficits in representing the
system structure and decomposition.
Table 4. 6 Papers on conceptual modelling grammars and type of evidence5
Code Focus #
Type of
Evidence #
G1 Evaluating ontologies based on empirical testing of their predictions. 0
G2 Evaluating grammars for ontological expressiveness. 6 Interview 1
Theoretical 5
G3
Assigning ontological meaning to constructs of design grammars and
generating ontologically motivated modelling rules. 8
Case Study 2
Experiment 4
Commentary 1
Theoretical 2
G4
Resolving outstanding ontological problems that impact conceptual
modelling (e.g., nature of the part-of relationship). 2
Experiment 1
Theoretical 1
G5
Empirically testing predicted strengths and weaknesses in new and existing
grammars based on their ontological expressiveness. 0
G6
Determining which combination of grammars best support users who
undertake conceptual modelling work. 0
G7 Empirically testing the predicted implications of construct deficit and
overload in grammars. 10
Survey 1
Experiment 8
Interview 1
G
Other Other 7
Survey 2
Experiment 1
Commentary 1
Theoretical 3
5 Note that one paper contained more than one type of empirical evidence and two papers focussed on more than
one research opportunity. These papers were coded in each element, respectively.
40
Studies on conceptual modelling grammars have also focused on understanding and
improving existing grammars (Allen & March, 2006; Bera et al., 2011; Evermann & Wand,
2005; Laurier & Poels, 2012; Rittgen, 2006; Soffer & Kaner, 2011; Soffer, Kaner, & Wand,
2010). In doing so, researchers aimed to assign real-world semantics or ontological meaning
to the constructs of the grammars to facilitate their use in conceptual modelling. Eight studies
focused on assigning ontological meaning to constructs of design grammars and generating
ontologically motivated modelling rules. The dominant idea of these studies is to develop
rules to guide the modellers in creating ontologically well-formed models. The findings
indicate that diagrams created with the ontologically motivated rules lead to better domain
understanding (Bera, Krasnoperova, & Wand, 2010).
However, there are still some important real-world phenomena that remain problematic
for conceptual modelling, as they are not ontologically well-defined. For example, Shanks et
al. (2008) argued that representing the part-of construct remains a contentious issue. To
overcome this issue, the researchers suggested representing composites as entities (Shanks &
Weber, 2012).
Several studies (10 out of 31), focused on empirically testing the predicted implications
of construct deficit and overload in grammars. The main argument of these studies is that the
existence of construct overload in conceptual modelling grammars decreases the clarity of
conceptual modelling and therefore decreases the understandability and usefulness of the
designed models (Shanks et al., 2010; Shanks et al., 2008).
However, there are also some contrary arguments. Bowen et al. (2009) conducted an
experiment to evaluate the relationship between the level of model expressiveness and
ontological clarity on user performance during query development. The results indicate that
participants using the implementation that exhibited greater expressiveness via greater
ontological clarity made significantly more semantic errors, took significantly more time to
compose their queries, and were significantly less confident in the accuracy of their queries in
contrast to the findings of Bowen et al. (2006), Parsons (2011), and Shanks et al. (2010).
Likewise, Allen and March (2012) also found that construct overload does not have a
significant effect on subjects’ task performance.
Chapter 4:Conceptual Modelling Research in Information Systems: What We Know Now and What We Still Do Not Know 41
Identified opportunities in addition to those proposed by Wand and Weber (2002)
Through our review of studies across conceptual modelling grammars, we found seven
studies that did not fall into any of the grammar-related research opportunities proposed by
Wand and Weber (2002). Three main themes emerged from these studies. First, the
researchers highlighted the importance of factors other than ontological elements in
evaluating conceptual modelling grammars. These included semiotic and perceptual
discriminability of the grammatical constructs (Figl, Mendling, et al., 2013), and logical
qualities of conceptual modelling grammars (Clarke et al., 2016). Second, apart from
evaluating the capabilities of conceptual modelling grammars, researchers investigated what
other factors affect the usage behaviour of conceptual modelling grammars. The results from
surveying practitioners showed that the decision about which type of modelling grammar to
use depends on the purpose of the task in hand (Dobing & Parsons, 2008), and the willingness
to continue using the selected grammars depends on various factors, including the modeller’s
experience in modelling, background, and familiarity with the modelling grammar (Recker,
2010a). Third, researchers highlighted the complexity of modelling grammars and difficulties
in learning how to use the grammars. For example, VanderMeer and Dutta (2009) focused on
the unified modelling language (UML) and more specifically on UML sequence diagram and
proposed guidelines based on cognitive theories and the theory of learner-centred design to
decrease the complexity of learning how to build UML sequence diagrams.
Opportunities proposed by Wand and Weber (2002) that have not been addressed
Some of the research opportunities related to conceptual modelling grammars have not
yet been answered, for example, evaluating different ontologies based on empirical evidence.
Weber (2003) argued that some other alternative ontological theories can be used to improve
the understandability of modelling techniques or to predict different sorts of conceptual
modelling phenomena. One of the arguments about ontologies is that the selected ontology
should be understandable and applicable for selected modelling techniques, the purpose of
modelling and the types of users (Green & Rosemann, 2004).
As noted above, we found only one study that replaced Bunge’s (1977) ontology with
Chisholm’s (1996) ontology to evaluate data modelling languages (Milton & Kazmierczak,
2004). However, we couldn't find any empirical evidence of evaluating Bunge’s (1977)
ontology in comparison with Chisholm’s (1996) ontology.
42
Another open research opportunity proposed by Wand and Weber (2002) concerns
empirically testing the impacts of potential weaknesses or strengths of conceptual modelling
grammars on users of scripts created by the grammars, as well as the usefulness of
ontologically based modelling rules.
Another important question that remains is which combinations of grammars best
support users who undertake conceptual modelling work. While there has been some work
regarding evaluating the ontological capabilities of multiple modelling grammars (Green et
al., 2007; Green et al., 2011), there is still no empirical evidence regarding which
combinations of grammars best support users in their design and analysis tasks even though
practical observations clearly indicate that IS professionals normally use multiple modelling
grammars in their design and analysis tasks (Dobing & Parsons, 2008).
4.4.2 Conceptual Modelling Methods
In all, 38 studies addressed research opportunities on conceptual modelling methods.
Studies in this area were primarily empirical, with 13 experimental studies, three design
science research, two practitioners’ interviews, one case study, one action research, one
prototype, and two studies that provided multiple empirical evidence using experiment and
interviews and case study with experiment. Fifteen studies chose a theoretical approach and
proposed methods and guidelines to improve conceptual modelling.
Identified research that addressed opportunities proposed by Wand and Weber (2002)
Research on conceptual modelling methods mostly focused on developing procedures
for using grammars and identifying instances of phenomena to model with the grammar.
Table 4.7 provides a summary of these studies.
Out of 38 studies on conceptual modelling method, nine papers investigated how
different methods lead to better conceptual modelling practices. Together, this research
indicates that methods such as the Resource-Event-Agent modelling patterns (Poels, 2011;
Poels, Maes, Gailly, & Paemeleire, 2011), ontologically motivated rules in modelling (Bera,
2012), reuse of analysis patterns augmented with learning mechanisms (Purao, Storey, & Han,
2003), or diagrammatic attention management (Dunn, Gerard, & Grabski, 2011) can lead to
improved conceptual modelling results, such that, for example, models developed following
these methods are easier to understand for users. Siau and Rossi (2011) reviewed evaluation
Chapter 4:Conceptual Modelling Research in Information Systems: What We Know Now and What We Still Do Not Know 43
techniques for conceptual modelling methods and argued that different evaluation techniques
fit different purposes.
Table 4. 7 Papers on conceptual modelling method and type of evidence6
Code Focus # Type of
Evidence #
M1 Evaluating how well different methods allow users to elicit and model
critical domain knowledge. 9
Experiment 7
Action Research 1
Theoretical 1
M2 Developing procedures to assist users of a grammar in identifying and
classifying phenomena according to the grammar's constructs. 19
Experiment 4
Interview 2
Design Science 3
Proto Type 1
Theoretical 10
M3 Determining the beliefs and values that underlie different methods and
evaluating the consequences of these beliefs and values for practice. 1 Interview 1
M
Other Other 10
Case Study 2
Experiment 5
Theoretical 4
A second noted main focus was research on conceptual modelling methods that
developed procedures to assist users of grammars in identifying and classifying phenomena
according to the grammars’ constructs; such as developing cognitive principals to guide
classification (Parsons & Wand, 2008). The dominant idea emerging from research in this
category was that cognitive principles and ontological guidelines can assist users of a
grammar to identify and classify phenomena according to the grammar’s constructs (Bera,
2012; Bera et al., 2010; Evermann, 2005; Hadar & Soffer, 2006; Parsons & Wand, 2013).
A third dominant focus was the phenomena of diversity in conceptual modelling
(Ågerfalk & Eriksson, 2004). Researchers argued that using an ontology-based modelling
framework can help modellers to reduce variety in developed models (Hadar & Soffer, 2006).
Identified opportunities in addition to those proposed by Wand and Weber (2002)
Several scholars went beyond the research opportunities suggested by Wand and Weber
(2002) and suggested other methods aimed at improving conceptual modelling. One foci of
this work was the question of whether ontological guidelines, per se, can sufficiently cover
the problems of conceptual modelling. Clarke et al. (2016) proposed a logical perspective to
6 Note that two papers contained more than one type of empirical evidence and one paper focused on more than
one research opportunity. These papers were coded in each element, respectively.
44
improve the semantics of conceptual modelling and suggested using logical principles
together with ontological principles.
Researchers highlighted the classification of phenomena according to the grammars’
constructs and information loss when model users use developed class-based models. An
approach to overcome the effect of classification on information loss was proposed by
Lukyanenko, Parsons, and Wiersma (2014). The proposed approached signified the use of the
instance-and-attribute based modelling approach in developing models.
Another emergent dominant idea was to represent temporal dynamics of an application
using the constructs of a structural model. Allen and March (2003) argued that temporal
dynamics are semantic rather than structural and that the existing constructs in the ER model
are sufficient to represent them. The proposed method was based on representing events as
entities. Methods of mapping and transforming the captured requirements in conceptual
models to relational database and platforms were other areas highlighted by the reviewed
studies. An, Hu, and Song (2010) offered Round-Trip Engineering to maintain the
consistency of conceptual-relational mappings. Pardillo, Mazón, and Trujillo (2011) proposed
an approach to automate transforming the captured requirements into the data-warehousing
platform following model driven architecture and QVT (query/view/transformation)
transformations, and Rittgen (2006) suggested a language mapping framework to transform
from/between UML and DEMO. With regards to automated database design, Storey,
Goldstein, and Ding (2002) suggested using common-sense knowledge to improve the quality
of the designed database.
While research on conceptual modelling methods proposed various approaches to
improve conceptual modelling, Batra (2008) asked a question regarding which pattern should
be used in different circumstances. To answer this question, Batra (2008) evaluated eleven
data modelling patterns suggested in textbooks. The findings indicated that the use and
frequency of patterns were different based on the type of audiences, for example, practitioners
vs. students. The suggestion of the study was to use a combination of patterns to improve
effective training and use strategy.
Chapter 4:Conceptual Modelling Research in Information Systems: What We Know Now and What We Still Do Not Know 45
4.4.3 Conceptual Modelling Scripts
Thirty-four out of the 105 considered papers focused on conceptual modelling scripts.
Notably, most of these 34 papers also fell into other categories, such as grammars and scripts.
Table 4.8 represents the distributions of papers across research opportunities related to
modelling scripts. Out of the 34 papers on conceptual modelling scripts, 25 papers were
experimental studies, one case study, one design science, one study’s evidence was from
leveraging the findings of the previous studies, two literature reviews, and four theoretical
papers.
Table 4. 8 Research on conceptual modelling scripts and type of evidence7
Code Focus # Type of Evidence #
S1 Evaluating competing scripts generated via the same grammar to
describe some phenomenon. 21 Experiment 20
Design Science 1
S2 Evaluating competing scripts generated via different grammars to
describe the same phenomenon. 3 Experiment 2
Leveraging previous data 1
S3 Evaluating different combinations of scripts to determine which
combination best supports the task at hand. 0
S4 Developing theory to predict and understand how humans use
scripts to accomplish various tasks. 5 Experiment 3
Theoretical 2
S
Other Other 7
Case Study 1
Experiment 2
Literature Review 2
Theoretical 2
Identified research that addressed opportunities proposed by Wand and Weber (2002)
Twenty-one out of 34 papers on conceptual modelling scripts evaluated competing
scripts generated via the same grammar to describe the same phenomena; only three papers
evaluated competing scripts generated using different grammars. With most conceptual
modelling scripts being the products of using modelling grammars and methods, most of these
papers also appeared in our discussion above on conceptual modelling grammars and
methods.
The main theme that emerged from studies on scripts was that most of the researchers
compared competing scripts against the existence of construct deficits and overload and
whether assigning ontological meaning to grammatical constructs leads to better conceptual
7 Note that two papers focus on more than one research opportunity. These papers were coded in each element,
respectively.
46
modelling scripts. The second main theme of these studies was evaluating scripts developed
using different grammars or methods.
The evaluations of scripts based on ontological factors provided arguments about the
effect of construct overload on the clarity of conceptual modelling scripts. Most of the studies
supported that the existence of construct overload decreases the clarity of scripts and therefore
decreases the performance of the users (Bowen et al., 2006; Milton et al., 2012; Parsons,
2011; Shanks et al., 2010; Shanks et al., 2008). However, some studies provided contrary
arguments. For example, Allen and March (2012) did not find significant evidence that the
existence of construct overload affects users’ performance. Similarly, results of Bera et al.’s
(2014) study showed that construct overload did not affect the performance of novice and
expert users; however, models without construct overload improved the performance of users
with moderate knowledge of the domain shown in the model. In this line, findings of the
study by Bowen et al. (2009) also challenged the arguments of the usefulness of ontological
clarity on users’ performance for all task types, especially in the case of high complexity and
increase in the size of scripts (models).
Research on conceptual modelling scripts also evaluated scripts that developed using
the same grammar but following different conceptual modelling methods. Bera (2012) argued
that users of models developed using ontological rules have less difficulty in achieving
problem-solving models, while in developing the models, modellers who followed the
ontological rules also have less difficulty. Some other methods where researchers indicated
that ontological rules led to higher quality conceptual modelling scripts were classification
rules (Parsons & Wand, 2008), decomposition principles (Burton-Jones & Meso, 2006),
Resource-Event-Agent modelling patterns (Poels, 2011; Poels et al., 2011), application-based
domain modelling approach (Reinhartz-Berger & Sturm, 2008), using diagrammatic attention
management (Dunn et al., 2011), and using gateway constructs (Recker, 2013).
In contrast to evaluating different scripts developed using the same grammar, only three
studies evaluated different scripts developed using different grammars. Figl, Mendling, et al.
(2013) evaluated the effect of notational deficiencies, viz., semiotic clarity deficiencies and
perceptual discriminability deficiencies on process models understandability by comparing
scripts developed using the grammars YAWL, EPC, BPMN, and UML. Their findings
showed that the existence of both types of deficiencies negatively affected the performance of
Chapter 4:Conceptual Modelling Research in Information Systems: What We Know Now and What We Still Do Not Know 47
model users, while the performance was worse than the other scripts for users of the script
with severe notational deficiencies. Khatri et al. (2006) investigated the performance of users
of ER and EER models considering the individual differences based on participants’ IS and
domain knowledge. Their findings did not show the differences between users of EER in
comparison with ER model users.
Five studies developed theories to predict and understand how humans use scripts to
accomplish various tasks. One dominant theory argued that using high-quality information
and a different form improves users’ performance (Burton-Jones & Meso, 2008). The finding
of the study by Burton-Jones and Meso (2008) was that users of well-decomposed models in
combination with textual narratives had better performance than users without textual
narratives and users with poorly decomposed models.
Another argument was that users of models developed following an ontological
guideline had less cognitive difficulties compared to a user of models developed without
ontological guidelines (Bera, 2012). The proposed theory that emerged from the studies was
that ontological guidelines alleviate the cognitive difficulties of model users in understanding
the developed models, as well as the cognitive difficulties of modellers during model
development (Bera, 2012; Bera et al., 2010). These arguments indicate that users of models
consistent with the ontological guidelines have less difficulty in understanding and achieving
task goals.
In developing theories on how practitioners use scripts to achieve tasks goals, Parsons
and Wand (2013) argued that the principles of conceptual modelling, such as classification,
can also be used in other disciplines. They further argued that classification principles used on
conceptual modelling can also be used in natural science. The theory argued that principles of
inference and economy guide the choice of individual classes and collections of classes.
Identified opportunities in addition to those proposed by Wand and Weber (2002)
We found seven papers that did not fall into any categories on conceptual modelling
scripts. Two main ideas emerged from these studies: first, quality measures that relate to
conceptual modelling scripts; second, the use of different types of additional information in
support of conceptual modelling scripts.
48
In terms of quality measures, our above review shows that researchers mainly used
ontological expressiveness, semantic quality, and notational discriminability to evaluate
conceptual modelling scripts. In addition, Burton-Jones et al. (2009) suggested guidelines and
a framework to evaluate conceptual modelling through experiments; Siau (2004) used human
information processing theory and cognitive science to propose an approach for evaluating
different scripts considering their informational and computational equivalence. Krogstie et
al. (2006) developed a quality framework for process modelling to ensure that developed
models fulfil the needs of models users and the objectives that underlie the modelling project.
Bandara, Gable, and Rosemann (2005) proposed that there are contextual factors, such as
project-specific factors, together with modelling related factors that lead to effective process
modelling and offer successful measures for conceptual modelling in general.
In terms of additional information, two studies suggested that using textual narratives
together with conceptual models improves domain understanding (Burton-Jones & Meso,
2008; Gemino & Parker, 2009). This line of research has been based on the Multimedia
Theory of Learning (Mayer, 2005), and reaffirms the importance of models and texts in
developing understanding.
Opportunities proposed by Wand and Weber (2002) that have not been addressed
The most notable unanswered opportunities concern the lack of empirical investigation
into the use of multiple models. To date, no research has been conducted into combining and
evaluating multiple scripts to best support analysis and design tasks.
4.4.4 Conceptual Modelling Context
The papers on conceptual modelling context fell into three main sub-categories, viz.,
individual factors, task, and social agenda factors, as summarised in Table 4.9. In all, 25
papers fell into the conceptual modelling context, out of these 25 papers, 22 papers focused on
individual factors, eight papers focused on task-related factors, and five papers focused on
social agenda factors.8
8 Papers could address more than one sub-category, and some papers focussed on more than one research
opportunity in each sub-category.
Chapter 4:Conceptual Modelling Research in Information Systems: What We Know Now and What We Still Do Not Know 49
Table 4. 9 Papers in conceptual modelling context and type of evidence9
Code Focus #
Type of
Evidence #
Co
nce
ptu
al M
odel
ling C
onte
xt
Indiv
idual
Dif
fere
nce
s
CI1 Development of knowledge-based tools to support conceptual
modelling. 5
Experiment 2
Case Study 1
Developing
Prototype 1
Design
Science 1
CI2 Predicting which cognitive and personality variables bear on a user's ability to undertake conceptual-
modelling work. 15
Experiment 9
Interview 2
Survey 1
Literature
review 1
Theoretical 3
CI3 Predicting and testing empirically which social skills affect the
outcomes of conceptual modelling tasks. 1
Literature
Review 1
Other Other 1 Case Study 1
Tas
k
CT Evaluating the strengths and weaknesses of conceptual modelling
grammars, methods, and scripts in the context of different tasks. 4
Survey 1
Experiment 1
Interview 1
Theoretical 1
Other Other 5
Case study 1
Interview 2
Experiment 1
Literature
Review 1
Theoretical 1
So
cial
Ag
enda
CS1 Understanding which values and beliefs underlie conceptual
modelling work in practice. 3 Interview 3
CS2 Determining the costs and benefits of adopting different values and beliefs when undertaking conceptual modelling work.
0
CS3 Articulating detailed conceptual modelling procedures that are
congruent with different beliefs and values. 0
CS4
Understanding how existing conceptual modelling grammars and
methods facilitate conceptual modelling work under different
values and beliefs.
0
Other Other 2 Experiment 1
Theoretical 1
Individual Differences Factors – Identified research that addressed opportunities
proposed by Wand and Weber (2002)
Human factors in conceptual modelling have gained widespread attention since 2002. In
this area of research, we reviewed previous literature to evaluate how well previous studies
covered research areas proposed in 2002.
9 Note that one paper contains more than one type of empirical evidence. This paper was coded in each element,
respectively.
50
First, five out of 22 studies on individual differences factors in conceptual modelling
focused on developing knowledge-based tools to support conceptual modelling. The main
theme arising from this stream of research was the importance of the use of collected and
learned knowledge in conceptual modelling. For instance, Purao et al. (2003) developed a
prototype to emulate the learning that occurs over time in analysis patterns. Storey et al.
(2002) developed a common-sense business reasoner containing general knowledge about the
world and a reasoning tool to apply this knowledge to a database design task. Similarly,
Koschmider, Song, and Reijers (2010) developed a recommendation-based process modelling
support system to receive recommendations through social features to improve the modelling
process, and as a result, the quality of the designed models.
Second, researchers also highlighted the importance of capturing all requirements and
the traceability of the system and argued that information systems should have the ability to
evolve as their environment changes. To achieve this objective, Loucopoulos and Kadir
(2008) developed tools to improve the traceability in information systems through the
Business Rules-Driven Object Oriented Design (BROOD) approach. In this vein of research,
and to be able to automatically transform the requirements captured in the early stages to the
entire data-warehousing platform, Pardillo et al. (2011) proposed an model-driven
architecture approach and query/view/transformation (QVT) transformations and developed
CASE support tools that support the approach.
The third important research opportunity was predicting which cognitive and
personality variables bear on a user's ability to undertake conceptual modelling work (Browne
& Parsons, 2012; Davern, Shaft, & Te'eni, 2012a, 2012b). Most of the research we reviewed
on individual differences factors in conceptual modelling fell into this category. The findings
of this research identified various factors such as domain knowledge (Bera et al., 2011; Khatri
et al., 2006), background, experience, modelling familiarity (Recker, 2010a; Recker et al.,
2010), cognitive load (Evermann, 2005; Figl, Mendling, et al., 2013), learning behaviour and
capability (Soffer & Hadar, 2007; VanderMeer & Dutta, 2009), understandability and
applicability of the selected ontology for IS professionals (Green & Rosemann, 2004), as well
as familiarity and cognitive load (Masri et al., 2008; Shanks et al., 2010).
The fourth main area of research was to predict and empirically test which social skills
affect the outcomes of conceptual modelling tasks. We only found one review paper on
Chapter 4:Conceptual Modelling Research in Information Systems: What We Know Now and What We Still Do Not Know 51
human factors on conceptual modelling (Topi & Ramesh, 2002), which only discussed studies
published prior to 2002. The finding of this research indicates that researchers mostly focused
on issues relevant to communications between analysts and developers, while the issues
important to communication between analysts and users have apparently received little
attention.
Individual Differences Factors - Identified opportunities in addition to those proposed
by Wand and Weber (2002)
One relevant study was conducted on the conceptual modelling project level (Bandara et
al., 2005). These researchers argued that it is not only the issues and human factors of
analysts, developers, and users that are important, support from the managerial level also
plays an important role in the success of conceptual modelling.
Task Factors – Identified research that addressed opportunities proposed by Wand
and Weber (2002)
We reviewed eight papers that focused on task factors on conceptual modelling. One
paper used more than one empirical method. The main arguments of these articles were that
the differences in task settings (Recker, 2010a), the purpose of conceptual modelling (Green
& Rosemann, 2004; Recker et al., 2010), and different stakeholders involved in conceptual
modelling (Green & Rosemann, 2004) play an important role in the conceptual modelling
process. By contrast, evaluating the strengths and weaknesses of conceptual modelling
grammars, methods, and scripts in the context of different tasks was not the main focus of any
of the reviewed studies (Wand and Weber 2002).
Task Factors – Identified opportunities in addition to those proposed by Wand and
Weber (2002)
Several researchers identified the availability of tools for different tasks as an important
factor in conceptual modelling (Bandara et al., 2005; Recker, 2012). Other important task-
related factors identified in conceptual modelling were domain tangibility (Soffer & Hadar,
2007), the modelling grammar choice in dependence of a task (Bandara et al., 2005), and task
complexity in general (VanderMeer & Dutta, 2009).
52
Social Agenda Factors – Identified research that addressed opportunities proposed by
Wand and Weber (2002)
Only three studies focused on understanding the values and beliefs that underlie
conceptual modelling work in practice. One of the main arguments emerging from the studies
on conceptual modelling context was that the definitions of success may differ by the unit of
analysis (e.g., developer, project, organisation) and that the relationship among these
definitions is complex because a large number of variables affect the success of conceptual
modelling (Larsen, Niederman, Limayem, & Chan, 2009). In terms of the values and beliefs
impacting the outcome of conceptual modelling tasks that have been studied, Hadar and
Soffer (2006) argued that an individual’s perception and interpretation of reality, professional
experience, and perception of model quality may influence an individual’s performance.
Another study revealed that modelling conventions play an important role in the process of
conceptual modelling (Recker et al., 2010).
Social Agenda Factors – Identified opportunities in addition to those proposed by
Wand and Weber (2002)
We found two studies that examined opportunities in addition to those proposed by
Wand and Weber (2002). The first emerging idea is to use knowledge from social networks in
order to improve the quality of conceptual models. Koschmider et al. (2010) developed a
recommendation-based process modelling system that receives recommendations on the basis
of the previous usage of modelling fragments to complete or edit a formal business process
model.
The second emerging idea concerns environmental considerations during conceptual
modelling. Environmental consideration is one of the most important concerns in the current
decade and has gained widespread attention among researchers. He Zhang, Liu, and Li (2011)
extended goal-oriented modelling to model the rationality behind IT systems design, in
particular, how environment-related considerations come into play in such design decision
making.
Social Agenda Factors - Opportunities proposed by Wand and Weber (2002) that
have not been addressed
Not all of the research opportunities in conceptual modelling proposed by Wand and
Weber (2002) that relate to different values and beliefs have been addressed. The remaining
Chapter 4:Conceptual Modelling Research in Information Systems: What We Know Now and What We Still Do Not Know 53
research opportunities are: 1) determining the costs and benefits of adopting different values
and beliefs when undertaking conceptual modelling work, 2) articulating detailed conceptual
modelling procedures that are congruent with different beliefs and values, and 3)
understanding how existing conceptual modelling grammars and methods facilitate
conceptual modelling work under different values and beliefs.
4.4.5 Identified articles that did not match the research framework by Wand and Weber
(2002) altogether
We identified 21 papers related to conceptual modelling that did not fall into any of the
categories or opportunities suggested by Wand and Weber (2002). Sixteen of these studies
were theoretical, one was a literature review, one a commentary, and three were empirical
(one case study, experiment, and one action research). From these, we identified four main
streams in addition to the research framework proposed by Wand and Weber (2002).
A first theme emerging from the studies that did not fit in the research opportunities
proposed by Wand and Weber (2002) was multidimensional conceptual modelling. Trujillo,
Luján-Mora, and Song (2004) proposed an approach for multidimensional modelling using
UML. In this approach, the authors provided a UML-compliant class graphical notation to
conceptualise the dynamic aspects and defined rules about how to use state and interaction
diagrams to model the behavioural aspects of a system (Trujillo et al. (2004). In line with this
research, Garrigós et al. (2012) proposed personalising multidimensional models for on-line
analytical processing to overcome the complexity of understanding large models. Finally,
research by de Albuquerque and Christ (2015) proposed a modelling method to explicitly
model different aspects of a system aiming to improve organisational flexibility.
Another theme emerging from these studies was the quality of knowledge captured and
represented in knowledge engineering. The researchers highlighted the need to separate
knowledge representation into human comprehension and computational reasoning (Chua,
Storey, & Chiang, 2012).
The third main argument was about the usefulness and role of ontologies in conceptual
modelling. Fonseca and Martin (2007) believed that ontologies and conceptual schemas
belong to two different epistemic levels. While ontologies and conceptual schemas are
complementary, they represent distinct aspects of conceptual modelling. Ontologies focus on
explanation and information integration grounded in assumptions about invariant conditions
54
that define the domain of interest, whereas conceptual schemas focus on linking the general
ontological categories with particular observations to be classified in IS. On the other hand,
Weber (2003) and Wand and Weber (2004) argued the usefulness of ontology and discussed
why ontological theories can be used to inform conceptual modelling practice, research, and
pedagogy.
The fourth emerging theme concerns model-driven architecture and different aspects of
model-driven architecture, such as security features in database and process design
(D'aubeterre, Singh, & Iyer, 2008; Fernández-Medina, Trujillo, & Piattini, 2007) and software
configuration and design patterns (Dreiling, Rosemann, van der Aalst, Heuser, & Schulz,
2006; Vergara, Linero, & Moreno, 2007; Wagelaar & Van Der Straeten, 2007).
4.5 GUIDING THE NEXT WAVE OF CONCEPTUAL MODELLING RESEARCH:
A NEW FRAMEWORK
Based on our literature review, we believe that Wand and Weber’s (2002) framework
was useful and necessary at its time. It organised key aspects of conceptual modelling
research to progress and assisted in ascertaining conceptual modelling’s place as a core
research stream in IS. The volume of literature published since 2002 also suggests that the
framework served its purpose of guiding the community of researchers.
However, in our own use of the framework for research and for the purpose of this
literature review, we identified several reasons as to why we believe that a new framework
may be more suitable to guide the next wave of conceptual modelling research than simply
following-up on the outstanding research opportunities identified above. Our main reasons are
the following:
First, Wand and Weber’s (2002) framework is script-centric; it places the creation of
modelling scripts (via grammars, methods and in a context) at the core of modelling activity.
This, for example, makes it difficult to accommodate cases where the modelling activity does
not give prominence to modelling scripts.
Second, the framework is focused on supporting IS development (via modelling).
While IS development is a major part of IS, the existing framework prohibits consideration of
the use of existing IS, interaction with the data provided through an IS (e.g., business
analytics), or indeed any impacts that stem from the use of IS (i.e., outcomes).
Chapter 4:Conceptual Modelling Research in Information Systems: What We Know Now and What We Still Do Not Know 55
Third, the framework is coined by the tacit assumption that modelling is typically
undertaken by professional IS analysts, knowledgeable in appropriate methods and
grammars. Recently, however, the proliferation of content-producing technologies that may
support the creation of digital representations by ordinary people (e.g., Twitter’s hashtags),
raises questions about modelling performed by ordinary people that may be more creative and
spontaneous than the traditional process (Chang, 2010; Lukyanenko et al., 2017; V. Ramesh
& Browne, 1999).
Fourth, consistent with the decades of conceptual modelling research preceding the
framework where many modelling grammars and approaches have been proposed, the
framework emphasises the evaluation of existing grammars, potentially to the neglect of the
design of entirely novel modelling artefacts or approaches. However, the dramatic changes to
the information technology landscape for revisiting traditional design assumptions suggests
the development of novel conceptual modelling methods, grammars, and scripts. An already
debated instance in this context is the use of conceptual modelling for agile development (e.g.,
Erickson, Lyytinen, & Siau, 2005; Lukyanenko et al., 2015), to name just one example.
Fifth, the framework is technology-agnostic. With a steady availability of design
automation tools (Orlikowski, 1993) and the increasing prevalence of technologies with
inherent agency even without human interventions, the modelling of domains, existing or
future, is not necessarily a function of human conceptualisation or behaviour alone any
longer. Mining techniques that automatically construct process models from event logs are a
case in point (van der Aalst, 2011). This calls for consideration of technology support,
enablement, or even embodiment of conceptual modelling.
Sixth, the framework is static and does not explicitly consider feedback resulting from
the creation and use of modelling. This makes it difficult to accommodate multi-stage study
designs, such as action design research (Akhigbe & Lessard, 2016; Sein, Henfridsson, Purao,
Rossi, & Lindgren, 2011) involving modelling phenomena.
In summary, while the Wand and Weber (2002) framework remains reflective of
existing practice and has been useful to the academic discourse until now, it under-represents
the ever-widening spectrum of phenomena that can be supported by conceptual modelling.
56
Therefore, in what follows we propose a new framework with the objective of capturing both
traditional, as well as emerging opportunities.
Key to the new framework is the view that digital representation of reality – which lies
at the core of conceptual modelling research – is becoming a major societal force as
information technology increasingly entwines with all human activities (Leonardi, 2011).
Representations can be either formal or informal conceptualisations of user views and
information requirements, structure, and behaviour of information systems, personal, social,
and business processes, and existing information records. Representations can take the form
of diagrams (e.g., such as ER diagram), but can also include narratives, images, and other
multimedia forms. From a cognitive perspective, the representations we refer to are
considered external representations (J. Zhang, 1997); artefacts that exist outside of any one
individual’s mind and contain knowledge and structure about a domain.
As human reliance on IS for daily functions grows, people routinely reason and act
based on their perceptions of representations of reality stored in digital systems and
increasingly shun direct and traditional interactions. Floridi (2012) coined this on-going
process the “enveloping” of society by an ever-increasing digital layer. We believe conceptual
modelling research brings an important array of theories, tools, methods, and objects of
research to develop, support, and interpret modern digital representations. While
representation is a research object for many scientific disciplines (Hoyningen-Huene, 2013),
the IS conceptual modelling community has unique expertise investigating representations in
the context of information technology. We thus propose a new research agenda of
investigating representations to support the development and use of information and
information technologies. This agenda remains cognisant and incorporates all issues related to
conceptual modelling scripts, grammars, methods, and context that Wand and Weber’s
framework stipulated, but is substantially broader as it explicitly recognises the role of the
conceptual modelling community in supporting a wide range of human interactions with
information technologies. At the same time, it retains the core of the traditional framework, as
the issue of representation constituted a major part of research on conceptual modelling
scripts, grammars, and methods (Browne & Parsons, 2012; Burton-Jones & Grange, 2013;
Kent, 1978; Rai, 2017; Wand & Weber, 1995). Figure 4.3 demonstrates our view of this
framework.
Chapter 4:Conceptual Modelling Research in Information Systems: What We Know Now and What We Still Do Not Know 57
Figure 4.3 A New Research Framework to support future conceptual modelling research
To illustrate the applicability of our new framework, consider several research
directions that follow from it.
First, while Wand and Weber’s framework was script-centric, our new framework does
not insist on this emphasis, which makes it easier to accommodate emerging forms of
representations. As the digital envelope expands, much of this process is spontaneous and
highly creative, through which novel forms of representation are born. Thus, many successful
systems (e.g., Facebook, Twitter) may not implement traditional modelling (e.g., ER
diagrams) or use traditional storage technology (e.g., relational), and their success paves the
way for novel modelling paradigms (e.g., agile modelling, noSQL databases). Many of these
emerging systems explicitly proceed without a modelling script, or use modelling in a
different way (e.g., for feasibility analysis or data interpretation) (Storey & Song, 2017). The
new framework calls for investigation of novel representational approaches and for
assumptions to be made when no script is involved (Kaur & Rani, 2013; Lukyanenko &
Parsons, 2013).
Second, while it remains important to study effective and appropriate representations to
support the development of new IS, with the growth of digital content, novel needs are
emerging. Repurposing data for unanticipated insights is at the heart of the increasing growth
58
of data mining, business analytics, and applied artificial intelligence (H. Chen, Chiang, &
Storey, 2012; Rai, 2017). Here, representations remain critical; however, their role changes –
they no longer guide IS development, but are needed to integrate, visualise, and interpret
massive volumes of heterogeneous data to make informed decisions. Furthermore, different
assumptions made when assembling information for the analytics process may result in
different model performance and predictive power, and may thus result in different actions
being taken.
Third, in moving beyond the conceptual modelling scripts, our new framework enables
exciting new synergies between conceptual modelling research and other research streams
that may be affected by the assumptions behind and the quality of the representations. This
includes studies that investigate the impact of new representations by ordinary users on
information quality, effective use, adoption, and more generally, IS success (Burton-Jones &
Grange, 2013; Lukyanenko & Parsons, 2014; Lukyanenko et al., 2014).
Fourth, as our new framework does not insist on the traditional modelling process, it
supports the emerging practice of information production by ordinary people. Currently, very
little is known about these more spontaneous kinds of models paving the way to an exciting
new direction for conceptual modelling research (Lukyanenko et al., 2017; Recker, 2015).
Fifth, our new framework explicitly recognises the need for ongoing design innovation
in response to technological change. For example, the requirements of open information
environments – where controls over information production are considerably weaker than in
traditional corporate settings, motivating the search for novel approaches to conceptual
modelling that is more adaptable, flexible, and open (P. P.-S. Chen, 2006; Liddle & Embley,
2007; Parsons & Wand, 2014). Likewise, the blooming practice of machine learning and
business analytics may require new forms of representations of data.
Sixth, the new framework proposes feedback as part of a research agenda. We explicitly
recognise that antecedents could influence other antecedents. For example, ontological
assumptions could influence grammars or creator’s capabilities may influence the method
employed in appropriating a grammar. Next, outcomes can have feedback loops to other
outcomes. Using a representation for communication about a domain could lead to better
Chapter 4:Conceptual Modelling Research in Information Systems: What We Know Now and What We Still Do Not Know 59
domain understanding (Anglim, Milton, Rajapakse, & Weber, 2009; Geiger, 2010; Hoffer,
Ramesh, & Topi, 2012; Power, 2011).
Lastly, outcomes can also impact the antecedents to representation. For example, lack of
effective use, adoption, or quality could lead to a change in the creator’s capabilities as the
creator may learn or realise a better way to create future representations to mitigate the issues.
Explicit modelling of feedback in the new framework should provide impetus for more
research of this type.
Our new framework creates significant new opportunities for the conceptual modelling
community and substantially expands what counts as conceptual modelling research. It also
stands to bring different research communities that deal with digital representation (e.g.
information quality and conceptual modelling) into closer contact, promising more
opportunities for cross-pollination of ideas and interdisciplinary collaboration. Our new
framework strongly suggests that conceptual modelling research impacts a broad range of
issues related to information and information technology.
4.6 CONCLUSION
As a research field, conceptual modelling has matured into an established research area
of IS. Perhaps it is not regarded in the same manner as research on technology adoption and
the business value of technology; however, conceptual modelling stands as a cornerstone of
the research discipline.
Yet, the standing and reputation of conceptual modelling within the discipline is not
stable. As with any other field, conceptual modelling research is rightfully under constant
scrutiny in terms of its validity, applicability, relevance, and utility in the ever-changing
world. To cement the place as a research field within IS and surrounding disciplines, it will be
important to constantly review and revise our own research efforts on conceptual modelling.
To that end, in this paper we have taken two important steps. We examined the
influence and consequences of a seminal research framework in the field, and we provided a
new research framework that we believe offers a reinvigorated and exciting new perspective
on conceptual modelling research challenges and opportunities. In doing so, we have created
new pathways to research conceptual modelling that: (a) both relax and challenge our own
assumptions about what conceptual modelling is, and (b) move our research efforts towards
60
the fringes of the conceptual modelling paradigm, to areas where we are required to explore
unknown territory rather than confirm known principles. Our new framework makes an
important step in this direction by drawing attention to significant new opportunities for the
conceptual modelling community and substantially expanding our view of what counts as
conceptual modelling research. It also stands to bring different research communities that deal
with digital representations (e.g. information quality and conceptual modelling) into closer
contact, promising more opportunities for cross-pollination of ideas and interdisciplinary
collaboration. Our new framework strongly suggests that conceptual modelling research is
impacted by and impacts a broad range of issues related to information and information
technology.
In following the agenda set by our work, we may discover that conceptual modelling
has its limits. However, we will certainly increase our confidence in where, how, and why
conceptual modelling is effective and useful – and we may discover that conceptual modelling
has premises and promises that we are yet to foresee.
Chapter 5:Combined Use of Conceptual Models in Practice: An Exploratory Study 61
Chapter 5: Combined Use of Conceptual Models
in Practice: An Exploratory Study10
ABSTRACT
Conceptual models are fundamental to system analysis and design. However, the actual
usage of conceptual models remains poorly understood, in particular, how and why
practitioners would use multiple models in combination when undertaking their work. In this
paper, we explore the reported use of multiple conceptual models for system analysis and
design to determine the circumstances that lead professionals to use multiple models. We
uncover both semantic and pragmatic reasons that influence the choice and selection of
different models for system analysis and design tasks. Contrasting these findings to existing
ontological theories demonstrates that the extent and type of multiple model use is determined
not only by ontological factors, but also contextual factors that can override ontological
qualities, and in so doing bring forth desired qualities for users. We offer several novel
propositions about the implications of ontological theory that will be worth exploring in
future research.
Keywords: combined use, conceptual models, ontology, context, completeness, semi-
structured interviews, representation theory, ontological analysis, qualitative research
10 Published as: Jabbari Sabegh, M. A., & Recker, J. (2017). Combined Use of Conceptual Models in Practice:
An Exploratory Study. Journal of Database Management, 28(2), 56-88.
62
5.1 INTRODUCTION
System designers and analysts often begin their work by developing and using graphical
representations of relevant features of the domain under examination (Burton-Jones & Meso,
2006). These representations are called conceptual models (Wand & Weber, 2002). They play
a significant role in the early detection and correction of system development errors and assist
analysts to better communicate with stakeholders (Moody, 2005; Saghafi & Wand, 2014).
Much research on conceptual models and the grammars with which they are created has
been conducted over the past decades, often using ontological analysis (Wand & Weber,
1990a, 1993, 1995). Ontological analysis allows researchers to suggest how grammars for
conceptual modelling might be modified to be ontologically sound and how well developed
models that correspond to ontological guidelines are understood. There is a strong track
record of studies showing empirical support for these guidelines (Saghafi & Wand, 2014).
Yet, noticeably, this rich research tradition has focused almost entirely on the evaluation of
single grammars (such as UML, ERD, BPMN, and so forth) or single models (such as those
that are ontologically clear versus unclear).
One important insight from this stream of research is that no single available grammar
is ontologically complete (e.g., Irwin & Turk, 2005; Opdahl & Henderson-Sellers, 2002;
Recker et al., 2009; Wand & Weber, 1993; Weber, 1996). This situation implies that users
will never be able to create a single model that fully represents all of the relevant aspects of
the real-world phenomena they wish to have represented.
This situation is not necessarily problematic; in fact, incompleteness seems to be a
design choice for many popular grammars: UML, for example, provides fourteen different
grammars to describe structure, behaviour, and interactions of a system from a variety of
perspectives, each of which is, by necessity, incomplete (Rumbaugh, Jacobson, & Booch,
2004). Other longstanding methodologies, such as Multiview (Avison & Wood-Harper,
1986), have promoted the use of multiple models with different perspectives for close to thirty
years.
In this paper, we explore two questions that follow from this situation: How do analysts
and designers deal with the fact that any model they have available is not a complete
representation? How do they select from a variety of possibly available models given that
Chapter 5:Combined Use of Conceptual Models in Practice: An Exploratory Study 63
each will only offer some representation but never a complete one? The proposition that we
put forward is that they use multiple models in combination such that the completeness of the
representation of their relevant real-world phenomena can be maximised.
We are not the first to make this proposition. Theoretically, this proposition has been
explored, firstly, by Weber (1997) and Green (1996) who suggested two principles, maximum
ontological completeness and minimal ontological overlap, to explain why designers might
select different grammars for conceptual modelling. More recently, Recker (2014) suggested
theoretical arguments in a theory of faithful use of conceptual model combinations. Yet, what
remains notably absent is empirical knowledge about how practitioners work with multiple
models. We take this step in this paper and explore two broad research questions:
RQ1: Why are system analysts and designers using multiple conceptual models?
RQ2: How are system analysts and designers using multiple conceptual models?
Our contribution is primarily empirical: we set out to explore current practices in the
actual use of multiple conceptual models, which is surprisingly scarce in the literature. In
doing so, we also provide theoretical contributions: through our work, we identify a range of
contextual and contingency variables that impact on the use of conceptual models; and we
will show how these variables extend upon the understanding that can be generated through
ontological analysis alone; and in turn, how ontological theory might be improved.
We proceed as follows: First, we review prior ontological analysis and empirical studies
on conceptual modelling. The research method and how we designed our research is then
discussed. Section 5.4 reports the results in two parts, in line with our research questions.
Section 5.5 summarises the main contributions, outlines implications for the research and
practice, and reviews potential limitations of the work. Section 5.6 closes by drawing some
conclusions from our work.
5.2 BACKGROUND
Two streams of research informed our study: first, existing theoretical research on the
ontological analysis of conceptual modelling grammars and models; second, empirical
research on the use of conceptual models in practice. We discuss each, in turn below.
64
5.2.1 Ontological Analysis as a Theoretical Foundation of Conceptual Modelling
Research on conceptual modelling is an active research area in information systems
research (Burton-Jones et al., 2009). Much of this research draws on Wand and Weber’s
(1990a, 1993, 1995) representation theory as a means to evaluate, explain, and improve
conceptual modelling grammars and their usage (e.g., Gemino & Wand, 2005; Green et al.,
2011; Recker, 2010b; Recker et al., 2010).
Representation theory builds on the philosophical field of ontology (Bunge, 1977) to
inform evaluations of conceptual modelling grammars and the models (also called scripts)
that can be generated with them. The theory postulates that conceptual modellers are
concerned with building models of some relevant phenomena or domain in the real world
(Burton-Jones & Weber, 2014) and that ontological factors, such as the completeness and
clarity of a representation of real-world phenomena, can be used to predict the strengths and
weaknesses of conceptual modelling grammars and scripts, and ultimately conceptual
modelling success.
In essence, Wand and Weber’s theory suggests two main evaluation criteria: ontological
completeness and ontological clarity. Ontological completeness is measured by the degree of
construct deficit, that is, the extent to which grammatical constructs exist that allow users to
represent the entirety of the real-world phenomena they desire to be represented in a model.
Ontological clarity is measured by the degrees of construct overload (where a grammatical
construct can represent multiple aspects of a real-world phenomenon), construct redundancy
(where some aspects of a real-world phenomenon can be represented by multiple grammatical
constructs), and construct excess (where a grammatical construct does not represent any real-
world phenomena). Deficiencies in terms of ontological completeness and ontological clarity
are typically established through a procedure that has become known as ontological analysis
(Fettke & Loos, 2005; Gehlert & Esswein, 2005; Hadar & Soffer, 2006; Rosemann, Green, &
Indulska, 2004; Rosemann, Recker, Green, & Indulska, 2009; Wand & Weber, 1993; Hong
Zhang et al., 2007).
Ontological analysis based on representation theory has been used in various studies to
examine the predictions of ontological completeness and clarity on modellers’ ability to
develop clear and complete models of real-world phenomena using different modelling
grammars. Structured overviews of these studies are provided by Green et al. (2011), Burton-
Chapter 5:Combined Use of Conceptual Models in Practice: An Exploratory Study 65
Jones et al. (2009), and Recker et al. (2009), amongst others. A meta-analysis of empirical
studies to date shows that the theory is indeed able to make largely valid predictions about the
use of conceptual modelling in practice (Saghafi & Wand, 2014).
Two aspects of this stream of research are noteworthy. First, virtually all of the existing
research has been on single grammars or single models, examining questions such as whether
ontological deficiencies in a grammar lower perceptions of grammar usefulness (Recker et al.,
2011), or whether the existence of ontological deficiencies in a grammar inhibit users’ ability
to faithfully describe a particular real-world phenomenon in a model constructed with that
grammar (e.g., Bodart et al., 2001; Parsons, 2011; Shanks et al., 2008). Second, all
ontological analyses performed to date have shown that all current available grammars have
some level of construct deficit (see Table 5.1). In other words, no grammar, and therefore no
single conceptual model, is ever ontologically complete: some aspect of real-world
phenomena cannot be expressed. This state is intuitively reasonable: most grammars have
been developed with a clear purpose in mind. For example, data modelling grammars, such as
ERD, have been developed to model substance and form in the real world – but not possibility
and change (Burton-Jones & Weber, 2014). Process modelling grammars, by contrast, model
behaviour, change, and dynamics – but not things and their properties. Moreover, often not all
constructs provided via a grammar are used in any given model (Siau et al., 2005; zur
Muehlen & Recker, 2008). In other words, most models are even less complete than the
grammars with which they have been created.
Nevertheless, any information system that is being analysed or designed incorporates
both aspects: substance and form (Which data needs to be stored about objects? Which
attributes characterise these objects?), as well as dynamics and behaviour (What happens to
these objects? Which processes are supported by the system?). During analysis and design,
therefore, users are likely to desire a more complete representation than any one model can
provide. In this paper, we therefore ask: How do practitioners achieve the level of
representation they desire?
66
Table 5.1 Examples of studies on ontological completeness of modelling grammars
Reference Example of Grammars
Evaluated
Identified Construct Deficit
Irwin and Turk (2005) Use Case Modelling
Grammar
e.g., system structure, system
decomposition
Recker et al. (2010) BPMN e.g., history, system decomposition
Wand and Weber (1989, 1993);
Weber (1996)
Entity Relationship
grammar
e.g., transformation
Data Flow Diagram e.g., state law
Green and Rosemann (1999) ARIS e.g., decomposition, state law
Weber and Zhang (1996) NIAM e.g., event, history
Green et al. (2007) BPML e.g., things, state law
ebXMLBPSS e.g., kind, system decomposition
BPEL4WS e.g., history
WSCI e.g., things, state law
Recker et al. (2009) Petri net e.g., property, history
IDEF3 e.g., class, kind
Merise e.g., state, system
ISO TC87 e.g., class, state
ANSI Flowchart e.g., things, property
Opdahl and Henderson-Sellers
(2002)
UML e.g., conceivable state space, lawful event
space
5.2.2 Empirical Studies of Conceptual Modelling Practice
We reviewed the empirical literature with two goals in mind: first, is there evidence to
support the assumption that professionals use more than one model or one grammar in their
tasks? Second, if so, how do they go about achieving the level of representation they require?
We quickly realised that practitioners do indeed often use more than one conceptual
model in their design and analysis tasks (Davies et al., 2006; Dobing & Parsons, 2008; Fettke,
2009). For example, Recker (2012) reported that over 30% of surveyed process modellers
access additional grammars when modelling business processes. Green et al. (2011) showed
that 80% of users of modelling tool environments select and use multiple grammars in
combination when engaging in conceptual modelling. Dobing and Parsons (2008) reported
that 90% of UML users employ at least two different UML grammars in at least one-third of
their projects.
We also realised that the literature is scarce with regards to the reasons for how and why
users use multiple grammars or models. Weber (1997) argued that if analysts or designers
encounter ontological incompleteness, they will somehow have to augment their model(s).
One plausible strategy is to use additional grammars or models, as indicated in some studies
(Green et al., 2011; Recker, 2012). Other reported workarounds include annotations and
Chapter 5:Combined Use of Conceptual Models in Practice: An Exploratory Study 67
colour highlighting (Recker et al., 2010), the reliance on organisational conventions (Samuel
et al., 2015), or the augmentation of models with textual descriptions (Gemino & Parker,
2009).
Another finding from our review was that empirical studies on multiple conceptual
models use are scarce, albeit their results are promising. We found three empirical studies on
multiple models (Gemino & Parker, 2009; J. Kim et al., 2000; Siau & Lee, 2004). The
findings of these three studies all implied that using multiple models results in better
performance than using a single model.
Thus, our conclusion from this review is that: a) multiple model use is an empirically
established phenomenon, with b) ascribed but not conclusively demonstrated theoretical
benefits, and c) that substantive knowledge is lacking with regards to at least three questions
concerning why users would use multiple models and how:
1) How do users realise that one model may not be sufficient for the tasks they set out
to complete?
2) Can the rationale for multiple model use be linked to ontological factors, such as
the completeness and clarity of representation?
3) Which contingencies, such as conventions or other workarounds, exist that
influence the decisions users make?
We now report on our exploratory study that aims to provide some answers to these
questions.
5.3 RESEARCH METHOD
5.3.1 Design
To answer our two broad research questions, we decided on an exploratory, qualitative
research design. Specifically, we opted for semi-structured qualitative interviews. Semi-
structured interviews are a main data collection tool in qualitative research (Yin, 2009). In our
case, they enabled us to examine our phenomenon of interest from the model users’
perspective (Benbasat et al., 1987; Myers & Newman, 2007; Yin, 2009).
To ensure reliability and validity, we mainly followed and applied the guidelines of
leading positivist qualitative researchers (Shanks, 2002; Yin, 2009). We used the extant
68
literature on empirical and ontological research on conceptual modelling to build a semi-
structured protocol to guide our interviews. The protocol is available in Appendix B. We
designed the protocol such that we could obtain information on the extent of and reasons for
the use of multiple models (Dobing & Parsons, 2008; Fettke, 2009; Green et al., 2011), the
contextual factors involved in the use of the conceptual models (Recker, 2010a, 2012; Wand
& Weber, 2002), and the potential impact of ontological deficiencies on the use of multiple
models (Recker et al., 2010; Recker et al., 2011).
The use of the semi-structured interview protocol helped us to be consistent in
structuring and conducting the data collection processes. It thereby assisted us in maintaining
the reliability of the collected data. For example, it assisted us with clear terminology; in the
literature several relevant terms have various definitions (e.g., a conceptual model), as do
factors that possibly influence the use of conceptual models (such as familiarity with different
types of models, existing domain knowledge, ontological factors). Our protocol allowed us to
use relevant terms understandably and consistently. With these considerations in mind, our
protocol consisted of four parts:
1. We started by clarifying terms used during the interview, such as conceptual
models (graphical representations vs other schemas such as XML schemas or meta-
models), interpretation (human-readable models vs machine-readable models), and
others.
2. We collected demographic information and performed control checks to be sure
that the interviewees were indeed users of conceptual models and whether they had
different types of conceptual models available for use.
3. In the main part of the interview, we asked questions regarding the interviewees’
use of conceptual models. For example, we asked how often practitioners used
conceptual models, whether they used multiple types of conceptual models, why
they used different models, and what types of models they used for which purposes
and tasks.
4. Finally, we collected information about the familiarity of participants with different
types of conceptual models to be able to control for potential preferential biases
(Figl & Recker, 2014) in choosing or using conceptual models. This part also
Chapter 5:Combined Use of Conceptual Models in Practice: An Exploratory Study 69
contained questions for individuals who reportedly did not use combinations of
conceptual models, which was useful to explore boundary conditions.
Throughout all interviews, we followed the interview protocol closely to increase the
reliability of data and to ensure that the same questions were asked of every interviewee.
5.3.2 Participants and Procedures
The population of interest for our study consisted of practitioners in systems analysis
and design roles that used already created conceptual models for their work tasks (as opposed
to modelers who would create such models). We used purposeful sampling techniques for
participant selection for this study instead of random sampling (Eisenhardt, 1989).
To approach potential participants, we initially recruited UML practitioners from
LinkedIn groups, such as UML professionals, Business Analyst Professionals, Software
Developers, UML, UML Lovers, and Executable UML. We began with this focus for two
main reasons. First, UML is reported to be one of the most widely used conceptual modelling
methods and a de facto standard during information systems development (Davies et al.,
2006; Fettke, 2009). Second, UML provides a wide variety of model diagrams, each of which
conveys different information (Dobing & Parsons, 2008), which increases the likelihood that
UML practitioners engage with multiple types of conceptual models during their work.
However, we did not limit our sample to only UML users. We also approached other
systems analysis and design practitioners by searching for LinkedIn professionals who were
endorsed for skills in requirement analysis, system analysis or development, software design,
engineering, architecture, analysis, or object oriented design. Through this sampling
procedure we also recruited practitioners with expertise in EPC or SysML, for example.
To offset potential bias stemming from particular modelling roles (Recker, 2010a) and
to accumulate a broader view on the actual use of conceptual models in practice, we recruited
interviewees from different companies and with different roles. The variety of organisational
roles in our interview sample was important for us to investigate conceptual model use from
both technical and business aspects.
Using these procedures, we collected interview data between December 2015 and May
2016 from 15 individuals, across eight countries, incorporating various roles in their
respective business environments (e.g., system/business analyst, system engineer, software
70
architect). We conducted our interviews mostly through Skype, as most of our participants
were from countries outside of Australia. Interviews with locally available participants were
conducted in their office or preferred meeting place. All interviews were audio recorded and
transcribed by external transcribers upon reception. Each interview lasted between 45-90
minutes. Overall, we obtained over 18 hours of interview data, resulting in 372 pages of
interview transcriptions for analysis. Table 5.2 provides basic demographic information. The
reported experiences classified our interviewees as highly experienced professionals.
Table 5.2 Demographic information
Role and Responsibilities # of Interviewees Experience in years Mean Min Max
Business Analyst/consultant 6 Experience in Business 23.74 7 38
Software Architect 3
System Engineer 2 Experience in IT 21.8 7 35
System Architect 2 Experience as a System
designer/analyst 15.3 7 25
Software Developer 2 Experience in Conceptual
Modelling 15.2 5 25
Total 15
We conducted interviews until we reached a sample size we deemed appropriate and
substantial for our analysis, on the basis of guidelines by Guest, Bunce, and Johnson (2006)
and Francis et al. (2010). Specifically, we tracked the derivation of new knowledge from each
interview. We did this by interviewing, transcribing, and analysing each interview in
sequence, and plotting new codes emerging from each interview analysis and the
accumulation of codes for all interviews (Figure 5.1) until the trends indicated data saturation.
Figure 5.1 shows that we achieved saturation after approximately thirteen interviews; thus, we
ceased data collection after completing our 15th interview.
5.3.3 Data Analysis
We analysed the transcribed the data using NVivo 10. We used multiple coding
techniques commonly associated with the grounded theory method (Urquhart, 2013;
Urquhart, Lehmann, & Myers, 2010). Specifically, during open coding we analysed each
interview line-by-line to identify concepts or key ideas potentially related to our research
questions. Table 5.3 provides several examples of how we guided our coding through the
open codes collected initially.
Chapter 5:Combined Use of Conceptual Models in Practice: An Exploratory Study 71
Figure 5.1 Interview sampling and theoretical saturation
Table 5.3 Examples of open codes collected
“Nugget” from interview transcript Initial open code
“So, an information system is vastly more than the paradigm of process only or data
only, it’s a whole lot of interrelated considerations.”
Size of domain
“I don’t think there’s a single model to rule them all, no. I think within domains I
expect different model types, greater or lesser.”
Deficit in model
“Every single diagram does not give you an accurate description of the system, it's by
the combination of the diagrams that you get the complete description.”
Complementary use of
models
“We use BPMN for all [of] the process models, and that's mandated, we must use
that.”
Mandated by
organisation
“It is basically repetition. You are basically repeating yourself over and over in three different dimensions that are saying the same things.”
Redundant information
We then used selective coding techniques to categorise and assemble the identified
concepts to see whether we could match the identified concepts to existing theoretical patterns
based on our understanding of the literature (such as ontological qualities), or alternatively to
define new patterns grounded in our interpretation of the data. Figure 5.2 illustrates how we
tracked and categorised codes in our attempt to explain why practitioners used different types
of models, and how they selected the types of models.
72
* Numbers indicate the frequency of code across all interviews.
Figure 5.2 Illustration of coding procedure
5.4 FINDINGS
We discuss our findings in two main steps, in alignment with our research questions.
We start by exploring our data regarding why practitioners use multiple types of models in
combination. We then explore the factors were unearthed that influence selection and use of
different types of models in combination.
Chapter 5:Combined Use of Conceptual Models in Practice: An Exploratory Study 73
5.4.1 Why practitioners use multiple types of models
Not surprisingly, all of our interviewees (15 out of 15) reportedly used more than one
type of models in their design and analysis tasks. One interviewee reported: “… a lot of the
time I’m interested in using [different] models as a view onto the domain from different points
of view and then having links behind the scenes between those representations.” In fact, using
multiple types of models seemed to be usual for practitioners; an interviewee noted: “I need
two types of diagrams, at least two, one structural that shows the physical topology of my
domain […] and the dynamic view of it.”
Table 5.4 summarises the types of models reportedly used by at least two interviewees.
The results are reasonably consistent with other studies (Dobing & Parsons, 2008; Grossman,
Aronson, & McCarthy, 2005) even though we focused on model users as opposed to model
designers and we did not only focus on UML diagrams but also other model types (e.g.,
BPMN). Moreover, our results show a higher usage for activity diagrams, state machines
diagrams, and BPMN diagrams compared to previous studies. Still, we learned that, as
expected, professionals use UML selectively, but never completely (i.e., all available model
types).
Table 5.4 Most widely used models
Type of Models Number of
Interviewees
Purpose of Use (Number of Interviewees)
Class Diagram 13 Capture and define requirements (10), Collaborate with developers and engineers (5), Documentation (1), and Database design (3)
Use Case Diagram 8 Capture and define requirements (8), Collaborate with business people
and end users (5), and Scoping the project and estimations (1)
Activity Diagram 8
Collaborate with business people and end users (3), Collaborate with
developers and engineers (1), Documentation (3), and Basic
development design (2)
State Machine
Diagram 7
Collaborate with developers and engineers (7) and Documentation (5)
BPMN 6
Collaborate with developers and engineers (5), Documentation (5),
Change Management (1), Migrate to a New System (1), and Validate
Processes (1)
Sequence Diagram 5 Collaborate with developers and engineers (4), Documentation (3),
and Validate Processes (1)
Deployment Diagram 3 System Configuration (2) and Aggregation of system parts (2)
Component Diagram 2 Collaborate with business people and end users (1), Aggregation of
system parts (1)
Use Case Scenario 2 Capture and define requirements (1) and Documentation (1)
Entity Relationship
Diagram 2
Database design (2)
Object Role Model 2 Collaborate with developers and engineers (1) and Scoping out
information and data requirements (1)
74
Table 5.4 also summarises the main purposes for which practitioners reported use of the
different models. Note that a model could be used for multiple purposes. For example,
practitioners mostly stated that they employed use case diagrams for both capturing and
defining requirements and also for communicating with business people and end users.
As presented in Table 5.4, practitioners used different types of models for different
purposes. The most notable result is that practitioners used various types of models to
collaborate and communicate with users vs. analysts. For example, class diagrams were
mostly used to collaborate with developers and engineers, while use case diagrams were
mostly used to collaborate with business people and end users.
These findings are consistent with earlier studies. For example, Topi and Ramesh
(2002) noted differences in model-based communications between analysts and clients (i.e.,
end users) versus that between analysts and developers. They argued that more expressive
models can convey more information to model readers; however, this would be more useful to
analysts than clients. Our data likewise suggests that practitioners often trade-off complexity
of models to promote simplicity. An analyst cited: “When you go to, let’s say, a process
owner, [...], you can’t show her a sequence diagram or a class diagram. She simply won’t
understand […]. But you can show her something like an activity diagram, or use case
diagram. So, it depends on your communication partner.”
Reasons for the use of multiple models were more varied. At the outset, multiple types
of models were not always and consistently used by practitioners in their system design and
analysis tasks. Instead, practitioners chose sets of multiple models based on several conditions
of their tasks and initiatives. These included: 1) the size and complexity of the domain, 2)
involvement of different stakeholders in the projects, and 3) incompleteness of any particular
model. We explore each factor in turn below.
Size and complexity of the domain
One of the first findings that became apparent in our study was that system designers
and analysts may not need to use graphical representations of a system at all to gain an
understanding of the domain. When dealing with a small and simple domain, they simply use
a single model, if at all. Practitioners reported that this often provided them with enough
information to help them to achieve their tasks goals. A small and simple domain was
Chapter 5:Combined Use of Conceptual Models in Practice: An Exploratory Study 75
characterised as one involving only a few interacting elements with stable and clear cause-
effect relationships. For example, a “loan payment” system was described as having a limited
number of elements with direct and stable relationships in comparison with a system with a
large number of interconnected components with non-linear relationships, such as banking.
As one system analyst commented: “So, a small project, I might get by with one class
diagram […]. For a bigger project, I might want to describe certain key processes that really
do matter […], using different diagrams.”
We also learned that for small and simple project development processes, conceptual
modelling may not only not be required, but could also slow down the development process.
However, for initiatives involving systems in big and complex domains, interviewees
consistently noted that it would be challenging to understand the domain through the use of a
single model only. One practitioner noted “[…] because our system is far more complex than
either data or process [model]”; while another mentioned “[the] banking domain is way too
complex to represent on one type of diagram”.
We learned that practitioners do indeed use conceptual models as specific abstractions
of a real-world domain from a particular perspective. When a system to be represented is
complex, multiple interrelated models were used to represent different aspects of the system,
much as suggested in the study by J. Kim et al. (2000). A practitioner commented: “Yes, in
order to depict everything that happens in a wire transfer, for example, I’d need static
diagrams, component and class diagrams, but also behavioural diagrams, like activity.”
Practitioners had their own way of classifying abstraction perspectives, often guided by
available textbooks that provided simple classifications, for example, categorisations into
behavioural and structural (Booch, Rumbaugh, & Jacobson, 2005). Figure 5.3 summarises the
most often named model types associated with a particular systems design perspective. It
shows that practitioners mainly used class diagrams or component diagrams to depict the
structural perspective of a domain, while activity diagrams or state machine diagrams were
used to describe the dynamic perspective of a domain. As one system analyst noted: “I need a
state machine to capture what’s going on inside it, and again if it really is a big project, class
diagrams will be good to describe certain key low-level aspects of the system.”
76
Figure 5.3 Example of diagram use to represent different aspects of a system
We further learned from our data that practitioners felt that understanding a complex
domain through the use of a single diagram alone would be challenging. A practitioner
commented: “Where that starts to be a real challenge, is where that system is very complex,
as it starts to include more and more aspects….” In fact, our interviewees consistently
commented that they often dealt with large and complex systems that included a large set of
interrelated aspects. In these cases, using different models in combination improved their
domain understanding. A user commented on why they used a combination of models: “Well,
the short answer is ideally you want to do everything. You want to get the best understanding
of a system as possible.”
Involvement of different stakeholders
Different stakeholders are involved in systems analysis and design, all of which require
an understanding of the relevant problem domain in some form or other (Darke & Shanks,
1997). In the context of our research, and akin to (Topi & Ramesh, 2002), our data suggested
that two categories of stakeholders are relevant in terms of the use of multiple models:
stakeholders as systems users and stakeholders involved in systems analysis and design.
Stakeholders as systems users occupied roles such as process owners, technical personnel, and
lay end users. Stakeholders involved in system analysis and design were managers, systems
analysts, and developers. Importantly, we learned that each of these types of stakeholders may
require different models. A system analyst commented: “My communication partners
currently are, as I said, line management, lower management, technical personnel in the
Chapter 5:Combined Use of Conceptual Models in Practice: An Exploratory Study 77
bank, analysts, and my own developers. And for all, each group, I need different sets of
diagrams.”
We further learned that practitioners use different types of models that are
understandable by different stakeholders in order to be able to communicate with all of the
stakeholders involved in the project. They do so because individuals involved in the project
glean a different level of understanding from different models. Out of our 15 interviewees, six
mentioned that they used multiple, different types of models when they wished to be able to
communicate with an entire project team beyond the technical staff responsible for analysis
and design. A system analyst commented: “When you go to a lady who basically just ticks off
issues in an issue management system, you can’t show her [a detailed diagram]. She simply
won’t understand, and within ten seconds her eyes will glaze over, and you’ve lost her.”
Another practitioner stated: “If you go and seek software engineers and you turn up with
something a bit too simplistic, they wouldn't take you seriously.”
Our data also suggests that not all stakeholders in a project desire to obtain the same
level of understanding about the details of a system. Managers and decision makers normally
wish for an overview representation of the entire system. According to our interviewees,
developers and analysts want to obtain a deep level of understanding about the system or
domain. A practitioner commented: “When you are talking to business people, maybe you
don’t like to use state diagrams because it’s too technical in most cases. […] but when talking
to business people they want to see simple diagrams, and they want to see what standard
value is.”
Incompleteness of single models
In probing for reasons why practitioners use multiple models, we also learned that, as
predicted by ontological analysis, ontological deficits of a model provided some key
explanations. Eighty percent of interviewees frequently used phrases such as “expressive
power”, “deficit”, or “ability to capture information”, all of which allude to the existence of
ontological deficit or representational incompleteness of a model. As one interviewee noted:
“Sometimes there is no language that represents some concepts in the domain.” A second
interviewee commented, “I mean, I don’t think you have a modelling technique that
succinctly captures issues in all of the areas in which you are trying to conceptualise the
system.” A third practitioner argued that he used different types of models: “Basically
78
because I can express certain aspects of the whole system [only] with specialised kinds of
diagrams.”
5.4.2 How practitioners decide on which models to use
Above we reported that all interviewees indicated that they used multiple models, at
least two, in most of their design and development processes. Reasons for doing so included
stakeholders, domain size, and complexity and incompleteness. This behaviour raises the
question of how practitioners select what types of models to use in their design and analysis
tasks. In this section, we explore the reported rationales that determined the selection of types
of models. We structure them broadly into semantic reasons – selection rationales grounded
in the capacity of models to represent a domain, and pragmatic reasons – selection rationales
grounded in the socio-technical context of conceptual model use.
Semantic Reason 1: Select models that increase the completeness of a representation
The most eminent reason for the selection of multiple models noted by our interviewees
was their desire to maximise the completeness of representation. To do so, almost 80% of
interviewees mentioned that they use additional models to that end. One analyst noted, “Well,
the reason you need [more than one model] is because each one of those diagrams only
depicts certain properties of the system”. Another practitioner noted, “In order to depict the
system completely, or more fully probably, so […] you need to use multiple diagrams to
depict it.” This suggests that system designers and analysts select those models in
combinations that are complementary to each other and enable them to achieve a maximal
level of ontological completeness.
The main way that practitioners attempted to achieve completeness was to select models
that featured alternative perspectives, from structure to behaviour, for instance. A user
commented: “Every single diagram does not give you an accurate description of the system,
it's by the combination of the diagrams that you get the complete description. So, I have to
look at my system from different angles. The structural angle, the behavioural…”
However, some practitioners selected models that described the same or similar
perspective as another, but provided more detailed information about the requirements within
a specific aspect. For example, some mentioned that if they were particularly concerned about
the behavioural view of the domain, they selected those models that provided more
Chapter 5:Combined Use of Conceptual Models in Practice: An Exploratory Study 79
information about the behaviour of the domain. For example, a system designer commented:
“So, the areas that I’m not comfortable with I tend to model, and say if it’s behaviour I’ll be
leaning on the dynamic models more, so the activity diagrams, the state transition or the third
one.” However, if their main concern was about information and structure of the domain they
selected the types of models that provided them with more detail in that view. Another system
designer noted: “The basic system design is by looking into the use cases and a collaboration
which tells you “Okay, I have an idea, this and that must be done, and I kind of need business
[uptakes] doing certain things and having certain attributes”. So then you have a basic class
model, and this class model is again linked to the use cases from where you derive it and the
collaboration.”
This data suggests that on the one hand the selection of the models is based on which
model gives more information, but on the other hand, also depends on which types of
information a user requires. In other words, selection also relates to the purpose of the task,
which we explore further below.
Semantic Reason 2: Select models with little overlap to minimise cognitive demands
The second rationale in the selection of models we found was that practitioners tried to
avoid overlapping models because they add unnecessary complexity. A practitioner noted:
“I’m trying to avoid a situation where I was using a mix of functional representations. You
know, I'm trying to stay with only one sort of representation in the functional view or in the
information view or whatever. Otherwise, it starts to get messy.” A second analyst noted:
“Do I allow competing model types that attempt to describe the same sorts of things using
different formats? […] Absolutely not, it is a disastrous practice, it always ends badly, it
always causes problems.”
When practitioners referred to terms such as “overlap”, “competing”, or “redundant
models”, they also consistently referred to terms such as “problems”, “complexity”, or
“cognitive effort” that such selections lead to. In other words, practitioners frequently
commented that high overlap between models would increase the cognitive demands on those
trying to use the models. For example: “[for] a general user community three different models
with three different conventions all attempting to describe the same sort of thing but
describing it in different ways – that’s a pretty big cognitive task for your average user to
have to get through.”
80
We also learned from three practitioners that cooperating with different stakeholders
may lead to an increase in overlapping models if and when those stakeholders favour
particular models and representations. A system analyst mentioned: “So, I’ve done some
database work where the software development has used UML class diagrams to represent
the logical data model but then there’s been a database expert who’s used an entity
relationship model from the same project to also represent the database, and generally
speaking we try to integrate these things and if we can use a single language, we use a single
language, because otherwise it just adds complexity.”
Semantic Reason 3: Select models that establish correspondence between them
We also learned that the extent of overlap mattered. Two practitioners commented that
in selecting models, they select those models that have some correspondence between them.
Correspondence in this context refers to the existence of semantically related or equivalent
elements that appear in multiple models (such as a construct representing “the customer” in
both a process and a data model). Model users believed that this correspondence between
models helped them to link models together. A practitioner indicated: “I mean, there are what
I call 'touch points' between the different paradigms there because you will have things that
are referred to in a business process functional view which are the artefacts which also
appear in the information log. There are these touch points between the different views. But
that's the value in having […] the diagrams representing a model because you can then start
and do checking between the different views.”
While six interviewees commented that overlap between models may increase
confusion and increase complexity, the existence of some level of correspondence between
models was also noted as a useful means to help users to better link different models and in
turn improve the understandability of the information.
Only one of our interviewees stated that she attempted to keep the total number of
models as minimal as possible because correspondences also present a burden. She believed
that too many models would make it too difficult for readers to assimilate the information: if
the number of models increases, readers need to find links between models and map them to
their own mental model of the phenomenon.
Chapter 5:Combined Use of Conceptual Models in Practice: An Exploratory Study 81
Pragmatic Reason 1: Select models based on organisational regulations and
standards
What types of models were selected also depended on organisational settings. Most
organisations define some internal regulations in a set of modelling conventions (Delfmann,
et al., 2015; Rosemann, 2006; Samuel, et al., 2015) to facilitate analysis and design of
information systems, as noted by a system designer: “It [use of models] depends on the
standards and the template that companies define.” Most comments showed that the selection
and use of conceptual models fairly strictly follows these predefined regulations:
“Organisations will rarely just leave it to an individual project team or an individual to just
arbitrarily choose whatever technique. They typically have these quality standards which they
set…”
We found that 60% of the practitioners we interviewed were indeed governed in their
use of models through organisational conventions. While these take different formats, the
most common objective of these guidelines was harmonisation. An interviewee commented,
“We have standards based on how we define for the sake of consistency.”
In terms of contents of such conventions, we learned that most of the organisations had
predefined regulations regarding not only on what types of models should be used but also
how these models should be represented. An analyst mentioned, “They create not only
standards of which techniques to use, but most importantly […] what conventions in these
techniques you should use.” This involved predetermination of which types of models were
available to use. For example, a practitioner commented “Yeah, in our case we use UML. We
use BPMN 1.1. We use activity diagrams. We use class diagrams. These are the agreed upon.
We don’t use ERD and we don’t really use context diagrams. So, those are the agreed upon.”
Moreover, the organisational conventions in use specified the types of model constructs that
should be used in specific occasions. A practitioner commented, “They have the standard on
[….] what specific constructs you should use and when you should use them and what degrees
of modelling apply to them and how you should describe [them].”
In summary, we learned that organisational regulations, if they exist, can overrule or
mitigate some of the semantic rationales for selection rooted in the models themselves (such
as completeness or overlap). For example, some organisations explicitly required
documenting a system in types of models that overlap with each other: We do use it right now
82
because it is an agreed on the requirement, which is a lot of work. But asking the question of
relevance. Is it always required? No. We are actually challenging that; we’re saying “No,
this is a lot of work”. It is basically repetition. You are basically repeating yourself over and
over…”
Pragmatic Reason 2: Select models based on involved stakeholders
Involvement of different stakeholders not only leads model users to use different types
of conceptual models but also affects the types of models IS professionals choose to use. An
analyst commented: “So, I think that the selection of a modelling language should be
influenced by your target audiences that you want to communicate with.” Different audiences
may require different types of models because not all audiences desire the same type of
information. A practitioner commented: “You've got the very high-level executive decision
makers, especially the ones who shine off of all the money and what they want is different
from the IT technical people who will actually implement the system.”
This situation also implies that practitioners may use the same types of models with
different stakeholders, but with different levels of details because not all stakeholders seek to
gain the same level of understanding of the details of the system. One practitioner mentioned:
“I may use use-case diagram with technical people, but then I would use it differently”, while
another practitioner stated: “So, who I want to say it to influences the type of diagram that I
choose and saying it concisely means that I want to restrict to an absolute minimum. The
things in that diagram should just include the things that I needed to include in order to
communicate whatever it was I wanted to say.”
We learned that practitioners select different types of models based on their audience’s
targeted level of understanding from different diagrams. A practitioner noted: “So you’re
trying to get a feeling for what type of diagram your communication partner would or could
intuitively understand and you choose that one. That’s why there’s no problem using a
sequence diagram or a class diagram with a developer. Most of them understand it. But if we
talk […] let’s say with a line manager, you’re basically reduced to component diagrams, and
for the behavioural part, you’d be limited to activity diagrams. […] So, I use activity
diagrams with them based on their level of understanding.”
Chapter 5:Combined Use of Conceptual Models in Practice: An Exploratory Study 83
We also learned that practitioners choose types of diagrams based on the target audience
skill level and familiarity with models. A system analyst commented: “With one of them, he’s
a senior experienced developer. He immediately understands something like a sequence
diagram, for example. With one of them, he’s […] very talented, but also very junior. So, I
can basically only use class diagrams or component diagrams.”
Finally, we uncovered that stakeholders sometimes had their own preferences and liked
to use specific types of models due to different purposes or because their previous projects
involved specific types of models, that is, they had already used those models. For example, a
system analyst noted “Sometimes the customer will already have whatever request to use
particular types of models. That might be because they are already using them for a
particular purpose or for historical reasons and things like that.”
Pragmatic Reason 3: Select models based on adoption levels
A third pragmatic factor that affects choosing types of models to be used is the overall
diffusion of the models in industry. We found that widely used models were preferred in
many system analysis and design tasks, for two main reasons: first because it is more likely
that stakeholders involved in the projects are familiar with more popular models. An analyst
commented: “[…] the other main driver is the fact that they're the most popular. So, they're
good enough and they're the most popular, that means […] you're more likely to find someone
who knows UML and BPMN than someone using something a bit more esoteric.” Second, for
documenting purposes, it is more likely that system designers use the most up-to-date
modelling grammars to document a system for the purpose of traceability. “But more recently
I think the industry has been dominated by UML […] I could use the data flow diagram, I
could use their equivalent of the state transition diagram […]. But, I tend not to use those
because what customers tend to want is some kind of more modernisation.”
Pragmatic Reason 4: Select models based on task goals
We also unearthed that the specific purposes of tasks affect the selection of different
types of models. Practitioners mentioned that not all types of models can be used for the same
purposes. One designer commented: “Yeah, I think there is, certainly, techniques that are
useful for particular purposes. I think that’s why those techniques continue to be long
standing.” Almost 80 percent of the practitioners we interviewed mentioned that they select
models that provide the best “fit” to the purpose of the task they are trying to complete. One
84
system engineer commented “Which diagrams you use is dependent on […] what the problem
is or the characteristics of the system. […] So, it depends really on what you’re trying to
understand or what you’re trying to specify.”
We also learned that practitioners select types of models based on the characteristics of
the domain and what level of understanding they are trying to glean from that domain. A
systems developer commented: “So, it might be that the problem requires more structural
modelling than behavioural modelling or more behavioural modelling than structural
modelling and it really depends on the nature of the problem. And so also it may be that
you’re dealing with a system that has no interesting states but does have interesting message
sequencing, or it could be the reverse.” Another practitioner stated: “If I am mostly interested
in the process and possible data flow in that process and the decisions during the process […]
then I would tend to want to use an activity diagram. On the other hand, if I was more
interested in my system being in a number of states, what I want to focus on is that it is very
driven and depending on receiving events it will transition between those states. Then, of
course, I would be more interested in using a state machine diagram.”
Pragmatic Reason 5: Select models based on tool support
Tool support was a fifth pragmatic factor we identified as a model selection rationale.
When different types of conceptual models were available, 40% of interviewees mentioned
that they preferred to use models that were supported by appropriate tools available to them.
A developer commented on how she selected models as: “It depends on what tool support
there is.” For example, another developer mentioned that “[I use state machines] because I
have a tool that translates my state machines into code.”
We also learned that the availability of tools can be in compliance with the
organisational standards. A developer noted: “So it is really critical to set standards not only
in what techniques you use, what tools you use and what conventions you set.” Another
practitioner commented: “Tools have to comply with the standards. […] Each team needs to
have the tool and the management that says okay…”
5.5 DISCUSSION
We explored how practitioners reportedly use conceptual models during systems design
and analysis, with a focus on why and how they use different types of models in combination.
Chapter 5:Combined Use of Conceptual Models in Practice: An Exploratory Study 85
Fairly immediately we found that indeed multiple types of models are used in today’s system
development and analysis, which confirms earlier reports (Davies et al., 2006; Dobing &
Parsons, 2008; Fettke, 2009). Our main objective then turned to exploring why IS
professionals use multiple models and how they select different types of models.
We first explored factors that lead IS professionals to use multiple models. Several of
the findings that we noted, in and by themselves, have been noted previously. For example,
our data analyses suggests that IS professionals use multiple models because 1) no one model
is a complete representation of a domain (e.g., Green et al., 2007; Irwin & Turk, 2005; Recker
et al., 2009; Siau, 2010); 2) complex and large domains are much broader than what can be
presented in a single model (e.g., J. Kim et al., 2000); and 3) different stakeholders are
typically involved in a project, with different levels of understanding and familiarity with
different models and posing different requirements towards a model (e.g., Darke & Shanks,
1997; Topi & Ramesh, 2002).
While many of these findings on their own are consistent and corroborative of earlier
studies, it is the interaction of these factors that is interesting and novel. Our view of the
model that describes the different rationales for using multiple models in combination during
systems analysis and design is visualised in Figure 5.4. The model suggests that the extent
and type of multiple model use is determined by two primary factors: the size and complexity
of the real-world domain, which the existing or intended representation is meant to support;
and the provided representation by any one conceptual model, in terms of the perspective
provided on the domain and the level of incompleteness this model has. The model further
proposes that both extent and type of multiple model use, given one existent model will be
moderated by the requirements of the involved stakeholders, in terms of the level of
understanding they seek from the models (e.g., surface versus deep) and the extent to which
they are familiar with the provided and any additional model representation.
86
Figure 5.4 The role of different factors that determine the use of multiple models
These two sets of rationales in our model provide two important extensions to the
literature:
1. The model emergent from our data is consistent with both Wand and Weber’s
(1990a, 1993) representation theory and their research framework on conceptual
modelling (Wand & Weber, 2002); and provides a synthesis of both: the model
suggests that the use of multiple models can be explained by both conceptual
modelling script and context factors, with both direct and mediated relationships
between them.
2. The model also extends the main premises of representation theory in itself: it
proposes that not only the incompleteness of single models is a key determinant
of the use of multiple models, but also that the supported perspective of the
models (e.g., structural vs behavioural) is a main direct determinant. Rosemann
Conceptual Modelling Context
Relevant stakeholders- required type of understanding
- familiarity with models
Relevant real-world domain- size
- complexity
Conceptual Modelling Script
Conceptual model- ontological incompleteness
- supported perspective
Use of multiple models- extent- type
necessitates
increases likelihood
influences choices
Chapter 5:Combined Use of Conceptual Models in Practice: An Exploratory Study 87
and Green (2002) suggested that such perspectives can be identified in the
ontological constructs underlying Wand and Weber’s theory (e.g., things and
their properties, states assumed by things, and so forth), which aligns with our
interview data.
Our second main objective was to explore how model users choose which types of
models to use. Figure 5.5 shows our view of the emergent model forthcoming from our
interpretation of the findings that describes the selection rationale.
Figure 5.5 The role of different factors that determine the selection of different types of models
Much like the reasons for using multiple models, our findings indicate that, broadly,
two sets of factors, about the semantics of the models as script combinations, as well as the
pragmatic contexts in which they are used, influence the selection. Again we note how the
script combination factors match concepts proposed in the different extensions of Wand and
Weber’s representation theory (Green, 1996; Recker, 2014; Weber, 1997); yet, again we also
discovered a new concept: that of correspondence. Correspondence implies that while
Pragmatic Reasons
Conceptual Modelling Context
Organizational regulations and standards- conventions
- construct specifications
Adoption- diffusion
- popularity
Semantic Reasons
Conceptual Modelling Script Combination
Overlap- redundant constructs- similar perspectives
Selection of models- extent- type
requires
overrides
varies Correspondence- model linkages
- visual cues
Completeness- additional representations
- varied perspectives
increasesdecreases
minimizes
Relevant stakeholders- required type of
understanding- cognitive demand
Task- objectives
- requirements
Tool Support- supported grammars- translation support
influences choices
constrains
88
practitioners avoid overlapping models, the existence of semantically related elements in pairs
of models influences the selection of these models.
Again, we also identified the strong prevalence of context, which has, much like
pragmatics, largely been ignored in most of the theoretical and empirical work (Bera et al.,
2014; Burton-Jones et al., 2009; Recker, 2007). We identified factors, such as organisational
conventions, tool support, and adoption levels of models, that clearly influenced the selection
of the types of models IS professionals use in their design and analysis tasks. In themselves,
these findings are not new; similar findings had been reported in Delfmann et al. (2015);
Recker et al. (2010); Rosemann (2006); and Samuel et al. (2015). For example, Recker et al.
(2010) identified tool support and organisational conventions that influence how analysts use
BPMN grammar. These factors were also prominent in our data, and in addition to these
factors, we exposed the factor of “adoption level” as also influencing the selection of types of
models. Furthermore, consistent with Samuel et al. (2015), we identified that objectives and
requirements of the task at hand, together with relevant stakeholder requirements, influence
the choice of different models to use.
Beyond the identification of these factors, our dominant novel finding is that the
relationships between attributes of the modelling context and the modelling script
(combination) are far from trivial: for example, organisational conventions may direct model
users to choose models that overlap. These organisational regulations may increase
ontological overlap, whereas in principle, they could be designed to mitigate it instead.
Together, the findings of our exploratory study and assumptions proposed by our two
theoretical models provide novel theoretical logic and demand further research to
operationalise and test our conjectures. Whereas our research was, deliberately, exploratory
and qualitative, in our view, the follow-up research should be falsificationist and quantitative.
5.5.1 Implications for ontological research on conceptual modelling
We identified three main implications for ontological research on conceptual modelling.
First, we have shown that ontological factors are among the main drivers of using multiple
models and selecting types of models. While this empirical knowledge is novel, it falls within
theoretical expectations. It corroborates the relevance of ontological qualities in the actual use
of conceptual models in system analysis and design tasks. One novel, empirically grounded
Chapter 5:Combined Use of Conceptual Models in Practice: An Exploratory Study 89
extension of our work is in the interaction between completeness and the overlap of models:
while, perhaps as expected, ontological incompleteness is often a starting point for the use of
multiple models, the interaction between completeness and overlap in the selection of models
is not always clear-cut: it would appear that in dependence of the context of use, overlap
between models (and the correspondence it provides) may sometimes bring forth desired
qualities for users.
Second, we also found that not only ontological qualities are important in using and
selecting multiple types of models; contextual and individual difference factors also play
important roles on why and how IS professionals use conceptual models. Importantly, again,
we identified not only main effects but also important interaction effects between context and
script factors. The main factors we found are not necessarily novel; however, the interaction
effects we discovered are novel and stimulating: sometimes, context influences the impact of
script qualities; sometimes it overrides them. This was most evident in the role of
organisational conventions and individual differences (Khatri et al., 2006; Recker et al.,
2010): depending on regulations and intended audiences, IS professionals sometimes choose
to entirely compromise or even reverse ontological qualities (both in terms of overlap and
completeness). For the ontological research streams on conceptual modelling, our study not
only suggests the adage that “context matters”, but also offers novel propositions about the
existence and type of interaction effects that are worth exploring.
A third implication concerns the ways in which our study can be extended: we focused
on two overlooked aspects that fall into the pre-usage stage of conceptual modelling: the
rationales for why and how users select multiple models. These rationales determine the
choices users make in light of the tasks they set out to complete. During these tasks, they will
be required to interpret the chosen models to glean real-world understanding from them and to
assist them in meeting their task goals. Therefore, an important extension to our work would
be field research that examines whether the identified factors and proposed effects also inform
the development of domain understanding from multiple models (Gemino & Wand, 2003)
and overall task performance of users and the extent to which they found the models useful
for their tasks (Maes & Poels, 2007).
90
5.5.2 Implications for empirical research on conceptual modelling
Our research adds to the empirical body of knowledge by exploring conceptual
modelling context and scripts factors as influential factors affecting the use of multiple types
of conceptual models. Previous studies have indicated that IS professionals normally use
multiple types of models; however, it is unknown how and why they are using different types
of models. We uncovered that there are important context factors that lead IS professionals to
use multiple models and also select what types of models they wanted to use. There is much
that can be done to continue this research. Some of the identified factors, such as individual
differences (Khatri et al., 2006), organisational conventions, tool support, and adoption levels
(Delfmann et al., 2015; Recker et al., 2010; Rosemann, 2006; Samuel et al., 2015), or task
goals (Figl & Recker, 2014), have previously been considered in studies of the use of single
models or grammars. Opportunities now exist to examine the role and interactions of these
factors on using different types of models. For example, future work could attempt to qualify
the relative importance of our identified factors in different contexts or usage stages. For
example, one replication of our study could explore different usage stages (e.g., model
interpretation, model application, or model creation stages) by using a similar research design.
A second way to replicate our study would be to explore differences in the study population,
for example, by interviewing both novice and expert model users (Batra & Davis, 1992;
Shanks, 1997). A third meaningful continuation would flow from the exploration of
particularly large and complex domains, to explore how systems analysis and design on the
basis of multiple models tries to mitigate challenges related to volume, emergence, and
complexity. Domains such as mission-critical systems (Pettit, 2008; Pettit & Street, 2005),
information systems used in geology or ecology (Goodchild, 2012), or in building information
management (Merschbrock & Munkvold, 2012) might be good settings for such
investigations.
5.5.3 Implications for the practice of conceptual modelling
Several practical implications flow from our work: one, our findings inform
components that would be important to be included in conceptual modelling conventions and
guidelines: What types of models should be used to meet task purposes, to satisfy different
stakeholder requirements, and that dictate desired levels of completeness and clarity of
representation?
Chapter 5:Combined Use of Conceptual Models in Practice: An Exploratory Study 91
Two, our study provides practical guidance for IS professionals for identifying
appropriate models for use. Specially, our findings suggest that in dealing with large and
complex domains, multiple types of models may improve domain understanding and inform
that those types of models should be selected that not only best fit the purpose of task, but
also meet different stakeholder requirements.
Three, another implication of this study is in regards to managerial decisions concerning
tool selections. Our findings uncovered that IS professionals prefer to use those models that
they are supported by tools available in their organisations. This suggests a managerial
direction for decision makers to invest in appropriate supporting tools to achieve the best
results from conceptual modelling.
Four, we found that organisational conventions may lead IS professionals to use
overlapping models, whilst, prima facie, users prefer to avoid using models that have high
overlap. These findings suggest that it is an important managerial decision to identify
appropriate types and contents of models to be included in such conventions. Decision makers
should (re-)design organisational conventions in such a way that not only provides system
designers and analysts with appropriate types of models for a given task, but also enables
them to provide a complete representation of the domain without overlapping information.
5.5.4 Limitations
We followed an inductive approach to determine how conceptual models are used in
today’s system analysis and design practice. This choice of research design brings forth
several limitations common to exploratory, qualitative, and inductive research (Shanks, 2002;
Yin, 2009), which we were aware of and mitigated where possible.
First, as is normal for interviews, it was time consuming to approach participants and
conduct the interviews. We continued the interviews until we reached appropriate saturation,
as explained; nevertheless, our absolute sample size of fifteen interviewees is arguably low.
Second, the quality of data from interviews is dependent on the capability of the
interviewer and performing reliable interviews. To manage this, we conducted five pilot tests
before the actual interviews, and in order to avoid bias in the questions, the interviewer
followed the designed protocol closely. Moreover, following the designed protocol helped us
to collect reliable data.
92
Third, our sample was drawn mostly from practitioners who were using UML in their
system design and analysis tasks. Our choice was motivated by the observation that UML
users were most likely exposed to a variety of grammars and models, whereas other types of
practitioners, for example, data or process models, may simply not have had exposure to
contemplate using multiple models. However, to increase the generalisability of our findings,
we also interviewed practitioners who were using other conceptual modelling techniques. We
also specifically recruited practitioners with different backgrounds, roles, and skills to
maximise the generalisability of our findings.
Fourth, we did not interview practitioners who did not use conceptual models in their
analysis and design tasks. Our findings are therefore susceptible to positive confirmation bias.
However, this was not by design: we recruited any practitioner involved in systems analysis
and design with some knowledge of UML or other models, independent of whether they used
it or not. It just so happened that all of our interviewees did in fact have experience in using
conceptual models in their work.
Finally, we did not consider using multiple conceptual models for a particular purpose
or in specific phases of system design and development process. As the purpose of tasks may
change during the life cycle of system development, and different stakeholders may be
involved in different phases of system development, practitioners may need to use different
types of models at different times. Moreover, there may be other or different factors
influencing this decision at different times. For example, it is conceivable that different phases
of the systems development life cycle may require models with different levels of
completeness and overlap. Future studies could expand our study by considering this
contingency.
5.6 CONCLUSION
Burton-Jones et al. (2009) highlighted the important role of the interaction between
semantics and pragmatics when understanding conceptual models. They noted a dearth of
research on this topic, as well as, more broadly, a need for “more theoretical work and more
exploratory studies of conceptual modelling in practice” (p. 514). In this paper, we have
attempted to provide both: an empirical account on the use of (multiple) conceptual models in
practice as reported by practitioners and the development of important extensions and
Chapter 5:Combined Use of Conceptual Models in Practice: An Exploratory Study 93
contingencies to existing theory in this field, mainly representation theory and the associated
ontological analysis.
We believe the value of our research lies particularly in its generativity and fertility: our
analyses and findings are generative because they lead us to ask several questions that
challenge the common assumptions about conceptual modelling practice and theory: for
example, one question our research asks is: Can we use existing theory about the
representational semantics of conceptual models to better formalise the perspectives these
models arguably (and perhaps intuitively) provide to practitioners?
Our results are also fertile because they provide clear impetus for theoretical
advancement and development: for instance, we demonstrated how constructs that describe
the modelling context (e.g., task factors, conventions or tools) interact with factors that
describe the semantics of a set of conceptual models (e.g., overlap and completeness). The
propositions from our emergent models can be used both to refine and extend the core
premises of representation theory and its extensions for multiple grammars and models
(Green et al., 2011; Recker, 2014) and to evaluate contingencies in the empirical domains in
which the theory predictions purportedly hold. These are important avenues to continue the
active and ongoing dialogue in our field.
94
Chapter 6: An Experimental Study to Test and
Extend the Theory of Faithful Use of
Conceptual Model Combinations11
ABSTRACT
Conceptual models are used to graphically capture the requirements of an information
system. Research on conceptual modelling has focused on improving the modelling process.
The vast majority of studies have focused on evaluation of single grammars or models.
However, practitioners typically use multiple, and different, types of models in system
analysis and design tasks. Recker (2014) proposed a theory of the faithful use of conceptual
model combinations. The theory develops propositions to explain the model selection, domain
understanding, and perceived usefulness of model combinations. On the basis of a laboratory
experiment with 131 students, this research presents the first dedicated empirical evaluation of
the propositions of Recker’s (2014) theory. Our results of the testing of the three propositions
lead to the conclusion that the level of combined ontological completeness and ontological
overlap impacts users’ selection of model combinations, as well as user’s understanding
across measures of comprehension and deep-level understanding. However, the impact on
users’ retention ability and perceived usefulness of model combinations was not significant.
Our research informs implications for research and practice by providing empirical evaluation
of the merits of the theory of faithful use of conceptual model combinations and guideline
development to better use of model combinations for different purposes.
Keywords: Combined Ontological Completeness, Ontological Overlap, Theory
Testing, Experiment Design, Ontology, Faithful Use
11 An Experimental Study to Test and Extend the Theory of Faithful Use of Conceptual Model Combinations, To
be submitted to MIS Quarterly
An initial version of the paper was published as: Jabbari Sabegh, M. A., Recker, J., & Green, P. (2016).
Designing Experiments to Test the Theory of Combined Ontological Coverage. Paper presented at the
International Conference on Information Systems, Dublin, Ireland
Chapter 6:An Experimental Study to Test and Extend the Theory of Faithful Use of Conceptual Model Combinations 95
6.1 INTRODUCTION
A key activity in systems analysis and design is to develop and use representations,
often graphical, of relevant features of the domain under examination (Burton-Jones & Meso,
2006). These representations are called conceptual models. Conceptual models are developed
using modelling grammars (Wand & Weber, 2002), that is, sets of constructs and rules that
show how to combine the constructs to model real-world domains. Studies about conceptual
modelling abound (Gemino & Wand, 2004), and they have investigated a wide variety of
aspects. However, they have almost exclusively focused on single models or single grammars,
for example, how the use of a model can be improved by additional notation elements such as
colours (Masri et al., 2008), or how much domain understanding can be created from reading
a model (Bera et al., 2014). By contrast, research on the use of multiple models is very sparse.
Results of empirical survey research, on the other hand, show that IS professionals typically
do not use just one conceptual modelling grammar, let alone one conceptual model in their
analysis and design tasks (Dobing & Parsons, 2008; Fettke, 2009). In fact, multiple types of
models, often developed using different grammars, are usually used in combination. For
example, UML provides grammars to design many different types of models, such as class
diagrams and sequence diagrams; and practitioner surveys show that many of these model
types are used together (Dobing & Parsons, 2008).
Recker (2014) proposed a theory of faithful use of conceptual model combinations to
predict and explain the use of multiple models in combination. This theory provides
propositions to explain and predict the selection of models, generation of domain
understanding from multiple models, and perceived usefulness of models in combination.
This research is the first dedicated empirical study with the aim of evaluating the propositions
of the theory of faithful use of conceptual model combinations.
The objectives of this paper are therefore: (1) to investigate how users select a model
combination from a set of available models to achieve given task goals, (2) to examine how
much domain understanding can be generated from a model combination based on different
levels of completeness and clarity, and (3) what the roles of combined ontological
completeness and ontological overlap are on perceived usefulness of a model combination. To
that end, we conducted an experimental study with 131 information systems students to test
and extend the theory of faithful use of conceptual model combinations.
96
The rest of this paper is structured as follows. In Section 6.2, we review relevant
literature about the use of conceptual models by system analysts and designers. In Section 6.3,
we briefly recall the propositions of the theory of faithful use of conceptual model
combinations. We then describe the research method, the design of the experiment, materials,
and how we operationalised the levels of combined ontological completeness and ontological
overlap. Section 6.5 reports the hypothesis development. Next, in Section 6.6, we explain the
results of data analysis. In Section 6.7, the discussions of the findings and the key
contributions of these findings are presented. The chapter ends with the conclusion of this
study.
6.2 USE OF MULTIPLE CONCEPTUAL MODELS
Conceptual models are used for a wide variety of purposes, such as communication,
improving domain understanding, designing and improving processes, documenting or
analysing requirements, and designing and developing software or databases (Davies et al.,
2006; Siau, 2004; Wand & Weber, 2002). Much research has been carried out to understand
the capabilities of conceptual models to achieve these objectives. Reviews of this literature
have been provided by Wand and Weber (2002), Burton-Jones et al. (2009), Recker et al.
(2009), and Green et al. (2011), amongst others.
Across this literature, we noted that, with the exception of a limited number of studies
on using multiple grammars (e.g., Green et al., 2007; Green et al., 2011) or models (e.g.,
Gemino & Parker, 2009; J. Kim et al., 2000; Siau & Lee, 2004), much of the research on
conceptual models focuses on the use of a single grammar or a single model (Bera &
Evermann, 2014; Bodart et al., 2001; Bowen et al., 2006; Gemino & Wand, 2005; Parsons,
2011; Recker, 2013). The dominant findings of research on conceptual modelling grammars is
that any one grammar, and as a result, any single model, suffers from some level of construct
deficit (Irwin & Turk, 2005; Recker et al., 2010; Recker et al., 2009; Hong Zhang et al.,
2007).
Weber (1997) argued that if analysts or designers encounter ontological incompleteness,
they will somehow have to augment their model(s). One plausible strategy is to use additional
grammars or models. For instance, Green et al. (2011) demonstrated how grammars with
different representational capabilities can complement each other. The findings of these
Chapter 6:An Experimental Study to Test and Extend the Theory of Faithful Use of Conceptual Model Combinations 97
studies show that modellers use multiple grammars to overcome construct deficits in any one
individual grammar. Studies on multiple models likewise indicate that using additional
models results in better performance than using a single model (Gemino & Parker, 2009; J.
Kim et al., 2000; Siau & Lee, 2004).
Practitioner surveys indicate that indeed, more than one conceptual model is often used
in most of the systems design and analysis tasks (Davies et al., 2006; Dobing & Parsons,
2008; Fettke, 2009). For example, Recker (2012) reported that over 30% of surveyed process
modellers access additional grammars when modelling business processes. Green et al. (2011)
showed that 80% of users of modelling tool environments select and use multiple grammars
in combination when engaging in conceptual modelling. Dobing and Parsons (2008) reported
that 90% of UML users employ at least two different UML grammars in at least one-third of
their projects.
Overall, these arguments lead us to conclude that: first, system analysts and designers
do use different types of grammars to create different types of models. Second, using multiple
grammars or multiple models may lead to better modelling outcomes. Both conclusions are
intuitive: as targeted systems become more complex, we need more than a single model to
represent them (J. Kim et al., 2000). Moreover, different types of models seem to represent
different aspects of a problem domain, for instance, use case diagrams and class diagrams
(Siau & Lee, 2004). Thus, multiple model use is apparently quite common in practice;
however, we still do not have an understanding regarding how and why that is the case.
6.3 THE THEORY OF FAITHFUL USE OF CONCEPTUAL MODEL
COMBINATIONS
Recker’s (2014) theory of faithful use of conceptual model combinations set out to
provide an explanation for the faithful use of multiple models by analysts and designers. The
primary conjecture of the theory is that faithful use of a combination of conceptual models
will be dependent on two factors: the models’ combined ontological completeness (level of
representational coverage a set of models provides about some real-world phenomenon) and
ontological overlap (level of redundant representations across a set of models) in representing
the focal real-world phenomena. Figure 6.1 provides a visual representation of these two key
factors.
98
Figure 6.1 Illustration of Ontological Completeness and Overlap of Two Conceptual Models (Recker, 2014)
The basic rationale behind this conjecture is that: first, users will select multiple models
to maximise their coverage of some relevant real-world phenomena (Green, 1996). This is
because ontological evaluations of modelling grammars show that any single modelling
grammar suffers from construct deficit (Green et al., 2011; Irwin & Turk, 2005; Siau, 2010).
In other words, no one grammar can be used to model all potentially relevant real-world
phenomena. Moreover, the actual level of completeness of a conceptual model is often less
than its potential level of completeness, because often not all constructs provided via a
grammar are used in any given model (Siau et al., 2005; zur Muehlen & Recker, 2008).
Therefore, users will use multiple models in combination to overcome ontological deficits in a
single model and to maximise combined ontological completeness through multiple models.
Second, users will also seek to minimise overlap in the representation to keep the coverage as
simple and clear as possible (Wand & Weber, 1990a, 1990b, 1993). On the basis of these two
main factors, the theory suggests three propositions that detail the faithful use of multiple
models:
Model selection: model users will select a combination of conceptual models with the
aim to reach maximal ontological completeness; however, they will select additional
models only until a particular level of overlap is reached that they perceive to be
bearable.
Completeness of
domain representation
of model A
Completeness of
domain representation
of model B
Resulting
ontological overlap
between models A
and B
Remaining construct
deficit of combined
representation
Chapter 6:An Experimental Study to Test and Extend the Theory of Faithful Use of Conceptual Model Combinations 99
Domain understanding: the level of understanding users can generate from reading
models will be increased when additional models present more elements of a domain;
however, the level of understanding that can be generated will be moderated by the
level of ontological overlap between models.
Perceived usefulness: Perceptions of the usefulness of model combinations will
increase when the ontological completeness of the achieved representation increases;
however, combinations of models with high ontological overlap between the models
will be evaluated as less useful.
The theory addresses an important and ecologically valid aspect of the use of conceptual
models in practice. The propositions of the theory provide foundations to examine which
combinations of models are faithful for use. Following the guideline proposed by the theory,
in the next section, we describe how we designed and conducted an empirical study to test the
propositions of the theory.
6.4 RESEARCH METHOD
6.4.1 Preliminaries
An experiment is a suitable research method to evaluate propositions of the theory of
faithful use of conceptual model combinations. This is because the theory has not been tested;
therefore, a primary objective is to maximise internal validity (Calder et al., 1981). The theory
suggests three sets of propositions about selection, domain understanding, and usefulness.
Selection is a preferential task, whilst generation of domain understanding is an experiential
task and usefulness perceptions are formed reflectively post hoc. The differences in the nature
of each set of the propositions raise issues with respect to designing an experimental study to
test the propositions effectively. Therefore, a two-part experiment is required: an experimental
design to examine model selection following another experimental design to investigate the
development of domain understanding through using model combinations. A post hoc test
following the second part can then evaluate perceptions of usefulness.
6.4.2 Experimental Design
We designed a two-phase experiment: a free simulation experiment and a 2x2 between-
groups experiment followed by a post hoc test. Each phase is discussed below:
100
In the first phase of the experiment, participants were asked to select conceptual models
for an upcoming model use task (which required developing a deep-level understanding of the
modelled domain). Therefore, in this phase, the experimental design corresponds to a type of
free simulation experiment. In this type of design, no experimental treatment is provided.
Participants were given four types of conceptual models for a domain, and they were asked to
select a combination of models out of a given set of models that would assist them in an
upcoming problem solving task. Participants were free to behave within the required
boundaries of the study (i.e., the prescribed tasks) and they were asked to make decisions and
choices as they saw fit (Figl & Recker, 2014). To measure the selection, we captured: (1) the
types of models participants selected, and (2) the number of models participants added to their
model combinations prior to engaging in the task. We employed two independent variables:
the level of combined ontological completeness the additional model added to the
combination, and the level of ontological overlap the additional model added to the
combination. The dependent variable in this phase was a binary variable that indicated
whether the difference between the level of combined ontological completeness and the level
of ontological overlap of the selected model combination was more than any other possible
combinations.12
In the second phase, we examined domain understanding and perceived usefulness of
model combinations. We implemented a 2x2 between-groups experimental design with two
between-group factors, combined ontological completeness and ontological overlap, each of
which had two levels: high and low. The purpose of having two levels of combined
ontological completeness and overlap is to determine whether different levels of combined
ontological completeness and overlap affect understandability and perceived usefulness of
model combinations in different ways. The two factors implemented by designing four
combinations of two conceptual models that vary in their level of combined ontological
completeness and ontological overlap, similar to the example given by (Recker, 2014). For
instance, in our study, one such combination could be a class diagram with a state machine
diagram (high combined ontological completeness and low ontological overlap). An example
for low combined ontological completeness and high ontological overlap would be a class
12 Note that the independent variables and the dependent variable will be based on the selected second model and
third model.
Chapter 6:An Experimental Study to Test and Extend the Theory of Faithful Use of Conceptual Model Combinations 101
diagram and a use case diagram. To measure the dependent variable, generation of domain
understanding, we employed three measures often used in conceptual modelling experiments,
viz., comprehension test, problem-solving test, and Cloze test.
6.4.3 Materials
To be able to develop and test hypotheses derived from the propositions of the theory,
we needed to apply the propositions to a set of conceptual models of some relevant real-world
phenomena. However, there is a possibility of a learning effect during a two-part experiment.
To avoid the potential learning effect, we developed two sets of models, one set to investigate
model selection and one set to evaluate domain understanding. In developing the conceptual
models, we chose a textbook case simply because the textbook featured a wide selection of
different models for this scenario.
The sets of models we selected are described below, and we then explain how we
manipulated the experimental treatments of combined ontological completeness and
ontological overlap using these models.
Deciding on Case and Models
We selected two different cases: the High Peak Bicycles case, (i.e., a bicycle rental
business that included concepts such as customer, rent, and return) and the Library
Management System case (i.e., a library in a university that had concepts such as students,
borrowing, returning, books, and articles). We adopted both cases from Whiteley (2013, pp.
228-263). We selected these cases for two reasons: first, we could use the instructions
explained in the textbook to develop different types of models to represent different aspects of
the domains; second, these are well-established domains that have also been used in related
prior studies (e.g., Bera et al., 2014) and in IS education.
For each domain we decided to develop four conceptual models, each developed with a
different UML grammar. Figure 6.2 presents these models for the Library Management
System, the models for the High Peak Bicycles case are available in Appendix C. We chose
UML diagrams for several reasons. First, UML is a widely used conceptual modelling method
with widespread adoption in practice and with extensive coverage in IS education (Davies et
al., 2006; Fettke, 2009). Second, UML provides a wide variety of models, each of which
conveys different information (Dobing & Parsons, 2008). Third, a substantial body of
102
research has analysed the relevant grammatical constructs of UML based on Wand and
Weber’s (1993) ontological theory (e.g., Evermann & Wand, 2005, 2006; Opdahl &
Henderson-Sellers, 2002; Siau, 2010). In turn, it was likely that the models would be different
in their coverage of real-world phenomena, and we could draw on established ontological
analyses to evaluate the levels of combined ontological completeness and overlap.
Post-Graduate Student
Borrow Books
Return Books
Pay Fines
Request Research JournalsLibrarian
Maintain Records
Manage Collection
Source New Books
Shelve Books
Calculate Fines
Under-Graduate Student
User
User
-Student ID*
-userName
-userInfo
Undergraduate
student
-Student ID
-Name
-DoB
-Major
-Faculty
-Address
Postgraduate
student
-Student ID
-Name
-DoB
-Major
-Faculty
-Address
-Thesis Topic
-Thesis supervisor
Transaction
-transactionID
Reservation
-dateReserve
-dateAvailable
-studentID*
Copy
-bookID*
-copyNumber
-copyAvailability
1..1
1..*
Book
-ISBN
-bookTitle
-bookAuthor
-Year
-Edition
-Topic
-Category
-Keywords
-Description
0..*
1..1
-bookID
-bookReservation
Loan
-copyNumber*
-dateOut
-dateBack
0..*
1..1
1..1
1..*
Author
-biography
-birthdate
-name
Research Journals
-journalTitle
-publisher
-volume
-issue
0..1
0..*
-author 1..*
1..*
1..*
1..*
Collection
-Name of collection
-relevant student major(s)
-number of books in
collection
0..*
1..*
Bill
-billNumber
-StudentID*
-billDate
-fineAmount0..4
1..1
+createBill()
+updateBill()
+createUser()
+checkUser()
+cancelUser()
+payFine()
+returnBook() +createReservation()
+getReservation()
+checkAvailableDate()
+createLoan()
+getLoan()
+completeDateBack()
+createJournal()
+createAuthor()
+createCopy()
+addBook()
+searchBook()
+displayBookDetails
+updateBookDetails()
+createCollection()
+createTransaction()
+getTransaction()
0..1
0..1
+updateAvailability()
-copyNumber*
+checkUser()
Library Management System
Transaction User
Request an ItemCheck the Request
Calculate Fine
Check Availability
Validate User
[Ret
urn
Book
]
[Request Book]
[Req
uest
Jour
nal A
rtic
le]
[Not
Ava
ilabl
e]
Inform User
[Available or Hired]
[Ava
ilabl
e]
Request Loan
[On
loan
]
Request Reservation
Validate User
Validate User
Make Reservation
Issue Book
Change Book Status
[Not
Val
id o
r Exc
eed
loan
or b
ill]
[Not
Val
id o
r Exc
eed
loan
or b
ill]
Submit Request
[Sto
p th
e or
der]
[Continue to Order]
Authorise Journal Access
[Pos
t-gr
ade
Stud
ent]
[Not Valid]Shelve Book
[Fin
e ==
0]
[Fin
e >
0]
Shelve Book Issue Bill
Change User Status
Check User Status
Check User Status
[Valid & No exceed loan and bill]
[Valid & No exceed loan and bill]
b: Bill
[unpaid]
Save as Debt
Pay the Bill
Change Book Status
b: Bill
[paid]
Change User Status
Change Book Status
Idle Request Submitted Item Searched
User Authorised
User Not Authorised
Book Status Updated
Journal AccessedFine Calculated
Bill Issued
su
bm
it re
qu
est
ch
eck
re
qu
est
[Req=Search]/search item
[Found]/check user validity
[Not Found]
[User valid]
[User not Valid]
[Req==Article]/access article
[Req==Return]/calculate fine
[Fine>0]/issue bill
[Fine==0]/update book status
update book status
[Req==Reserve]/update book status
[Req==Loan]/update book status
Figure 6.2 Four Conceptual Models for the Library Management System Case (Whiteley, 2013, pp. 228-263)
We developed a use case diagram, a class diagram, an activity diagram, and a state
machine diagram for each case based on three reasons:
Chapter 6:An Experimental Study to Test and Extend the Theory of Faithful Use of Conceptual Model Combinations 103
1. We chose a use case diagram, class diagram, activity diagram, and state machine
diagram because together they specify a systems’ function, structure, and behaviour
(Booch, Rumbaugh, & Jacobson, 2010; Burton-Jones & Meso, 2006).
2. The results of our semi-structured interviews indicate that these four diagrams are
the most widely used UML diagrams IS professionals use in practice (Jabbari
Sabegh & Recker, 2017).
3. We selected these diagrams based on their level of combined ontological
completeness and ontological overlap. For example, there wouldn’t be a meaningful
difference between the levels of combined ontological completeness of a
combination of a use case diagram with a sequence diagram compared to the
combination of a sequence diagram with a class diagram. That is, we could not
explain the effect of our treatment if there would be a significant difference
between the performance of the users of combinations of models with the same
level of combined ontological completeness.
6.4.4 Establishing Levels of Combined Ontological Completeness and Ontological
Overlap
To manipulate our treatment variables, we first analysed of the representations offered
by any one model in a given combination of models. First, we performed an interpretation
mapping. That is, we mapped the constructs represented in each model to ontological
constructs that describe things in the real-world (Bunge, 1977; Wand & Weber, 1993). We
then conducted overlap analysis to establish the levels of combined ontological completeness
and ontological overlap for any possible combinations of models.
Interpretation mapping
In carrying out interpretation mappings, we followed established procedures for the
ontological analysis of models, as demonstrated by Recker (2014) and detailed by others (e.g.,
Rosemann et al., 2004; Rosemann et al., 2009). We also followed the multi-coder mapping
procedure, as described in Recker et al. (2010), to establish the reliability of this mapping. We
followed three steps:
104
1. All three researchers individually read the constructs represented in each model
and then interpreted and mapped the ontological constructs to constructs
represented in the models (Recker et al., 2010). In conducting the interpretation
mapping we relied upon previous literature (Wand & Weber, 1993, p. 221).
UML grammars have been evaluated by researchers (e.g., Evermann & Wand,
2001; Evermann & Wand, 2005, 2006; Irwin & Turk, 2005; Opdahl &
Henderson-Sellers, 2002). For each model, we identified grammar constructs
represented in the model. We then evaluated the meaning ascribed to each
construct and the mapping rationale to map it to the corresponding ontological
construct. We did so because not all constructs provided via a grammar are
presented in the models13
(Siau et al., 2005; zur Muehlen & Recker, 2008). For
example, the constructs “system” or “extend” are not represented in any of the
use case diagrams for High Peak Bicycles and Library Management System.
2. The researchers then met to discuss the interpretations of the representation
mapping.
3. Finally, the second revised version of the interpretation mappings was discussed
and refined in several meetings with all of the researchers until a 100%
agreement on mappings was reached.
The results of these processes are represented in Table 6.1 and details of the ontological
evaluation of the conceptual models are presented in Appendix C.
Overlap Analysis
After we implemented interpretation mapping for each of the eight models developed
for both cases, we performed an overlap analysis (Green et al., 2007). To evaluate the level of
combined ontological completeness of any possible combination of models, we counted those
constructs of the models in a combination that corresponded to different ontological
constructs.
13 Moreover, interpretation mappings are not always a 1:1 correspondence of ontological to grammatical
constructs. The details of the mappings are presented in Appendix C.
Chapter 6:An Experimental Study to Test and Extend the Theory of Faithful Use of Conceptual Model Combinations 105
Similarly, we measured the level of ontological overlap of any possible combination of
models by counting those constructs of models in a combination that could be mapped to the
same ontological construct. Table 6.2 shows the results of overlap analysis for the pairwise
combinations of models for cases. Table 6.3 shows the level of combined ontological
completeness and ontological overlap of possible combinations of three models and the
combination of all four models for the High Peak Bicycles case.
Table 6.1 Ontological evaluation of conceptual model constructs - Library Management
System and High Peak Bicycles
Ontological construct Class Use Case Activity State Machine
Thing
Class *(*) *(*) *(*)
Composition Thing *
Property in General *(*)
Hereditary Property *
Binding Mutual property *(*) *(*)
State * *(*)
State Law *(*) *
Stable State *(*)
History
Event *(*)
External Event
Transformation *(*) *(*) *(*) *(*)
Lawful Transformation *(*)
Stability Condition *(*) *(*)
Coupling *(*)
Acts on
System *(*)
Sum out of 17 8 (6) 3 (3) 8 (6) 4 (4)
The corresponding mappings of models represented in the High Peak Bicycles case are presented in brackets.
Table 6.2 Overlap Analysis
High Peak Bicycle Case
Diagram Type Use Case
Diagram Class Diagram Activity Diagram
State Machine
Diagram
Use Case Diagram 3 2 1
Class Diagram 6 2 1
Activity Diagram 7 10 2
State Machine Diagram 6 9 8
Library Management System
Diagram Type Use Case
Diagram Class Diagram Activity Diagram
State Machine
Diagram
Use Case Diagram 3 2 1
Class Diagram 8 3 1
Activity Diagram 9 13 3
State Machine Diagram 6 11 9 Combined ontological completeness of the combinations is given in the dark grey cells below the diagonal; ontological overlap is given in the light grey cells above
106
Table 6.3 MOC and MOO for combinations of three and four models - High-Peak Bicycles
Case
Use Case
Diagram
Use Case
Diagram
Use Case
Diagram Class Diagram
Use Case
Diagram
Class Diagram Class Diagram Activity Diagram Activity Diagram Class Diagram
Activity Diagram State Machine
Diagram
State Machine
Diagram
State Machine
Diagram Activity Diagram
State Machine
Diagram
MOC 10 9 9 12 12
MOO 3 3 3 3 4
6.4.5 Measures
It is difficult to test domain understanding directly. For this reason, we used three
measures – models comprehension test, problem-solving questions, and a Cloze test – to
assess domain understanding. The comprehension test was used to determine a users’ ability
to glean information from the models. The performance on the comprehension test was
measured by the number of correct answers given to eight comprehension questions. The
problem-solving test was used to examine using knowledge gained from reading the
combination of models to solve a problem not directly answerable from models. A typical
problem-solving question defines a problem in the domain, for example, an element that is not
working (Burton-Jones & Meso, 2008). We devised three problem-solving tasks. Participants
were asked to: 1) describe the main functionality of the system, 2) identify a problem in the
system and describe how it occurred, and 3) to provide a solution for a problem based on their
understanding. The performance on the problem-solving test was measured by the total
number of correct answers given to the three problem-solving questions (Gemino & Wand,
2003, 2005). After the problem-solving test, the models were removed. In the Cloze test,
participants were given a passage about the domain to read; however, some of the words in
the passage were missing. Thus, the Cloze test was about participants’ ability to recall
information gained from the models to fill in the blanks of a textual description of the domain
(Masri et al., 2008). The performance on the Cloze test was measured by the number of
blanks filled in with correct answers in the Cloze test (Burton-Jones & Meso, 2006; Recker &
Dreiling, 2011).
After completing the second phase of the experiment, a post hoc test was designed to
measure the perceived usefulness of model combinations to complete the assigned tasks. To
Chapter 6:An Experimental Study to Test and Extend the Theory of Faithful Use of Conceptual Model Combinations 107
measure participants perceived usefulness of model combinations we adopted measurement
items from Recker and Rosemann (2010b). The measurement items for the dependent
variables are presented in Table 6.4.
Table 6.4 Measurement instrument for dependent variables
Measurement Items for the Dependent Variable - Model Selection
MOC and MOO
= 1, if (combined ontological completeness - ontological overlap) is maximum compared to
any other possible model combinations
= 0, otherwise
Measurement Items for the Dependent Variable - Comprehension
1. Are there any limitations for a student to extend a loan?
2. Can a post-graduate student borrow more than six research journals at the same time?
3. Can a user of a library be an employee of the university? 4. Do students need to pay fines to access research journals?
5. Can a student request books that are loaned by someone else?
6. Are there a limited number of copies of each book in the library?
7. Can an undergraduate student borrow a book and a research journal at the same time?
8. Do students need to pay fines for a delay in returning books?
Measurement Items for the Dependent Variable- Problem Solving
1. Explain the main functionality of the library management system. Write down as many supported activities
as you can.
2. Under what circumstances would a student fail to borrow an item s/he needs? Write down as many possible
reasons as you can think of.
3. How can the library management system be improved to ensure that students in each major can have access
to the relevant books and journals they need? Write down as many possible solutions that should be
considered.
Measurement of the Performance on Cloze Test
The library management system is designed to track books and journal articles in the library of a university. It
also provides functions to …(1)… and control orders and loans, and calculate fines for …(2)… . Users of library
management system are either …(3)… or …(4)… .
The library has books in a wide variety of areas and topics. Normally, there are multiple …(5)… of a particular
book in the library. The library also provides access to research journals.
…(6)… do not have access to research journals . However, books in the library are categorised in specific
…(7)… to be sure that the library provides relevant books for students in different majors.
Users can search items using title, authors name, topic and any keywords related to their majors. In fact, …(8)…
can search items in the library, but only …(9)… can borrow books or have access to journal articles.
The results of a search for a book can be threefold: … (10) …, …(11)…, or …(12)…. If a book is on loan by
someone else, a user can still …(13)… the book. There are no limitations on how many …(14)… a …(15)…
student can have access at the same time.
Measurement Items of Perceived Usefulness
On a 1-7 scale (“Strongly disagree” to “Strongly agree”)
1. Overall, I found the combination of diagrams useful for the given tasks.
2. I found the combination of diagrams useful for achieving the purpose of the given tasks.
3. The combination of diagrams helped me in meeting the objectives of the given tasks.
Responses to the problem-solving questions and the cloze test were coded14
by two
independent graduate students. We prepared an extensive coding scheme that described the
coding procedure and provided examples of correct answers for the questions. The first author
14 For details on coding of the participants’ responses to problem solving questions and the Cloze test, please see
Appendix C.
108
explained the coding procedure to both coders. Before starting to code the responses, the
coders coded a sample of responses for both problem solving-questions and the cloze test.
Upon completion of the sample coding, the author and the coders discussed the rationale and
coding to ensure that the coders fully understood the nuances of the responses and how they
were to be coded (Khatri et al., 2006). The coders then independently coded the participants’
responses. After the coding procedure was completed, both coders and the first author met
and compared and discussed their coding to try to arrive at a consensus regarding the coding;
they then changed their coding accordingly. This process was repeated until an acceptable
reliability score was achieved. The inter-coder reliability was assessed using both raw
agreement and the kappa statistic (Cohen, 1988). After multiple rounds of coding, the raw
agreement achieved for the problem solving was 0.83, and the kappa statistic was 0.81 (i.e.,
81% agreement between coders). The cloze test was much more objective. The raw agreement
for the Cloze test was 0.92 while the kappa statistic was 0.91 (i.e., 91% agreement between
coders). The kappa value of more than 0.81 is almost perfect. However, the coders and the
first author met and discussed and refined their coding until a 100% agreement was obtained
between the two coders.
6.4.6 Covariates
To eradicate potential exogenous factors and evaluate rivalling hypotheses throughout
both phases of the experiment, we also measured several control variables, namely individual
difference factors, and cognitive load (Figl, Mendling, et al., 2013; Khatri et al., 2006; Masri
et al., 2008). We operationalised individual differences by examining participants’ prior
experience in conceptual modelling and familiarity with the conceptual models used in the
experiments. The rationale was to ensure that participants had experience of using relevant
conceptual models and to evaluate their levels of experience (Figl & Recker, 2014).
Familiarity with models was also used as a potential exogenous variable, which could affect
the usage of a model (Figl & Recker, 2014; Recker, 2010a) because participants with high
familiarity may have found using a combination of conceptual models easier than participants
with low familiarity, because of inherent preferences. To that end, we adopted the model
familiarity test from Figl and Recker (2014). Likewise, we measured perceived experienced
cognitive load during the tasks (Marcus, Cooper, & Sweller, 1996). We did so because a
complex combination of models (Figl, Recker, et al., 2013) and high ontological overlap
Chapter 6:An Experimental Study to Test and Extend the Theory of Faithful Use of Conceptual Model Combinations 109
(Moody, 2009) is likely to increase cognitive load and may adversely affect the development
of understanding. We measured experienced cognitive load using a subjective rating scale
(Figl, Mendling, et al., 2013) after the second part of the experiment. The measurement
instruments developed to measure concepts of relevance to our study are available in
Appendix C.
6.4.7 Participants
The target population for the experiments comprised users of conceptual models. The
participants of this study were 131 undergraduate students in a large university in Australia
who completed courses related to systems analysis and design and conceptual modelling of
real-world domains. Students were regarded as reasonable proxies of junior working experts
as a sample for this study, first, because prior studies have found little difference between
students and junior professionals (Arisholm & Sjøberg, 2004). Second, this research was
relevant to the students who passed UML related courses because they had covered the
requirements of analysis, communication, system design, and modelling concepts. Finally,
using students is relatively common in conceptual modelling research (e.g., Bera et al., 2014).
Participation was voluntary, and remuneration and incentives were provided for
participants in the study. The experiment was conducted using an online experimentation
system in computer labs between September and November 2016 and was monitored to
ensure that individuals completed the experiment independently.
6.4.8 Procedures
Participants started with pre-tests that collected users’ demographic information and
UML familiarity levels. A short tutorial then introduced the basics of a use case diagram,
class diagram, state machine diagram, and activity diagram. No time limitation was placed on
the tutorial and performing the tasks. Figure 6.3 presents an overview of the experimental
procedures.
110
Pre-test
Modeling Familiarity
UML Diagrams Familiarity
Domain Familiarity
Tutorials
Use Case Diagram, Class
Diagram, State Machine
Diagram, Activity Diagram
Model Selection
High Peak Bicycles
Comprehension
Library Management System
Problem Solving
Library Management System
Cloze
Library Management System
Post hoc test
Perceived Usefulness
Library Management System
Remove models
Fir
st
ph
ase
Se
con
d p
hase
Figure 6.3 Overview of experimental procedure
First Phase
The first part of the experiment began with the free simulation experiment designed for
model selection. Participants were then instructed about the upcoming task and asked to make
selection decisions for this task, similar to the study by (Figl & Recker, 2014). The given task
was twofold: a) to identify reasons for a problem in the domain, and b) to suggest solutions
for a problem in the domain. We believe that if a user is asked a problem-solving question
that requires a deep understanding, he or she is more likely to examine different models
(Burton-Jones & Meso, 2008). Participants were provided with a use case diagram, class
diagram, state machine diagram, and an activity diagram presenting the High Peak Bicycle
case. Participants knew the types of models, but they could only see the models once selected.
This ensured that participants only used the models that they selected, and allowed us to
control for potential learning effects from unselected models. Through multiple pilot tests, we
learned that if participants were presented with all of the diagrams, they would still glean
information from the models they did not add to their selection.
Participants were free to start with any model to answer the problem-solving questions.
Participants could continue to add models to their selections if they felt they needed more
models to be able to answer the questions. However, the selection of models was stepwise. In
each step, participants were asked to either complete the task if they felt that the selected
Chapter 6:An Experimental Study to Test and Extend the Theory of Faithful Use of Conceptual Model Combinations 111
model/s provided enough information to accomplish the task or to select additional models if
they needed to do so. In the latter case, the problem-solving questions and the remaining
model types were displayed, and participants were asked to choose one additional model.
These steps were repeated if participants continued to prefer to select a third or fourth model.
After the fourth model, participants were asked whether they felt they could complete the task
using all four models or if they were not sure they could answer the questions.
Participants were asked to select one model in each step and continue to add models in
the following steps. This stepwise procedure was implemented to observe the users’ selection
procedure of models. It also enabled us to examine whether participants compromised the
level of combined ontological completeness. Participants might not aim to have a
combination of models with a higher level of combined ontological completeness when the
addition of more models increases ontological overlap.
At the end of this phase, participants were asked to rate the extent to which they relied
on the information in each model they selected, their own knowledge, and the extent to which
they made assumptions in answering the two problem-solving questions. Participants were
also asked to identify which model/s they would like to remove if they had an opportunity to
remove a model, or which models they felt were not useful in answering the questions. These
measures allowed us to gather data for control checks and to measure cognitive engagement
with the models.
Second Phase
For the second phase of the experiment, participants were randomly assigned to four
groups. Table 6.5 presents the distribution of the participants across the treatment groups.
Each group had between 31 and 34 participants.
Table 6.5 Distribution of participants in each group
Combined Ontological Completeness
Low High
Ontological Overlap
Low
Use Case +
State Machine
(N=34)
Class +
State Machine
(N=33)
High
Class +
Use Case
(N=31)
Class +
Activity
(N=33)
112
Each group was given a combination of two models with different levels of combined
ontological completeness and ontological overlap (high or low). Each participant completed
three tasks (models comprehension task, problem-solving task, and Cloze test). After the
problem-solving test, the models were removed. The sequence of tasks was fixed to ensure
the internal validity of the experiment (Gemino & Wand, 2005). For all tasks, participants in
each group were given the same combination of two models representing the library
management system. Similar to the first phase, in the second phase, we used four UML
models (use case diagram, class diagram, activity diagram, and state machine diagram)
describing the Library Management System case. The models were described in Section 6.4.1.
Post Hoc Test
After the Cloze test, participants were asked to answer the post hoc test about the
perceived usefulness of the model combinations. Participants finished the experiment by
answering the questions about cognitive load.
6.5 HYPOTHESIS DEVELOPMENT
We developed our hypotheses in three parts following the model selection, domain
understanding, and perceived usefulness propositions. We begin with the selection
proposition.
6.5.1 Selection of Model Combination
The use of conceptual models occurs as part of a particular task (Recker, 2014) and a
decision on which model or model combination to use is characterised by task goals. A
selection decision about which model to use is influenced by performance outcome
expectations (Compeau, Higgins, & Huff, 1999). On the assumption that several models about
a real-world domain are available, the selection proposition concerns which models would be
selected by users to maximise the level of domain understanding they can generate to
complete an upcoming task (Recker, 2014). Recker’s (2014) theory suggests model users will
select a set of models from a number of available models to increase the ontological
completeness of representation of some focal real-world phenomena because any one model
will have construct deficits. Users will do so because they have a desire to have a more
complete representation that provides all relevant information about a real-world domain that
users will need.
Chapter 6:An Experimental Study to Test and Extend the Theory of Faithful Use of Conceptual Model Combinations 113
Users make decisions about which model to use based on an evaluation of the potential
value of alternatives (Thompson, 1967), which leads them to achieve upcoming task goals.
The outcome expectations of using conceptual models is to generate a complete and clear
understanding of real-world phenomena (Weber, 1997). Users create a mental model of real-
world phenomena based on the information directly presented in the models (Gemino &
Wand, 2003). The outcome expectation of using models is to create understanding of the
domain that requires users to develop a more complete mental model (Masri et al., 2008) and
a more complete mental model demands more information. This reasoning suggests that,
given one starting model out of the four presented diagram types, users will select an
additional model in which the selected model provides users with more information compared
to the alternatives. Therefore, given a set of models, users will select additional models for
task completion such that the ontological completeness of the model combination is
maximised. For our experimental materials, we propose that users would select an additional
model for use according to the following rules:
H1: Given any model participants start with, they will select an additional model/s
that add/s maximum ontological completeness to the model combination.
By increasing the level of ontological completeness of the representation of a domain,
more information will be available for users to integrate a mental model of more aspects of
the represented real-world phenomenon. On the other hand, through selecting additional
models, the ontological overlap between models may also increase (Green et al., 2007).
Increasing ontological overlap between models will decrease the clarity of representation of
the real-world phenomena because several representations are in the set of models that
represent the same real-world phenomenon (Wand & Weber, 1993). It is likely that high
levels of ontological overlap will lead to misunderstanding, because users need to identify
overlapping constructs and reconcile their meaning.
The increase in the level of ontological overlap will also increase the complexity of the
representation of the domain (Gemino & Wand, 2005). If we consider a simple representation
of a domain as a representation without ontological overlap, based on Occam’s razor
(Domingos, 1999), which argues that simplicity is a goal itself, model users will avoid adding
more ontological overlap to the representation of the domain by selecting additional models.
However, the outcome expectation of model use is to create a complete understanding of a
114
domain, therefore adding more information, increasing ontological completeness, will
improve domain understanding. Therefore, we argue that model users will also follow the law
of parsimony: the least number of different models with the most explanatory power to
represent the real-world domain of interest will be used in combination. Based on these
reasonings, we argue that model users will use additional models to reach maximal level of
completeness only if adding one more model will increase the level of combined
completeness and is accompanied by a minimal level of increase in the overlap.
We also argue that developing a mental model is limited to the capability of individuals’
working memory. That is, an increase in ontological overlap between models will require
additional reasoning to develop understanding and increase the cognitive load (Sweller &
Chandler, 1994). The burden on the limited capacity of working memory can be reduced by
using models that present multiple constructs that can be interpreted as a single element (Paas,
Tuovinen, Tabbers, & Van Gerven, 2003; Soffer, Wand, & Kaner, 2015); therefore, they
require further integration to construct a mental model that increases cognitive load and
reduces efficiency in creating a clear and accurate understanding. Thus, we argue that model
users will not select an additional model if adding another model only increases the
ontological overlap.15
In the case of our materials, we therefore believe the following logic
would be used.
H2: Given any model participants start with, they will select an additional
model/s that add/s a minimum ontological overlap to the model combination.
6.5.2 Development of Domain Understanding
Recker’s (2014) theory argues that the level of domain understanding that can be
generated from the reading of a model combination will be a function of the combined
ontological completeness of the model combination. In fact, one of the primary objectives of
using conceptual models is to create understanding of some aspects of the social and physical
world around us (Gemino & Wand, 2004). Any model-based task requires constructing
knowledge from the information presented in the model and integrating that knowledge with
15 Recker’s (2014) theory also argues that users will select additional models until they reach a bearable level of
ontological overlap. The bearable level of ontological overlap is constrained by users’ processing capability. If
this capability is maximised, users will not select additional models even if adding another model would increase
combined ontological completeness. However, users processing capability is volatile and contextual; therefore,
we have no basis to speculate ex ante what the bearable threshold could be; thus, we cannot offer a hypothesis on
this element of the proposition.
Chapter 6:An Experimental Study to Test and Extend the Theory of Faithful Use of Conceptual Model Combinations 115
the mental model developed by using the model. A combination of models will provide more
representation elements that convey meaning about the phenomena in a domain if and when
the combination has higher ontological completeness than any one model alone. More
information is then available for assimilation into the users’ mental model about all relevant
aspects of focal real-world phenomena (Recker, 2014). This combination should improve
domain understanding.
Moreover, interpretation of a model for domain understanding is a cognitive process
(Bodart et al., 2001; Gemino & Wand, 2005). The existence of a construct deficit in any
single conceptual model results in a lack of relevant information presented in the model.
Based on cognitive load theory, the way information is presented, that is, lack of information
in a domain presentation, can increase extraneous cognitive load (Kirschner, 2002). A
combination of models with high ontological completeness will not only reduce extraneous
cognitive load, but also present different forms of information (Burton-Jones & Meso, 2008)
(e.g., structural and behavioural). The availability of different forms of information may
promote germane cognitive load, and as a result facilitated learning and understanding. We
propose that model users who read a combination of models with high levels of combined
ontological completeness will generate higher levels of domain understanding than users who
read a combination of models with a low level of combined ontological completeness; thus
we developed the following hypotheses:
H3a: Users of the combination of models with a high combined ontological
completeness will achieve a higher score in the comprehension test compared to
the users of the combination of models with a low combined ontological
completeness.
H3b: Users of the combination of models with a high combined ontological
completeness will achieve a higher score in the problem-solving test compared to
the users of the combination of models with a low combined ontological
completeness.
H3c: Users of the combination of models with a high combined ontological
completeness will achieve a higher score in the Cloze test compared to the users
of the combination of models with a low combined ontological completeness.
Recker’s (2014) theory of faithful use of conceptual model combinations proposes that
an increase in the level of ontological overlap in a model combination will impede the
positive effect of combined ontological completeness. This is because, while a combination of
models with a high level of combined ontological completeness and a high level of
116
ontological overlap will provide more information than a single model, this achievement often
comes at the cost of decreased clarity if and when additional models overlap partially (i.e.,
they share at least partially redundant information with an existing model). In such a situation,
reading combinations of models with ontological overlap will require additional cognitive
effort to develop domain understanding, because users will need to identify and discriminate
elements that appear in multiple models and that convey the same meaning about some real-
world phenomena. Thus, this may increase the extraneous cognitive load experienced by
model users (Gemino & Wand, 2005). In turn, users will have less cognitive capacity
available to receive and integrate relevant information from models. We propose that the
positive impact of combined ontological completeness of model combinations on users’
ability to generate domain understanding would be decreased by the extent of ontological
overlap in the combination of models; thus leading to the following hypotheses:
H4a: Users of the combination of models with a high ontological overlap will
achieve a lower score in the comprehension test compared to the users of the
combination of models with a low ontological overlap.
H4b: Users of the combination of models with a high ontological overlap will
achieve a lower score in the problem-solving test compared to the users of the
combination of models with a low ontological overlap.
H4c: Users of the combination of models with a high ontological overlap will
achieve a lower score in the Cloze test compared to the users of the combination of
models with a low ontological overlap.
6.5.3 Perceived usefulness of model combinations
Finally, we examined the perceived usefulness of multiple models for a given task. The
ontological completeness of a modelling grammar has a direct effect on perception about the
usefulness of the grammar (Recker et al., 2011). We argue that similar mechanisms will affect
the perceived usefulness of conceptual models (Maes & Poels, 2007). Perceived usefulness
refers to the extent that model users believe a model combination helps them to achieve task
goals (Davis, 1989; Maes & Poels, 2007; Recker et al., 2011). We assumed that one of the
primary objectives of using models is to create complete and clear domain understanding
(Recker, 2014). Construct deficit that exists in a single model results in the lack of relevant
information in which a model reader seeks to develop a mental model about the domain.
Similarly, a model combination with a low level of combined ontological completeness
provides a representation with some level of relevant information about the domain.
Chapter 6:An Experimental Study to Test and Extend the Theory of Faithful Use of Conceptual Model Combinations 117
Therefore, model readers require additional cognitive demand that will reduce their
understanding.
Model combinations with high levels of combined ontological completeness will be
perceived as more useful (Recker, 2014) because high combined ontological completeness
provides more information, and therefore greater ability for users to manifest relevant aspects
of real-world phenomena, which we call the direct information effect; thus leading to the
following hypothesis:
H5: Users of the combination of models with a high combined ontological
completeness will perceive the combination as more useful than the users of the
combination of models with a low combined ontological completeness.
Yet, in model interpretation tasks, a combination of models with high combined
ontological completeness and high ontological overlap will be evaluated as less useful,
because the representation with ontological overlap will undermine the gains in
representational coverage, in turn making the model combination less useful than a less
complete but less overlapped combination of models (Recker, 2014). This is because a high
level of ontological overlap in the model combination decreases the clarity (Wand & Weber,
1993) and increases the complexity of the representation of the domain (Gemino & Wand,
2005). The additional added complexity of the representation with ontological overlap will
increase the cognitive effort of interpretation of the information presented in the models
(Sweller, 1994). The high cognitive load will decrease the user’s capacity for cognitive
processing of information (Mayer, 1989). The high cognitive effort will undermine the
perception of usefulness (Recker, 2010a); we therefore hypothesised that:
H6: Users of the combination of models with a high ontological overlap will
perceive the combination as less useful than the users of the combination of the
models with a low ontological overlap.
6.6 DATA ANALYSIS
We analysed our data in four steps. First, we checked the results of the reliability and
homogeneity of variances. We then tested the model selection hypotheses. In the third step,
we tested the domain understanding hypotheses, and in the fourth step, we analysed the
perceived usefulness hypotheses.
118
6.6.1 Preliminary Tests
Our data analyses began with screening the data for its conformance with the
assumptions of the designed tests. We started by assessing the reliability and validity of the
Likert-type measures of the pre-test modelling familiarity measures, cognitive load, and post
hoc test, perceived usefulness measures through an exploratory factor analysis implemented
in IBM SPSS Statistics 22 (Tabchnick & Fidell, 2007). The results of factor analyses
indicated that one measurement item of the cognitive load needed to be eliminated due to low
loading (0.48). Therefore, we used three measurement items that appropriately captured the
cognitive load in engaging with the experimental tasks. All measurement items for modelling
familiarity and perceived usefulness showed an acceptable factor loading of more than 0.65
(Fornell & Larcker, 1981). Table 6.6 shows the average factor loadings and Cronbach’s alpha
for the measurement items of the three constructs. Appendix C summarises the factor loading
for each measurement items.
Table 6.6 Reliability and validity analysis of scaled measures
No. Items Average. F. Load Cronbach's Alpha
Modelling Familiarity 4 .774 .778
Perceived Usefulness 3 .860 .822
Cognitive Load 3 .837 .806
The homogeneity of variances assumption was verified using the Levene test. Table 6.7
shows the results. Levene statistics for each of the dependent measures suggest that the
hypothesis of equal variances was not rejected across any of the variables at the 0.05 level.
Table 6.7 Test of homogeneity of variances for dependent measures
Measure Levene Statistic df1 df2 Sig.
Comprehension .050 3 127 .985
Cloze Test 1.108 3 127 .348
Problem Solving .669 3 127 .572
Perceived Usefulness .433 3 127 .730
Cognitive Load 1.173 3 127 .323
6.6.2 Hypothesis Testing: Model Selection
Selection of a second model
One hundred and seven (82%) of 131 participants selected more than one type of
models. Only twenty-four of 131 participants selected one model to achieve the goals of the
upcoming task in the model selection experiment. Table 6.8 summarises this information.
Chapter 6:An Experimental Study to Test and Extend the Theory of Faithful Use of Conceptual Model Combinations 119
Table 6.8 Frequency of participants and the number of models in combination
One model Two models Three Models Four Models All
24 31 41 35 131
18.32% 18.32% 23.66% 31.3% 100
Table 6.9 shows the descriptive statistics of the number of participants and the types of
models they selected as the first and second choice. The most widely selected model as a first
choice was the activity diagram. The activity diagram was also the most widely selected
model as the second choice. The least frequently selected model was the class diagram. Table
6.10 shows the mean and standard deviation of the constructs investigated in the model
selection. On average, users selected a second model that added almost three levels of
ontological completeness to the representational coverage of the model combination.
Table 6.9 Frequency and descriptive statistics for second model selection
Started Model # (%) Second Model #(total)
Model A) Use Case Diagram 47 (35.88)
Model B) Class Diagram 10 (19)
Model C) State Machine Diagram 9 (22)
Model D) Activity Diagram 22 (36)
None 6 (24)
Model B) Class Diagram 16 (12.21)
Model A) Use Case Diagram 5 (30)
Model C) State Machine Diagram 1 (22)
Model D) Activity Diagram 7 (36)
None 3 (24)
Model C) State Machine Diagram
19 (14.5)
Model A) Use Case Diagram 6 (30)
Model B) Class Diagram 3 (19)
Model D) Activity Diagram 7 (36)
None 3 (24)
Model D) Activity Diagram 49 (37.40)
Model A) Use Case Diagram 19 (30)
Model B) Class Diagram 6 (19)
Model C) State Machine Diagram 12 (22)
None 12 (24)
Table 6.10 Mean and standard deviation of constructs for model selection
Constructs Mean SD
Domain Familiarity 3.43 1.38
Modelling Familiarity 4.1939 .911
Familiarity with Selected Diagram .41 .49
Level of Completeness Added by the Second Model 2.84 1.23
Level of Overlap Added by the Second Model 1.95 .57
120
The correlation between constructs is presented in Table 6.11. The results of the
correlation matrix indicate that the selection of the second model had a statistically significant
positive correlation with the level of combined ontological completeness added by the model,
while it also had a statistically significant but negative correlation with the level of
ontological overlap added by the model.
Table 6.11 Correlation matrix for selection of second model
Constructs 1 2 3 4 5 6
1 Domain Familiarity 1
2 Modelling Familiarity 0.174 1
3 Familiarity with the Selected Model -0.150 0.041 1
4 Level of Completeness Added by the Second
Model -0.159 -0.86 0.077 1
5 Level of Overlap Added by the Second Model 0.026 -0.14 .202* -0.131 1
6 MOC and MOO -0.049 -0.012 -0.066 .716** -.355** 1
We conducted logistic regression analysis to test the predictors of the selection of the
second model using the level of combined ontological completeness and the level of
ontological overlap added by the second model, familiarity with the selected model,
modelling familiarity, and familiarity with the domain as covariates.
The results of model fit, as presented in Table 6.12, suggest that between 59.3 percent
and 79.4 percent of the variation in selecting the second model can be explained by the set of
independent variables (Pallant, 2007). We ran hierarchical logistic regression analysis with
and without the ontological variables. Based on the results presented in Table 6.12, the ability
of the model to predict the selection of the second model was higher when we used
ontological factors. The Wald test (Pallant, 2007) also supports that the level of completeness
added by the second model (β = 4.006, p = 0.00) and the level of overlap added by the second
model (β = - 4.217, p = 0.014) explains the selection of the model to achieve the maximal
combined ontological completeness and minimal ontological overlap. In line with Hypotheses
1 and 2, the results indicate that the more ontological completeness the second model added to
the combination, the more likely it was that the users would select the model. However, the
negative beta value for the level of ontological overlap the model added to the combination,
indicating that the more ontological overlap the model added, the less likely a user was to
select the model to add to the combination.
Chapter 6:An Experimental Study to Test and Extend the Theory of Faithful Use of Conceptual Model Combinations 121
Table 6.12 Hierarchical logistic regression for selection of second model
B S.E. Wald Sig. Exp (B)
Model One
Domain Familiarity -.090 .146 .378 .538 .914
Modelling Familiarity .005 .219 .001 .982 1.005
Familiarity with Selected Model -.311 .404 .595 .441 .733
Constant .206 .995 .043 .836 1.229
Model Two
Domain Familiarity -.118 .259 .210 .647 .888
Modelling Familiarity .736 .427 2.969 .085 2.087
Familiarity with Selected Model -.516 .768 .451 .502 .597
Level of Completeness Added by the Second Model 4.006 1.127 12.631 .000 54.930
Level of Overlap Added by the Second Model -4.217 1.713 6.056 .014 .015
Constant -8.232 3.197 6.628 .010 .000
Model One Model Two
Chi-square 2.99 3.85
-2 Log Likelihood 146.34 50.93
Cox & Snell R Square 0.008 0.593
Nagelkerke R Square 0.011 0.794
Selection of a third model
Based on the data presented in Table 6.8, 76 participants selected a third or fourth
model to add to the combination. We evaluated the determinant factors for selection of the
third model. Table 6.13 presents the correlation between the factors. Similar to the selection
of the second model, the correlation matrix indicates that the selection of the third model had
a statistically significant positive correlation with the level of combined completeness added
by the third model.
Table 6.13 Correlation matrix for the selection of the third model
Constructs 1 2 3 4 5 6
1 Domain Familiarity 1
2 Modelling Familiarity 0.142 1
3 Familiarity with Selected Diagram 0.001 0.173 1
4 Level of Completeness Added by the Third Model -0.031 -0.115 0.142 1
5 Level of Overlap Added by the Third Model 0.010 -0.003 -0.024 -.266* 1
6 MOC and MOO -0.094 -0.14 0.192 .628** 0.205 1 *. Correlation is significant at the 0.05 level (2-tailed). **. Correlation is significant at the 0.01 level (2-tailed).
To test how the level of combined ontological completeness and ontological overlap
added by the third model could predict the selection of the model, we conducted hierarchical
logistic regression. As presented in Table 6.14, and similar to the selection of the second
122
model, the model with ontological factors was stronger in explaining the variations in the
selection of the third model (dependent variable: MOC and MOO of the combination of three
models).
Table 6.14 Hierarchical logistic regression for the third model selection
B S.E. Wald Sig. Exp(B)
Model One
Domain Familiarity -.100 .174 .333 .564 .904
Modelling Familiarity -.434 .270 2.592 .107 .648
Familiarity with Selected Model 1.268 .601 4.454 .035 3.554
Constant 2.254 1.223 3.399 .065 9.529
Model Two
Domain Familiarity -.179 .334 .287 .592 .836
Modelling Familiarity -.179 .566 .100 .752 .836
Familiarity with Selected Model 1.819 1.155 2.481 .115 6.168
Level of Completeness Added by the Third Model 4.113 1.188 11.975 .001 61.108
Level of Overlap Added by the Third Model 5.182 2.089 6.154 .013 178.086
Constant -.179 .334 .287 .592 .836
Model One Model Two
Chi-square 6.84 6.35
-2 Log Likelihood 97.74 32.99
Cox & Snell R Square 0.072 0.596
Nagelkerke R Square 0.097 0.808
Table 6.14 also shows the results of logistic regression to assess the impacts of
covariates and the treatment variables on the selection of the third model. Based on the results
of the Wald test, both the level of combined ontological completeness (β = 4.113, p = 0.001)
and the level of ontological overlap (β = 5.182, p = 0.013) the model added to the
combination had a statistically significant effect on the selection of the third model. However,
unlike the selection of the second model, the level of ontological overlap added by the third
model had a positive effect on the selection of the model. This may be due to the fact that in
selecting a third model, only four options out of 24 possible remaining options existed where
model users could select a model without ontological overlap. For any other options, the third
model added at least one level to the ontological overlap of the model combinations.
Therefore, we conducted a post hoc analysis to investigate the selection of more than two
models in more detail.
Chapter 6:An Experimental Study to Test and Extend the Theory of Faithful Use of Conceptual Model Combinations 123
Post hoc test
To better understand the behaviour of participants in model selection, we looked at
participants’ responses to two questions: first, which models they relied on in answering the
questions, and second, whether they would like to remove any diagram/s from the
combination they had already selected. Table 6.15 shows the number of models participants
selected and the number of models they used in combination to answer the problem-solving
questions.
Table 6.15 Number of models selected and number of models they used in combination to
answer the questions
No. of models
selected
No. of
participants
No. of models used
in combination
No. of
Participants
Most frequently used
model combination
Three models 41
3 9 use case, class and activity
2 25 activity and use case
1 7 activity
Four models 35
4 2 + 7 all four models
3 17 use case, state, activity
2 6 activity and state/use case
1 3 activity
As presented in Table 6.15, only 35 participants used three or four models in
combination. Seventy-six percent of the participants who selected a third or fourth model
would have liked to remove at least one of the models they had selected.
Out of 41 participants, only nine participants who selected the third model used all three
models in combination to answer the questions, while 32 participants would have liked to
remove some of the models. Overall, out of 131 participants, 26 participants used three
models in combination to answer the questions, and 73.3% of participants used two or less
models in combination. Another dominant finding was that only two participants who
selected four models used all four models to answer the questions and seven mentioned that
they were not sure whether they could answer the questions using all four models. If we
remove those seven participants, 93% of the participants who selected the fourth model used
three or less models in combination to answer the questions. Table 6.16 shows the number of
users and types of models selected and dropped.
124
Table 6.16 Frequency of types of models
Types of Models Number of Users
Selected Dropped Used
Use Case 99 26 73
Class 67 25 42
State Machine 73 21 52
Activity 110 4 106
These findings indicate that participants would have liked to remove some models from
the combination, and as a result decrease the combined ontological completeness and also the
ontological overlap.
6.6.2 Hypothesis Testing: Domain Understanding
We began by screening the data collected in the second part of the experiment. Table
6.18 presents descriptive statistics and Table 6.19 shows the correlation matrix. We ran an
ANOVA test to compare the means for modelling familiarity, diagram familiarity, and
domain familiarity across groups. The results show that there were no statistically significant
differences across groups in terms of the level of the familiarity with modelling, diagrams,
and domains (see Table 6.17). These results indicate that the participants were effectively
randomised across treatments.
Table 6.18 shows that the standard deviations of the means for modelling, diagrams,
and domains familiarity across groups were almost more than one unit. This indicates that
there was sufficient variation in the level of familiarity within the groups. From the
correlation matrix, we noted that performance on the comprehension test and ontological
overlap had a statistically significant correlation. We also noted that cognitive load had a
statistically significant correlation with modelling familiarity (at the 0.01 level) and with
UML diagrams familiarity (at the 0.05 level). We then evaluated the performance of
participants in domain understanding. Responses to the comprehension test were evaluated
based on the number of correct answers to the eight multiple choice questions.
Table 6.17 ANOVA test to compare familiarity difference between groups
Sum of Squares df Mean Square F Sig.
Familiarity Self Rate Library 1.682 3 .561 .350 .789
Familiarity Self Rate 5.203 3 1.734 1.967 .122
Familiarity Diagrams Test Correct Answers 7.113 3 2.371 1.278 .285
Chapter 6:An Experimental Study to Test and Extend the Theory of Faithful Use of Conceptual Model Combinations
125
Table 6.18 Descriptive statistics
Constructs Modelling
Familiarity
Diagram
Familiarity
Domain
Familiarity
(Library)
Cognitive
Load Comprehension
Problem
Solving Cloze
Perceived
Usefulness
Scale 1-7 Count. 1-7 1-7 Count. Count. Count 1-7
Group A
M 4.48 2.42 4.09 4.46 3.73 8.79 6.76 4.97
SD 0.81 1.56 1.21 0.92 1.42 2.33 3.61 1.23
Group B
M 3.92 1.81 3.77 4.19 4.32 8.35 7.29 4.9
SD 1.03 1.3 0.96 0.77 1.58 2.78 2.89 1.16
Group C
M 4.14 1.91 3.88 4.48 3.5 7.53 7.47 5
SD 0.99 1.19 1.36 0.66 1.44 3.35 3.07 1.09
Group D
M 4.18 2.06 3.94 4.37 4.18 7.42 7.58 5.37
SD 0.91 1.37 1.46 0.85 1.33 3.06 3.79 0.99
126
Table 6.19 Construct correlations
1 2 3 4 5 6 7 8 9 10
1 Combined Ontological Completeness 1
2 Ontological Overlap .023 1
3 Modelling Familiarity .156 -.137 1
4 Diagram Familiarity .140 -.083 .326** 1
5 Domain Familiarity (Library) .074 -.050 .131 -.033 1
6 Performance on Comprehension Test .021 .218* -.097 .044 -.116 1
7 Performance on Problem Solving Test .031 -.047 -.029 .029 .080 .229** 1
8 Performance on Cloze Test -.033 .048 .009 .137 -.002 .298** .352** 1
9 Perceived Usefulness .097 .072 -.017 .071 -.037 .164 .182* .186* 1
10 Cognitive Load .072 -.075 -.312** -.173* .014 -.099 -.100 -.242** .138 1
*. Correlation is significant at the 0.05 level (2-tailed).
**. Correlation is significant at the 0.01 level (2-tailed).
Chapter 6:An Experimental Study to Test and Extend the Theory of Faithful Use of Conceptual Model Combinations 127
Comprehension
To test hypotheses 3a and 4a, we ran ANCOVA analysis with combined ontological
completeness and ontological overlap as factors, modelling familiarity, diagram familiarity,
and cognitive load as covariates, and comprehension performance as the dependent measure.
Table 6.20 shows the results.
Table 6.20 Results for model comprehension
Difference among groups: Means (Std. Deviation)
Combined ontological Completeness
Low High
Ontological Overlap
Low 3.50
(1.44) 3.73
(1.42)
High 4.32
(1.58)
4.18
(1.33)
ANCOVA
Source df F Sig. Partial
2
Corrected Model 7 1.602 .141 .084
Intercept 1 30.814 .000 .200
Modelling Familiarity 1 1.604 .208 .013
Domain Familiarity 1 1.045 .309 .008
UML Diagrams Familiarity 1 .572 .451 .005
Cognitive Load 1 1.435 .233 .012
Combined Ontological Completeness 1 .198 .657 .002
Ontological Overlap 1 4.837 .030 .038
Combined Ontological Completeness * Ontological
Overlap 1 0.492 .484 .004
R Squared = .084 (Adjusted R Squared = .031)
As Table 6.20 shows, comprehension performance was driven by ontological overlap
rather than combined ontological completeness. Partial eta square value of 0.038 for
ontological overlap indicates that there was a small effect size based on Cohen (1988)
guidelines.
In line with Hypothesis 3a, the results indicate that when there was low ontological
overlap an increase in the level of combined ontological completeness improved model users’
performance in the comprehension test; however, the effect was not statistically significant.
While the positive impact of the combined ontological completeness decreased when
there was high ontological overlap in the model combinations, contrary to the Hypothesis 4a,
users of the combinations of models with high overlap still performed better compared to the
128
users of the model combinations with low overlap. The difference between groups was not
statistically significant.
Problem-Solving
To test Hypotheses 3b and 4b, we ran an ANCOVA test with combined ontological
completeness and ontological overlap as factors, modelling familiarity, diagram familiarity,
and cognitive load as covariates, and problem-solving performance as the dependent measure.
Table 6.21 shows the results.
Table 6.21 Results for problem-solving
Difference among groups: Means (Std. Deviation)
Combined ontological Completeness
Low High
Ontological Overlap
Low 7.53
(3.35)
8.79
(2.33)
High 8.35
(2.78)
7.42
(3.06)
ANCOVA
Source df F Sig. Partial
2
Corrected Model 7 1.198 .309 .064
Intercept 1 24.459 .000 .166
Modelling Familiarity 1 1.250 .266 .010
Domain Familiarity 1 .988 .322 .008
UML Diagrams Familiarity 1 .059 .809 .000
Cognitive Load 1 2.124 .148 .017
Combined Ontological Completeness 1 .262 .610 .002
Ontological Overlap 1 .542 .463 .004
Combined Ontological Completeness * Ontological
Overlap 1 4.468 .037 .035
R Squared = .064 (Adjusted R Squared = .011)
The results indicate that combined ontological completeness and ontological overlap did
not have statistically significant effects on a user’s performance in the problem-solving test.
However, the results of ANCOVA show a statistically significant interaction effect between
combined ontological completeness and ontological overlap. Partial eta square values of
0.001, 0.002, and 0.035 respectively for ontological overlap, combined ontological
completeness, and the interaction effect indicate the small effect size based on Cohen’s (1988)
guidelines.
Chapter 6:An Experimental Study to Test and Extend the Theory of Faithful Use of Conceptual Model Combinations 129
Therefore, we conducted an additional simple effect analysis to identify the origins of
the interaction effect (Elliott & Woodward, 2007). Table 6.22 summarises the results of the
simple effect analysis.
Table 6.22 Simple effect analysis for problem-solving performance
Combined Ontological
Completeness
Ontological Overlap
Mean Difference
Std.
Error Sig.
Low Low (I) High(J)
-.697 .732 .343
High Low (I) High(J) 1.464 .728 .047
Ontological Overlap Combined Ontological Completeness
Mean Difference Std. Error Sig.
Low Low (I) High(J)
-1.350 .728 .066
High Low (I) High(J) .811 .739 .275
The results presented in Table 6.22 show that when there was a high combined
ontological completeness, users of the model combination with a low ontological overlap
performed better than the users of the model combination with high ontological overlap (mean
difference = 1.464, p = 0.047). The results are in line with Hypothesis 4b.
In line with Hypothesis 3b, the findings indicate that when there was a low ontological
overlap, an increase in combined ontological completeness improved model users’
performance by almost 10 percent. However, there was no statistically significant difference
between groups (mean difference = -1.350, p = 0.066).
Similarly, to analyse the differences of the means of performance on the Cloze test
across groups, we ran ANCOVA analysis with combined ontological completeness and
ontological overlap as factors, modelling familiarity, diagram familiarity, and cognitive load
as covariates, and performance in the Cloze test as the dependent measure. Table 6.23
presents the means and standard deviations across groups and the ANCOVA test.
130
Table 6.23 Results for Cloze test
Difference among groups: Means (Std. Deviation)
Combined Ontological Completeness
Low High
Ontological Overlap
Low 7.47
(3.07) 6.76
(3.61)
High 7.29
(2.89)
7.58
(3.79)
ANCOVA
Source df F Sig. Partial
2
Corrected Model 7 1.676 .121 .087
Intercept 1 24.233 .000 .165
Modelling Familiarity 1 1.237 .268 .010
Domain Familiarity 1 .073 .788 .001
UML Diagrams Familiarity 1 2.246 .136 .018
Cognitive Load 1 7.329 .008 .056
Combined Ontological Completeness 1 .039 .844 .000
Ontological Overlap 1 .090 .764 .001
Combined Ontological Completeness * Ontological
Overlap 1 .964 .328 .008
R Squared = .087 (Adjusted R Squared = .035)
As the results indicate, combined ontological completeness and ontological overlap did
not have statistically significant effects on model users’ performance in the Cloze test.
However, cognitive load appeared to be a significant determinant of the performance on the
Cloze test.
In contrast with Hypothesis 3c, the results show that when there was low ontological
overlap, an increase in the level of combined ontological completeness decreased the retention
ability of model users. However, the effect was still not statistically significant.
The findings of the ANCOVA test reveal that when there was high ontological overlap,
an increase in the combined ontological completeness improved model users’ performance in
the Cloze test. The users of the model combination with high ontological overlap and high
combined completeness performed better on the Cloze test compared to the users of a model
combination with high combined ontological completeness but low ontological overlap.
These results are in contrast with what Hypothesis 4c predicted, with no statistically
significant effect of ontological overlap.
Figure 6.4 shows the estimated marginal means of model users’ performances on
comprehension, problem-solving, and the Cloze test and illustrates the direction of the effects
Chapter 6:An Experimental Study to Test and Extend the Theory of Faithful Use of Conceptual Model Combinations 131
of the combined ontological completeness and ontological overlap on domain understanding.
To be able to compare the marginal means, we normalised the graphs by dividing the mean of
the performance on each task by the maximum number of the correct answers provided for
each task to achieve a percentage score ranging from 0 to 1.
The results show that the effect of combined ontological completeness and ontological
overlap varied for each task. For both comprehension and problem-solving tasks, when there
was a low level of ontological overlap, an increase in combined ontological completeness
improved participants’ performance. However, this effect was not statistically significant on
participants’ performance in both tests. The direction of this effect on participants’
performance in the Cloze test was in contrast to both comprehension and the problem-solving
tests; however, the effect was not statistically significant.
As Figure 6.4 shows, an increase in ontological overlap had a major impact on
participants’ performance in the comprehension test. In contrast, an increase in ontological
overlap decreased participants’ performance in the problem-solving test. In particular, there
was a statistically significant difference between the performance of users of a model
combination with high ontological completeness and high ontological overlap and the users of
a model combination with low ontological overlap and high ontological completeness.
However, for the Cloze test, an increase in the level of ontological overlap improved users’
performance when the level of combined ontological completeness of models was high.
132
Covariates appearing in the model were evaluated at the following values: Domain Familiarity = 3.9237, Diagram Familiarity = 2.05, Modelling
Familiarity = 4.1870, Cognitive Load = 4.1170
Figure 6.4 Estimated Marginal Means of Performance on Domain Understanding
0.4
0.45
0.5
0.55
0.6
Low High
Combined Ontological Completeness
Low
High
Pe
rfo
rm
an
ce
on
Co
mp
re
he
ns
ion
Te
st Ontological
Overlap
0.4
0.45
0.5
0.55
0.6
Low High
Combined Ontological Completeness
Low
High
Ontological Overlap
Pe
rfo
rm
an
ce
on
Pro
ble
m S
olv
ing
T
es
t
0.4
0.45
0.5
0.55
0.6
Low High
Combined Ontological Completeness
Low
High
Ontological Overlap
Pe
rfo
rm
an
ce
on
C
lo
ze
Tes
t
Chapter 6:An Experimental Study to Test and Extend the Theory of Faithful Use of Conceptual Model Combinations 133
To analyse Hypotheses 5 and 6, we ran ANCOVA analysis with combined ontological
completeness and ontological overlap as factors, modelling familiarity, diagram familiarity,
and cognitive load as covariates, and perceived usefulness as the dependent measure. Table
6.24 illustrates the results.
Table 6.24 Results for perceived usefulness
Difference among groups: Means (Std. Deviation)
Combined ontological Completeness
Low High
Ontological Overlap
Low 5.00
(1.09) 4.97
(1.23)
High 4.90
(1.16)
5.37
(0.99)
ANCOVA
Source df F Sig. Partial
2
Corrected Model 7 1.042 .406 .056
Intercept 1 28.607 .000 .189
Modelling Familiarity 1 .006 .936 .000
Domain Familiarity 1 .166 .684 .001
UML Diagrams Familiarity 1 1.072 .302 .009
Cognitive Load 1 2.774 .098 .022
Combined Ontological Completeness 1 .681 .411 .006
Ontological Overlap 1 .977 .325 .008
Combined Ontological Completeness * Ontological
Overlap 1 1.667 .199 .013
R Squared = .056 (Adjusted R Squared = .002)
The results did not show any statistically significant effect of combined ontological
completeness and ontological overlap on perceived usefulness. Users of the model
combination with high combined ontological completeness and low ontological overlap
perceived the model combination as less useful than the users of the model combination with
low combined ontological completeness and low ontological overlap. While the effect of
combined ontological completeness was not statistically significant, the results do not support
Hypothesis 5. Figure 6.5 demonstrates that between the users of the model combinations that
had a high combined ontological completeness, users of a combination of models with high
ontological overlap perceived the combination as more useful than the users of the
combination with low ontological overlap. While the effect of ontological overlap was not
statistically significant, this result is in contrast with the expectation of Hypothesis 6.
134
Covariates appearing in the model were evaluated at the following values: Domain Familiarity = 3.9237,
Diagram Familiarity = 2.05, Modelling Familiarity = 4.1870, Cognitive Load = 4.1170
Figure 6.5 Estimated Marginal Means of Perceived Usefulness
6.7 DISCUSSION
6.7.1 Summary of Findings
Table 6.25 summarises the insights gained from the data analysis.
Table 6.25 Summary of hypothesis testing
Hypotheses Results Interpretation
Mo
del
Sel
ecti
on
H1 Given any model participants
start with, they will select an
additional model/s that add/s
maximum ontological
completeness to the model
combination.
Combined ontological completeness had
a statistically significant positive effect
on model selection.
The hypothesis is
supported.
H2 Given any model participants
start with, they will select an
additional model/s that add/s
minimum ontological overlap
to the model combination.
Ontological overlap had a statistically
significant negative effect on the
selection of a second model. Seventy-six
participants who selected three or more
models did not use at least one of the
models in combination.
The direction of the
effect is as
expected; however,
the hypothesis is
not supported.
Dom
ain U
nder
stan
din
g
H3a Users of the combination of models with a high combined
ontological completeness will
achieve a higher score in the
comprehension test compared
to the users of the combination
of models with a low combined
ontological completeness.
The group that had a model combination with high combined ontological
completeness and low ontological overlap
performed better on the comprehension
test compared to the group that had a
model combination with low combined
ontological completeness and low
ontological overlap; however, the
difference did not reach statistical
significance.
The direction of the effect is as
expected; however,
the hypothesis is
not supported.
H4a Users of the combination of
models with a high ontological
overlap will achieve a lower score in the comprehension test
compared to the users of the
When there is high combined ontological
completeness, users of the model
combination with high overlap performed better on the comprehension test
compared to users of the model
The direction of the
effect is not as
expected and the hypothesis is not
supported.
0.6
0.65
0.7
0.75
0.8
Low High
Combined Ontological Completeness
Low
High
Perc
eive
d U
sefu
lnes
sOntological
Overlap
Chapter 6:An Experimental Study to Test and Extend the Theory of Faithful Use of Conceptual Model Combinations 135
combination of models with a
low ontological overlap.
combination with low ontological
overlap; however, the difference did not
reach statistical significance.
H3b Users of the combination of
models with a high combined
ontological completeness will
achieve a higher score in the
problem-solving test compared
to the users of the combination
of models with a low combined
ontological completeness.
Users of a model combination with high
combined ontological completeness and
low ontological overlap performed better
in the problem solving test compared to
users of a model combination with low
combined ontological completeness and
low ontological overlap; however, the
difference did not reach statistical
significance.
The direction of
effect is as
expected; however,
the hypothesis is
not supported.
H4b Users of the combination of
models with a high ontological overlap will achieve a lower
score in the problem-solving
test compared to the users of
the combination of models with
a low ontological overlap.
Model users of a model combination with
high combined ontological completeness and high ontological overlap achieved a
lower score in the problem solving test
compared to users of a model
combination with high combined
ontological completeness and low
ontological overlap, and the difference
reached statistical significance.
The hypothesis is
supported.
H3c Users of the combination of
models with a high combined
ontological completeness will
achieve a higher score in the
Cloze test compared to the users of the combination of
models with a low combined
ontological completeness.
Users of a model combination with high
combined ontological completeness and
low ontological overlap achieved a lower
score on the Cloze test compared to users
of a combination with low combined ontological completeness and low
ontological overlap; however, the
difference did not reach statistical
significance.
The direction of the
effect is not as
expected and the
hypothesis is not
supported.
H4c Users of the combination of
models with a high ontological
overlap will achieve a lower
score in the Cloze test
compared to the users of the
combination of models with a
low ontological overlap.
Model users of a model combination with
high combined ontological completeness
and high ontological overlap performed
better on the Cloze test compared to users
of a model combination with high
combined ontological completeness and
low ontological overlap.
The direction of the
effect is not as
expected and the
hypothesis is not
supported.
Per
ceiv
ed U
sefu
lnes
s
H5 Users of the combination of
models with a high combined
ontological completeness will perceive the combination as
more useful than the users of
the combination of models with
a low combined ontological
completeness.
Users of a model combination with high
combined ontological completeness and
low ontological overlap perceived the combination less useful compared to
users of a model combination with low
combined ontological completeness and
low ontological overlap; however, the
difference did not reach statistical
significance.
The direction of the
effect is not as
expected and the hypothesis is not
supported.
H6 Users of the combination of
models with a high ontological
overlap will perceive the
combination as less useful than
the users of the combination of
the models with a low ontological overlap.
Users of a model combination with high
combined ontological completeness and
low ontological overlap perceived the
combination less useful compared to
users of a model combination with high
combined ontological completeness and high ontological overlap; however, the
difference did not reach statistical
significance
The direction of the
effect is not as
expected and the
hypothesis is not
supported.
136
With respect to the model selection, our results were reasonably clear. Our findings are
in line with Hypothesis 1, indicating that model users added an additional model if the model
maximised the combined ontological completeness of the model combination. Specifically,
we found that the level of combined ontological completeness an additional model added to
the combination had a statistically significant positive impact on the selection of the model.
This finding is significant in that it confirms that model users desire a complete representation
of a domain (Recker, 2014), which also underlines the importance of design consideration
with regards to types of models. The results were also in line with Hypothesis 2, when model
users selected the second model to add to the model combination. The findings indicate that
the level of ontological overlap added to the model combination had a statistically significant
negative impact on the selection of the model. However, the hypothesis was not supported
when model users selected a third model to add to the model combination. Because this result
was a surprise, we set out to determine a possible explanation. A deep investigation of the
collected data indicated that more than 76% of the participants who picked more than two
models did not use all of the models in combination. Indeed, they would have liked to remove
some of them from the combination. By removing some models, they decreased the level of
ontological overlap; however, they also decreased the level of combined ontological
completeness. These results indicate that model users preferred less complete with less
overlap combinations than a more complete but also with more overlap combination. In line
with the prediction of the theory of faithful use of conceptual model combinations, the
selection of which model to use in combination was governed by the principles of clarity over
completeness (Recker, 2014). Some studies have also indicated that model users may trade-
off completeness to achieve an easier model to use (Bowen et al., 2009; Siau & Lee, 2004; zur
Muehlen et al., 2007).
With respect to domain understanding, our results were equivocal. To evaluate the
understanding gained from model combinations, we used three measures: a model
comprehension test, a problem-solving test to measure the deep-level domain understanding,
and a Cloze test to measure surface-level domain understanding by evaluating the ability of
participants to reproduce (or recall) absorbed information (Bera et al., 2014; Bodart et al.,
2001; Recker & Dreiling, 2011; Recker et al., 2014; Shanks et al., 2010).
Chapter 6:An Experimental Study to Test and Extend the Theory of Faithful Use of Conceptual Model Combinations 137
In line with Hypotheses 3a and 3b, we found that when there was low ontological
overlap, an increase in combined ontological completeness improved both comprehension
performance and problem-solving performance; however, the effect was not statistically
significant. In contrast to Hypothesis 4a, our data analysis shows that ontological overlap had
a statistically significant and positive impact on participants’ comprehension performance.
However, in line with Hypothesis 4b, we found a statistically significant negative impact of
ontological overlap on problem-solving performance when participants used combinations of
models with high combined ontological completeness.
These findings are interesting because ontological overlap had a major positive impact
on model users’ performance on the model comprehension test. On the other hand,
ontological overlap had a statistically significant but negative effect on problem-solving
performance. One possible reason for these findings is the difference between the nature of
the model comprehension test and the problem-solving test. The model comprehension test
measures a surface understanding of the model combinations, while the problem-solving test
measures a deep understanding of the domain. Surface-level understanding requires users to
answer the questions based on the information plucked from their working memory (Mayer,
1989), and involves surface processing of the information without interpretation, reasoning, or
adding new elements (Offir, Lev, & Bezalel, 2008). Deep understanding involves deep
processing of information (Mayer, 1989, 2005), connecting the facts and concepts and
integrating the ideas with prior knowledge and collected information plucked from long-term
memory.
A model combination with high ontological overlap contains repeated constructs
representing the same element of a real-world domain. Studies on learning preferences
indicate that repetition could have a positive effect on learning (Bromage & Mayer, 1986;
Mayer, 1983). Repeating the information requires only surface processing without quotes or
interpretation. Therefore, the repetition of constructs in the model combination may help
participants to better link the elements of the domain represented in the models and create
higher surface level understanding.
While learning preferences hypotheses argue that repetition can improve learning
performance, the cognitive theory of multimedia learning, on the other hand, argues that
human working memory has a limited capacity. Repeated information would require
138
additional capacity out of the limited capacity of human working memory (Mayer, 2009, p.
126) and create extraneous cognitive load (Sweller, 1994; Sweller & Chandler, 1994). Given
that a deep understanding requires a deep cognitive process of handling of information
(Mayer, 1989, 2005), a more complete representation of a domain provides users with more
information that entails more cognitive demand to absorb information for problem-solving.
Therefore, additional extraneous cognitive load, introduced by the ontological overlap, will
greatly shrink the capacity of the human working memory to acquire information and generate
deep-level understanding.
In terms of Hypotheses 3c and 4c, our results show no evidence of significant
differences between treatment groups for the recall test (Cloze test). A lack of significant
differences between all treatment groups for the Cloze test may result from the fact that
participants responded to the Cloze test without having access to the models, in contrast to the
comprehension and the problem-solving test. Our findings also indicate that there was no
significant difference across treatment groups for the perceived usefulness of the model
combinations. However, the directions of the potential effects of combined ontological
completeness and ontological overlap on the recall test were similar to that of perceived
usefulness. While any explanation of these unexpected results is speculative, it is worth
noting that the perceived usefulness questions were immediately after the Cloze test. Thus,
there may have been a carry-over effect from the previous task.
6.7.2 Theoretical Implications
Our research provides the first dedicated empirical evaluation of the theory of faithful
use of conceptual model combinations. Similar to the argument of the theory, we identified
that combined ontological completeness and ontological overlap can explain the faithful use
of conceptual model combinations. In particular, our findings indicate that both combined
ontological completeness and ontological overlap are major determinants of users’ selection
of a model combination. Similar to the prediction of the theory, we also found that model
users compromised combined ontological completeness over ontological clarity. Our findings
indicate that users would have liked to remove some models from the combination they had
selected. A possible explanation is that the model combination reached the maximum bearable
level of ontological overlap. However, in this study, we could not offer a measurement to
evaluate the bearable level of overlap.
Chapter 6:An Experimental Study to Test and Extend the Theory of Faithful Use of Conceptual Model Combinations 139
Our findings also extend the theory in a number of ways. First, we found that the effect
of combined ontological completeness and ontological overlap varies based on different task
settings. Accordingly, different attributes of different tasks should be considered when
evaluating the importance of combined ontological completeness and ontological overlap.
Studies on task setting indicate that based on the type of task, the information seeking
behaviour and use of information may vary (Byström & Järvelin, 1995; Campbell, 1988). Our
results corroborate this argument: overlap helps when model users are involved with tasks
that require surface-level understanding, whereas overlap has a negative effect when users are
involved with tasks that require deep-level understanding.
Second, we have determined that there is an interaction effect between ontological
overlap and combined ontological completeness. Recker’s (2014) theory suggested that the
level of understanding would be moderated by the level of ontological overlap. Our findings
indicate that the negative effect of ontological overlap overcomes the positive impacts of
ontological completeness. However, our findings indicate that when there is low ontological
overlap, an increase in combined ontological completeness improves model users’ problem-
solving performance.
These results can be interpreted by drawing on principles from the cognitive theory of
multimedia learning: this theory suggests that learning has three outcomes, no learning, rote
learning, and meaningful learning (Mayer, 2009, p. 21). No learning occurs when participants
have poor performance on both a retention test and a transfer test. Rote learning occurs when
participants only have a good performance on a retention test, and meaningful learning occurs
when participants have good performance on both a retention test and a transfer test. Our
results indicate that ontological overlap improves rote learning. However, when there was low
ontological overlap, an increase in combined ontological completeness improved participants’
performance on both the comprehension test (retention test) and the problem-solving test
(transfer test). These results indicate that a model combination with low ontological overlap
but high combined ontological completeness leads to meaningful learning.
6.7.3 Practical Implications
Our findings inform a largely neglected aspect of conceptual modelling practice – the
use of multiple types of conceptual models in combination. Two main practical implications
140
flow from this work. First, the findings of this study provide practical guidance that can assist
conceptual modelling practitioners to select and use appropriate combinations of conceptual
models for different systems analysis and design tasks. This study provides evidence that the
primacy of ontological completeness and clarity varies for different tasks. Therefore, model
end users should be aware of the essence and objectives of tasks and whether tasks require a
parsimonious representation or a complete representation of the real world.
Second, while our study focused on model interpretation rather than model design, we
believe that our findings also provide practical implications for model designing practices.
This is because conceptual modelling process – either the design or use of models – are
ultimately dependent on how well individuals can understand the domain presented by the
models (e.g., Burton-Jones & Meso, 2008; Recker et al., 2014). Our findings show that
different levels of completeness and overlap are useful for different levels of understanding.
Therefore, model designers should design partially redundant representations or complete
representations depending on the purpose of the design of the conceptual models. These
findings inform model designers with regards to which types of models should be designed to
effectively support an upcoming task.
6.7.4 Limitations
Several limitations pertain to our study. First, the experimental procedures tasks, and
models used in our experiment may not faithfully reflect the scale of real-world modelling in
practice. This is common for experimental studies (e.g., Bera et al., 2014; Figl, Mendling, et
al., 2013). Experiments are artificial situations mainly focused on controlling variables and
maximising internal validity (Calder et al., 1981).
Second, we used a sample of students rather than a sample of practitioners. Despite
concerns regarding the limitations of using students as a sample (Compeau, Marcolin, Kelley,
& Higgins, 2012), we believe that our sample was still suitable for our study for several
reasons. First, our experiment was designed to test the faithful use of conceptual model
combinations. The use of conceptual models was professionally and personally relevant for
our students because they had enrolled in conceptual modelling and systems analysis and
design courses. Second, in the context of our study, students are regarded as reasonable
proxies of junior working experts. For instance, Arisholm and Sjøberg (2004) found little
Chapter 6:An Experimental Study to Test and Extend the Theory of Faithful Use of Conceptual Model Combinations 141
difference between students and junior professionals in understanding UML diagrams and
maintaining Java programs.
Third, we used combinations of UML diagrams in our experiment. However, our
findings from the interviews (Jabbari Sabegh & Recker, 2017) and some previous studies
indicate that IS professionals may integrate UML diagrams with other types of models, such
as ERD or BPMN (e.g., Petre, 2013). This limits the generalisability of our findings; however,
our primary objective in the experimental study was to control our variables and increase our
internal validity. We only used UML diagrams to avoid the potential extraneous effect of
individuals’ different levels of familiarity with UML diagrams compared to other types of
models.
6.8 CONCLUSION
Through this research, we have contributed to conceptual modelling research by
providing an empirical evaluation of the theory of faithful use of conceptual model
combinations. Our findings show the important role of completeness and clarity of model
combinations on model selection and domain understanding. These findings provide an
understanding of the use of conceptual models in practice, and in turn, can guide model end
users in their decisions when choosing to use conceptual models and inform executive
decision makers with regards to their investment to implement conceptual modelling
initiatives. In summation, our study informs the use of model combinations in practice and
may ultimately lead to more effective and efficient use of conceptual models in information
systems analysis and design tasks.
142
Part C: Retrospective
Chapter 7: Discussion
7.1 SYNTHESIS OF FINDINGS FROM STUDIES
This section summarises the findings of the three studies and discusses the implications
of the findings. The importance of conceptual modelling in information systems analysis and
design, a lack of knowledge about the use of multiple models in practice, and the constantly
changing and developing world motivated the first study of this dissertation. Study 1
answered the first research question: “What is the current state of knowledge of conceptual
modelling in IS?” To address this research question, Study 1 reviewed the current research
efforts in conceptual modelling and provided a revised research framework that suggests new
opportunities for research and stresses the need for ongoing innovation in conceptual
modelling in response to the emerging practices in IS.
Study 1 reviewed studies published in top IS journals and examined the extent to which
the current state of knowledge in conceptual modelling has addressed the challenges and
opportunities underlined by a formative and influential framework proposed by Wand and
Weber (2002). The results disclosed that some of the opportunities noted in 2002 have been
substantially addressed; however, some aspects of conceptual modelling remain entirely
overlooked. The findings of the first study also revealed several innovative studies that did
address some aspects that were not highlighted by the initial framework proposed by Wand
and Weber (2002). Study 1 also provided a new research framework that draws the attention
of conceptual modelling research to the interplay between digital representations and
outcomes and stimulates a new, relevant wave of conceptual modelling research. The findings
from a review of the literature are detailed in Chapter 4. One of the most notable unanswered
opportunities concerns the lack of empirical investigation on the use of multiple models. To
date, little research has been conducted to evaluate multiple models and investigate how a
model combination can support system analysis and design tasks.
Chapter 7:Discussion 143
Motivated by the practical observations and the lack of knowledge in IS literature about
the use of multiple models, the second study of this dissertation set out to answer the second
research question: “Why do systems analysts and designers use multiple conceptual models?”
To address this research question, Study 2 explored the reported use of multiple conceptual
models for system analysis and design tasks. Study 2 provided empirical evidence to address
the concerns highlighted in Study 1 about the lack of research in regards to the use of model
combinations. It provided qualitative support for the argument of previous studies (Dobing &
Parsons, 2008; Fettke, 2009; Whittle et al., 2014) that practitioners do indeed use multiple
models in system analysis and design tasks. The findings uncovered three main reasons why
practitioners use multiple types of models. These reasons are: 1) the size and complexity of
the domain, 2) involvement of different stakeholders in the projects, and 3) incompleteness of
any particular model.
As with the predictions of previous ontological analyses (e.g., Green et al., 2007;
Recker et al., 2009; Siau, 2010), incompleteness of single models led IS professionals to use
multiple models. In line with the findings of J. Kim et al. (2000), the finding of Study 2
indicated that practitioners use multiple models to understand different aspects of the domain,
especially when the domain is large and complex. Another reason why professionals use
multiple models is to meet the requirements of different stakeholders involved in the project,
such as project owners, end-users, or system analysts and designers. This finding is also in
line with previous studies (e.g., Darke & Shanks, 1997; Topi & Ramesh, 2002).
The findings of Study 2 also proposed theoretical relationships between these factors.
The results indicated that both the size and complexity of the domain and incompleteness of
single models are the primary determinants of how IS professionals use different types of
models combination. However, the extent and type of models they select will be moderated
by the requirements of the involved stakeholders. These relationships are consistent with
Wand and Weber’s (1990a, 1993) representation theory, while they also extend the main
premises of the theory. The findings uncovered that in addition to the ontological capabilities
of a model, the supported perspective of the model is also a key determinant of the use of
models. In line with Wand and Weber’s (2002) framework on conceptual modelling, scripts
(incompleteness of a single model) and context (size and complexity of domain and
144
involvement of different stakeholders) explain why IS professionals use different types of
conceptual models in combination.
Study 2 also explored the semantic and pragmatic reasons that influence the choice and
selection of different models for system analysis and design tasks. The semantic reasons
indicated that practitioners use multiple models in combination where: 1) they increase the
completeness of representation, 2) have low overlap, and 3) there is correspondence between
them.
The semantic reasons are in line with the propositions of the theory of faithful use of
conceptual model combinations (Recker, 2014), and other extensions of ontological theories
(Green, 1996; Weber, 1997). The findings indicate that IS professionals select a combination
of models that increase the completeness of the representation while also attempting to avoid
overlapping models, because overlap adds unnecessary complexity. However, Study 2
discovered a new concept: that of correspondence. Correspondence implies that while
practitioners avoid overlapping models, the existence of semantically related elements in pairs
of models influences the selection of these models.
The pragmatic reasons indicate that IS professionals use different types of models in
combination that: 1) are required by organisational regulations and standards, 2) are required
for different types of stakeholders involved in the project, 3) are widely used in industry, 4)
support the task goals, and 5) are supported by the available tools. In and of themselves, these
findings are also consistent with previous studies (e.g., Delfmann et al., 2015; Recker et al.,
2010; Rosemann, 2006; Samuel et al., 2015). However, the dominant finding of Study 2 is the
relationships between pragmatic factors, semantic factors, and the use of model combinations.
The findings indicate that pragmatic reasons might overrule semantic reasons. For instance,
similar to the findings of Green et al. (2011) on the use of multiple grammars, the findings of
Study 2 also indicate that organisational conventions may direct model users to choose
models that overlap. These findings are detailed in Chapter 5.
Motivated by the principles of the theory of faithful use of conceptual model
combinations, Study 3 set out to test the propositions of the theory (Recker, 2014) and
provide a deep understanding and empirical insights into the combined use of conceptual
models. Study 3 was a theory-testing study aimed at addressing the third research question:
Chapter 7:Discussion 145
“What characteristics of models affect users’ selection, understanding, and perceived
usefulness of model combinations?” To address this research question, Study 3 tested the
propositions of the theory of faithful use of conceptual model combinations through a
laboratory experiment.
The findings of Study 3 partially confirmed the selection proposition of the theory of
faithful use of conceptual model combinations. In line with the semantic reasons revealed in
Study 2, and conforming to the predictions of the theory (Recker, 2014), the findings of Study
3 indicate that the level of completeness a model added to the model combination had a
statistically significant positive effect on user’s selection of the model. These findings also
indicate that the level of ontological overlap a model added to the model combination had a
statistically significant negative effect on users’ selection of the second model. While the
tolerable level of ontological overlap could not be measured (Recker, 2014), the findings of
Study 3 indicate that participants would have liked to remove some models from the
combination they had selected to decrease the overlap, which also comes at the cost of
decreasing the completeness of the model combination. Results indicate that the use of
multiple models in combination can be governed by the principles of clarity over
completeness. Figl and Recker (2014) found that users’ preference for different representation
formats varies dependent on task purpose. The findings of Study 3 indicate that, for an
upcoming problem-solving task, participants selected models that increased the completeness
of the representation; however, they did not select additional models if the model increased
the ontological overlap of the model combination.
The findings of Study 3 reveal the different role of ontological overlap in different task
settings. These findings provide empirical evidence to support the theoretical relationship
between pragmatic reasons and semantic reasons, as proposed in Study 2. The results of Study
3 confirm that pragmatic reasons may overrule the ontological factors depending on the
context of use. The findings indicate that ontological overlap had a statistically significant
positive impact on users’ performance in models comprehension. However, an increase in
ontological overlap had a statistically significant negative effect on users’ problem-solving
performance. The propositions of the theory of faithful use of conceptual model combinations
argue that an increase in ontological overlap will decrease the understandability of the model
combination. The proposed reason is because ontological overlap between models decreases
146
the clarity of the model combination and increases the cognitive load. However, the findings
of Study 3 were equivocal. In the surface-level understanding of model combinations, the
ontological overlap between models helped users to better comprehend the elements presented
in the models. This finding can be interpreted by drawing on the learning preferences
hypothesis and surface-level understanding. Learning preferences hypotheses argue that
repetition can improve learning performance (Bromage & Mayer, 1986; Mayer, 1983).
Surface processing of information involves repeating the information without interpretation
and adding new elements (Offir et al., 2008). It appears that the existence of ontological
overlap between models would assist users to link the elements presented in the models.
On the other hand, in line with the predictions of the theory of faithful use of conceptual
model combinations, the findings of Study 3 reveal that ontological overlap decreased users’
performance in the problem-solving test. It appears that the problem-solving test requires a
deep-level understanding. Given that possibility, repetition would increase extraneous
cognitive load (Sweller, 1994; Sweller & Chandler, 1994) and human working memory has a
limited capacity (Mayer, 1989, 2005), high ontological overlap in a model combination will
greatly shrink the capacity of the human working memory to acquire information and
decrease its ability to generate deep-level understanding. These findings are detailed in
Chapter 6.
In summary, Studies 2 and 3 addressed the research opportunities highlighted in Study 1
and investigated the use of model combinations and examined the factors that define the best
combination of models. The findings of these studies corroborate the predictions of
ontological theories (Recker, 2014) and confirm the importance of pragmatic factors, as has
been noted in previous studies (e.g., Burton-Jones et al., 2009; Green et al., 2011; Recker et
al., 2010; Rosemann, 2006; Samuel et al., 2015). However, Study 2 also identified the
important interaction effects between context and script factors and proposed that sometimes
context influences the impact of script qualities and may override them. Experimental results
support these findings and indicate that the role of ontological overlap varies depending on
task settings. The findings of Study 2 and 3 are consistent with the framework proposed by
Wand and Weber (2002), and indicate that both the pragmatic context and semantic factors
explain the use of model combinations. In addition, the findings of Studies 2 and 3 also
correspond to the new framework proposed in Study 1. The findings provide empirical
Chapter 7:Discussion 147
support for the view that antecedents (i.e., context) can influence other antecedents (i.e.,
ontological qualities) and identified the role of these factors in the use of representation (i.e.,
combinations of models representing a domain) and the outcome (i.e., domain understanding).
7.2 IMPLICATIONS
7.2.1 Implications for Research
This thesis provides several implications for research. These implications are outlined in
each study and briefly summarised again here. First, Study 1 provided a review of research on
conceptual modelling in IS. Study 1 applied the framework proposed by Wand and Weber
(2002) to investigate the extent to which the research opportunities proposed by the
framework have been considered by the extant research on conceptual modelling. The
findings of the review highlighted the areas that require further research and indicated that
some of the research areas are entirely neglected in IS research. For example, further in-depth
investigations are required to understand the role of social agenda factors on conceptual
modelling. The study underlined the contributions of past research and can serve as a guide
for future research.
Second, Study 1 proposed new research opportunities and identified some path-breaking
studies, such as multidimensional modelling and systems flexibility (e.g., de Albuquerque &
Christ, 2015); or relevant aspects of model-driven architecture, such as security features (e.g.,
D'aubeterre et al., 2008). Study 1 also provided a new framework that draws the attention of
conceptual modelling research to unknown aspects of conceptual modelling (Recker, 2015);
emerging needs, such effective use of big data (Rai, 2017); and a new wave of modelling
paradigms, such as agile modelling or approaches when no script is involved (Lukyanenko &
Parsons, 2013). The proposed framework creates significant new opportunities for the
conceptual modelling community and substantially expands what counts as conceptual
modelling research.
Third, the second study explored the use of multiple conceptual models in practice.
While previous studies have indicated that IS professionals normally use multiple types of
models (Dobing & Parsons, 2008; Fettke, 2009), it was unknown how and why they are using
different types of models. Study 2 uncovered important semantic and pragmatic factors that
affect the extent and types of models IS professionals use in most of their systems analysis
148
and design tasks. These results extend the body of knowledge about conceptual modelling,
provide empirical insights, and also develop the theoretical foundations to explain why IS
professionals use multiple models.
Fourth, the study identified contextual factors other than the ontological factors as the
main determinants of using model combinations. Contextual factors not only have primary
effects on the combined use of conceptual models, there are also important interaction effects
between context and ontological factors. The results of the study confirm previous findings
that suggest the important role of contextual factors, such as organisational conventions or
tool supports (Recker et al., 2010), individual difference in understanding conceptual models
(Khatri & Vessey, 2016; Khatri et al., 2006), task objectives (Samuel et al., 2015), or different
stakeholder requirements (Topi & Ramesh, 2002). In addition, the results of Study 2 propose
that sematic and pragmatic factors interact in the use of model combinations. In particular, the
results of Study 2 indicate that contextual factors may dominate ontological factors. These
findings highlight the importance of considering both semantic factors and pragmatic factors
in further research about conceptual modelling (Bera et al., 2014; Burton-Jones et al., 2009).
Fifth, Study 3 tested the theory of faithful use of conceptual model combinations. This
study identified the role of completeness and overlap on users’ choice in selecting types of
models. These findings indicate that ontological completeness and overlap play a significant
role in model selection. However, users may choose not to select a model combination with
high ontological overlap, even if the combination provides them with a high complete
representation. These findings corroborate previous studies that have suggested that users
prefer to use representations that are easier to understand and compromised on completeness
of the representations (Samuel et al., 2015; zur Muehlen et al., 2007). Previous studies have
also indicated that the use of representations depends on the purpose of tasks and individual
differences (Figl & Recker, 2014). These findings indicate that the primacy of ontological
completeness or overlap may vary across different task settings and other pragmatic factors.
There are also future opportunities to evaluate how decisions about the selection of models
may connect to the performance of users across different tasks.
Sixth, while model users preferred a combination of models with low ontological
overlap, the results of the experiment revealed that ontological overlap improved the
performance of model users when they were dealing with a task that required a surface level
Chapter 7:Discussion 149
understanding. In contrast, when users were involved with a task that required a deep-level
understanding, ontological overlap decreased the performance of the users. These results
corroborate the equivocal arguments about the usefulness of ontological clarity on users’
performance. While most studies have argued that ontological clarity has a positive effect on
users’ performance (e.g., Milton et al., 2012; Parsons, 2011; Shanks et al., 2010), some
findings have challenged this argument (Allen & March, 2012; Bera et al., 2014; Bowen et
al., 2006), in addition, some studies have also identified that using different forms of
representation with redundant information improves users performance in domain
understanding (Gemino & Parker, 2009). Previous studies have demonstrated the interaction
effect between domain knowledge and ontological clarity (Bera et al., 2014; Burton-Jones &
Weber, 1999), the findings of the current research extend the research on ontological clarity
and identify the different roles of ontological clarity in different task settings. There would be
further opportunities to extend this stream of research, for instance, future studies could
examine a three-way interaction of domain knowledge, task settings, and ontological clarity.
7.2.2 Implications for Practice
Several practical implications flow from this research. First, the results indicate that the
role of ontological factors on the use of multiple models is not always straightforward. While
IS professionals prefer models with a high level of completeness and less overlapping
constructs, the correspondence between models may lead IS professionals to use partially
overlapping models. It would appear that, depending on the context, the correspondence
between models may sometimes bring forth desired qualities for users. This suggests a
practical direction for model users to find appropriate models to be able to integrate
information presented in different models and to also provide guidance for system designers
to avoid inconsistency in different representations of the domain.
Second, this research revealed that IS professionals prefer to use multiple models,
especially when dealing with large and complex domains. When the domain is large and
complex, IS professionals found it difficult to understand the entire domain through a single
model. The findings of the study also suggest that those models should be used in a
combination that meets different stakeholder requirements and fits the purpose of the task.
This provides practical guidance for IS professionals for identifying appropriate models to
use.
150
Third, the findings of this research provide practical guidance for decision makers to
invest in appropriate modelling tools and to define effective organisational conventions. The
findings indicate that organisational rules may lead IS professionals to use overlapping
models, which may lead to unnecessary complexity in the related tasks. Moreover, the results
indicate that IS professionals use the types of models in which they are supported by available
tools.
Chapter 8:Conclusions 151
Chapter 8: Conclusions
8.1 CONTRIBUTIONS
The central contribution of this thesis is that it provides results from an extensive
investigation into the combined use of conceptual models, and attempts to validate and extend
the theory of faithful use of conceptual model combinations. In addition to this central
contribution, this research brings forth a number of further contributions to IS research and
practice.
8.1.1 For research
First, the current research provides an overview of the current state of knowledge about
conceptual modelling. This research may assist researchers to understand which components
of conceptual modelling, for example, modelling grammars, methods, scripts, and context,
and what aspects of these components, as suggested by Wand and Weber (2002), have been
examined by current research efforts. The findings shed light on the aspects of conceptual
modelling that require further research. In addition, while the research framework proposed
by Wand and Weber (2002) certainly provides significant research directions, Study 1
proposed a new research framework to stimulate new pathways to research on conceptual
modelling to address the emerging practices and ongoing innovations in the IS field. This
research hopefully creates significant new opportunities for the conceptual modelling
community.
Second, this research extends the existing literature on conceptual modelling. It is the
first reported research, to the best of my knowledge, that provides a comprehensive and
empirical analysis into the use of multiple models. This research has explored the
circumstances that lead IS professionals to use multiple models and identified the factors that
affect the extent and type of models IS professionals select to use. The research also
developed a procedure and measurements to evaluate the faithful use of model combinations.
The results provide both theoretical logic and empirical evidence to explain how and why IS
professionals use model combinations.
152
Third, the research empirically demonstrates that ontological guidelines provide a
foundation and rationale to explain the combined use of conceptual models. The results
corroborate the principles of existing ontological theories (Green, 1996; Recker, 2014; Weber,
1997), and also reveal that in addition to ontological completeness and ontological overlap,
the existence of semantically related elements (i.e., correspondence) in pairs of models
influences the selection of the model combinations.
Fourth, the research reveals a strong prevalence of context, which has largely been
ignored in much of the theoretical and empirical work (Bera et al., 2014; Burton-Jones et al.,
2009; Recker, 2007). The research discloses the main effects of contextual factors on the
combined use of conceptual models, and it also indicates the important interaction effect
between ontological capabilities and contextual factors. The findings suggest that contextual
factors may override ontological principles. The research thereby sets a direction that can
assist fellow researchers to investigate the use of models in system analysis and design tasks
considering both model-related factors and contextual factors.
Fifth, this research provides the first empirical evaluation of the theory of faithful use of
conceptual model combinations. The results indicate that both ontological completeness and
ontological overlap appear to be major determinants of users’ model selection. However,
users made a trade-off between completeness and overlap. They selected models in
combinations that provided them with a less complete representation, but also less overlap. In
other words, the selection of models was governed by principles of clarity over completeness.
Sixth, the research indicates that different combinations of models are useful for
different types of tasks. The results indicate that a combination of models with high
ontological overlap can be helpful for tasks that require surface understanding, but can
decrease the performance of users who are dealing with a task that requires deep-level
understanding. These results confirmed the value of combining ontological theories with other
theories, such as cognitive theory of multimedia learning, in analysing the use of conceptual
models.
8.1.2 For practice
Given the importance of conceptual modelling in system analysis and design tasks, this
study provides practical contributions by: 1) clarifying the relevant and relatively important
Chapter 8:Conclusions 153
factors regarding the combined use of conceptual models, and 2) providing an evaluated
theory that explains the faithful use of model combinations. These contributions are of
interest, not only for conceptual model users and relevant stakeholders, but also for
conceptual model designers and decision makers in selecting an appropriate approach and
improving overall performance of conceptual modelling processes.
While often different models are developed and used in most of the system analysis and
design tasks (Mussbacher et al., 2014), it is expected that the IS literature provides practical
guidance to assist practitioners to use appropriate types of models to improve the outcome of
the conceptual modelling processes. However, the current review of conceptual modelling
research showed that the IS literature provided little knowledge with regards to how to use
different type of models. The results of this thesis can therefore contribute to practice by
identifying the circumstances that lead IS professionals to use model combinations and which
attributes of models define the best possible combinations.
The results of this research indicate that IS professionals use models that are supported
by available tools and defined by organisational conventions. However, not all types of
models are helpful for all tasks. Different combinations of models with different levels of
overlap may lead to different levels of understanding. These results can assist model users to
select appropriate combinations of models depending on task purposes. The results also shed
light on decision making to (re-)design organisational conventions in a way that not only
provides system designers and analysts with appropriate types of models for a given task, the
results can also assist decision makers to make investment decisions about appropriate tools
to achieve better results from conceptual modelling practices. From a broader view, this
research assists model designers to develop multiple models to suit different task purposes.
With the expectation that any developed model will be used to generate an understanding of
the domain that is presented by models and based on the different role of ontological overlap
in different tasks, model designers may design representations with different levels of overlap
based on the purpose of an upcoming task that involves conceptual model use.
8.2 LIMITATIONS
Although this research makes contributions, like all research, it has limitations. The
choice of research methods brings forth the first set of limitations.
154
First, a literature review is limited to the approach used to identify the relevant studies
and the purpose of the review. Study 1 followed a systematic approach and provided a
descriptive review of the extant studies on conceptual modelling. The systematic literature
review in Study 1 was limited to the papers published in specific journals. The IS literature is
broad, and research on conceptual modelling appears in vast numbers of academic journals.
An alternative approach could be to conduct reviews using different methods (e.g., theoretical
review or meta-analysis) and consider a larger set of journals. However, in the current
research, this issue was mitigated by selecting journals considered by the academic
community to be appropriately representative of mainstream research in the field (i.e., AIS
basket of eight journals). Moreover, different databases were used to locate relevant studies,
and multiple coders analysed the articles to ensure the reliability of descriptive analysis and
evaluation of the papers.
Second, as detailed in Chapter 5, semi-structured interviews are linked to a number of
limitations, such as time limitations and the small sample of participants. An alternative
approach to investigate the use of model combinations, for example, could incorporate case
study methods. However, to increase external validity, semi-structured interviews were
conducted to collect data from professionals from different companies across various
countries. Moreover, to ensure the reliability and validity of the study, the guidelines
proposed by the leading literature were followed, and schemes and protocols were developed
to conduct high-quality interviews.
Third, there were limitations concerning the generalisability and validity of the
laboratory experiments (as summarised in Chapter 6). Experiments are artificial and may not
reflect reality. However, the primary objective of the experiment in this research was to
increase the internal validity and control potentially extraneous effects. To increase the
internal validity of the experiment, measurement items used in previous studies were adapted,
and experimental setting procedures designed in previous related studies were followed (e.g.,
Bera et al., 2014; Burton-Jones & Meso, 2008; Figl & Recker, 2014).
Fourth, limitations result from the scope of the study. This research primarily
investigated the use of UML models. UML was selected because it provides a variety of
diagrams and is popular for object-oriented system development (Batra, 2009). To increase
the generalisability of the findings, practitioners who used other modelling grammars (e.g.,
Chapter 8:Conclusions 155
BPMN) were also interviewed; however, only UML models were used in the experiment to
control potential extraneous effects (e.g., different level of familiarity).
Fifth, the scope of this thesis is the individual level – that is, the individual user of
model combinations. Therefore, the findings are limited in terms of generalising to the group-
level or organisational-level model use. Fellow researchers can therefore consider multi-level
research when investigating the combined use of conceptual models.
8.3 FUTURE RESEARCH
In addition to the remaining research opportunities identified in Study 1 and the new
research opportunities proposed by the research framework proposed in Study 1, several
research opportunities exist to extend the current research and address the limitations of this
work.
First, Study 2 developed theoretical models and proposed theoretical logic to explain the
use of model combinations. Therefore, future research opportunities exist to use empirical
methods to operationalise and validate those theoretical conjectures proposed in Study 2.
Second, multiple conceptual models were not considered for a particular purpose or for
a specific system analysis and design approach. Further opportunities exist for future
researchers to use similar approaches to those proposed in this research to investigate the use
of model combinations for specific purposes (e.g., multidimensional database design) or in
different contexts (e.g., different system development approaches). For example, future
researchers could investigate how the use of model combinations can support agile modelling
(Ambler, 2002).
Third, the results of interviews and previous studies have indicated that IS professionals
normally integrate UML diagrams with other types of models, for example, BPMN or ERD
(Petre, 2013). Future studies could therefore use a setting similar to that proposed in this
research to replicate the study by using different types of models.
Fourth, the use of conceptual models is part of a task, and previous studies have
indicated that different task settings may influence how practitioners use conceptual models
(e.g., Figl & Recker, 2014). Future studies could therefore investigate the use of model
combinations for specific task settings, for example, query development (Bowen et al., 2009).
156
Fifth, previous studies have indicated the role of individual differences in the use of
conceptual models (e.g., Khatri & Vessey, 2016). Fellow researchers could therefore extend
the current experiment by using different domains (e.g., familiar and unfamiliar domains), or
different subjects as participants (e.g., novices versus experts). Future studies could also
extend the current experiment by using combinations of models with different levels of
completeness and overlap, for example, low, moderate, and high levels of completeness and
overlap.
Sixth, future studies could extend this study by investigating cognitive processes in
using model combinations. The results of the experiment show that ontological overlap
improves the models’ comprehension and apparently assists users to link the models.
However, high ontological overlap decreased the performance of users in the problem-solving
test. Future studies could therefore investigate how an increase in the level of ontological
overlap affects users’ information-seeking behaviour and how this behaviour varies based on
different task settings.
Finally, further studies could extend the scope of this thesis and investigate the
combined use of conceptual models at the group or the organisational level, or compare the
combined use of models in different levels, such as group-level versus the individual-level.
8.4 CONCLUSION
This research provides contributions for both research and practice. It has attempted to
provide both an empirical understanding of the combined use of conceptual models in practice
and the development of important extensions and contingencies to existing theory in
conceptual modelling research. This research identified substantive research opportunities that
can hopefully draw attention to further research on conceptual modelling and expand the
current state of knowledge with regards to conceptual modelling.
Bibliography 157
Bibliography
Ågerfalk, P. J., & Eriksson, O. (2004). Action-oriented conceptual modelling. European
Journal of Information Systems, 13(1), 80-92.
Aguirre-Urreta, M. I., & Marakas, G. M. (2008). Comparing conceptual modeling techniques:
a critical review of the EER vs. OO empirical literature. ACM SIGMIS Database,
39(2), 9-32.
Akhigbe, O., & Lessard, L. (2016). Situating requirements engineering methods within design
science research. In Breakthroughs and Emerging Insights from Ongoing Design
Science Projects: Research-in-progress papers and poster presentations from the 11th
International Conference on Design Science Research in Information Systems and
Technology (DESRIST) (pp. 35-42). Canada: DESRIST 2016.
Allen, G. N., & March, S. T. (2003). Modeling temporal dynamics for business systems.
Journal of Database Management, 14(3), 21-36.
Allen, G. N., & March, S. T. (2006). The effects of state-based and event-based data
representation on user performance in query formulation tasks. MIS Quarterly, 30(2),
269-290.
Allen, G. N., & March, S. T. (2012). A Research Note on Representing Part-Whole Relations
in Conceptual Modeling. MIS Quarterly, 36(3), 945-964.
Ambler, S. (2002). Agile modeling: effective practices for extreme programming and the
unified process. New York: John Wiley
An, Y., Hu, X., & Song, I.-Y. (2010). Maintaining mappings between conceptual models and
relational schemas. Journal of Database Management, 21(3), 36-68.
Anglim, B., Milton, S. K., Rajapakse, J., & Weber, R. (2009, 8-10 June). Current trends and
future directions in the practice of high-level data modeling: An empirical study.
Paper presented at the European Conference of Information Systems, Verona, Italy.
Arisholm, E., & Sjøberg, D. I. (2004). Evaluating the effect of a delegated versus centralized
control style on the maintainability of object-oriented software. IEEE Transactions on
Software Engineering, 30(8), 521-534.
Athenikos, S. J., & Song, I.-Y. (2013). CAM: A Conceptual Modeling Framework based on
the Analysis of Entity Classes and Association Types. Journal of Database
Management, 24(4), 51-80.
Atkins, D., Best, D., Briss, P. A., Eccles, M., Falck-Ytter, Y., Flottorp, S., . . . Henry, D.
(2004). Grading quality of evidence and strength of recommendations. BMJ (Clinical
Research ed.), 328(7454), 1490-1490.
Ausubel, D. P. (1968). Educational psychology: A cognitive view. New York: Holt, Rinehart
and Winston.
Avison, D., & Wood-Harper, A. T. (1986). Multiview—an exploration in information systems
development. Australian Computer Journal, 18(4), 174-179.
Bandara, W., Furtmuller, E., Gorbacheva, E., Miskon, S., & Beekhuyzen, J. (2015).
Achieving rigour in literature reviews: insights from qualitative data analysis and tool-
support. Communications of the Association for Information Systems, 37(8), 154-204.
Bandara, W., Gable, G. G., & Rosemann, M. (2005). Factors and measures of business
process modelling: model building through a multiple case study. European Journal of
Information Systems, 14(4), 347-360.
158 Bibliography
Batra, D. (2008). Conceptual data modeling patterns: Representation and validation. Journal
of Database Management, 16(2), 84-106.
Batra, D. (2009). Unified modeling language (UML) topics: cognitive issues in UML
research. Journal of Database Management, 20(1), I-X.
Batra, D. (2012). An Event-Oriented Data Modeling Technique Based on the Cognitive
Semantics Theory. Journal of Database Management, 23(4), 52-74.
Batra, D., & Davis, J. G. (1992). Conceptual Data Modelling in Database Design: Similarities
and Differences between Expert and Novice Designers. International Journal of Man-
Machine Studies, 37(1), 83-101.
Benbasat, I., Goldstein, D. K., & Mead, M. (1987). The case research strategy in studies of
information systems. MIS Quarterly, 11(3), 369-386.
Bera, P. (2012). Analyzing the Cognitive Difficulties for Developing and Using UML Class
Diagrams for Domain Understanding. Journal of Database Management, 23(3), 1-29.
Bera, P., Burton-Jones, A., & Wand, Y. (2011). Guidelines for designing visual ontologies to
support knowledge identification. MIS Quarterly, 35(4), 883-908.
Bera, P., Burton-Jones, A., & Wand, Y. (2014). How Semantics and Pragmatics Interact in
Understanding Conceptual Models. Information Systems Research, 25(2), 401-419.
Bera, P., & Evermann, J. (2014). Guidelines for using UML association classes and their
effect on domain understanding in requirements engineering. Requirements
Engineering, 19(1), 63-80.
Bera, P., Krasnoperova, A., & Wand, Y. (2010). Using ontology languages for conceptual
modeling. Journal of Database Management, 21(1), 1-28.
Bhattacherjee, A. (2012). Social science research: principles, methods, and practices. North
Charleston: CreateSpace Independent Publishing Platform.
Bodart, F., Patel, A., Sim, M., & Weber, R. (2001). Should optional properties be used in
conceptual modelling? A theory and three empirical tests. Information Systems
Research, 12(4), 384-405.
Boell, S. K., & Cecez-Kecmanovic, D. (2015). On being ‘systematic’ in literature reviews in
IS. Journal of Information Technology, 30(2), 161-173. doi:10.1057/jit.2014.26
Booch, G. (1986). Object-oriented development. IEEE transactions on Software Engineering,
12(2), 211-221.
Booch, G., Rumbaugh, J., & Jacobson, I. (2005). The Unified Modeling Language User Guide
(2nd ed.). Upper Saddle River, NJ: Addison Wesley Longman Inc.
Booch, G., Rumbaugh, J., & Jacobson, I. (2010). The Unified Modeling Language User
Guide, 1999. MA: Addison-Wesley.
Bowen, P. L., O'Farrell, R. A., & Rohde, F. H. (2006). Analysis of competing data structures:
Does ontological clarity produce better end user query performance. Journal of the
Association for Information Systems, 7(8), 514-544.
Bowen, P. L., O'Farrell, R. A., & Rohde, F. H. (2009). An empirical investigation of end-user
query development: the effects of improved model expressiveness vs. complexity.
Information Systems Research, 20(4), 565-584.
Bromage, B. K., & Mayer, R. E. (1986). Quantitative and qualitative effects of repetition on
learning from technical text. Journal of Educational Psychology, 78(4), 271-278.
Browne, G. J., & Parsons, J. (2012). More Enduring Questions in Cognitive IS Research.
Journal of the Association for Information Systems, 13(12), 1000-1011.
Bibliography 159
Bubenko, J. A. (1979). On the Role of'Understanding Models' in conceptual schema design.
In A. l. Furtado & H. L. Morgan (Eds.), 5th International Conference on Very Large
Data Bases (pp. 129-139). Rio de Janeiro, Brazil: IEEE.
Bunge, M. (1977). Treatise on basic philosophy: Ontology I: the furniture of the world (Vol.
3). Dordrecht, The Netherlands: Kluwer Academic Publishers.
Burton-Jones, A., & Grange, C. (2013). From use to effective use: A representation theory
perspective. Information Systems Research, 24(3), 632-658.
Burton-Jones, A., & Meso, P. N. (2006). Conceptualizing systems for understanding: an
empirical test of decomposition principles in object-oriented analysis. Information
Systems Research, 17(1), 38-60.
Burton-Jones, A., & Meso, P. N. (2008). The effects of decomposition quality and multiple
forms of information on novices' understanding of a domain from a conceptual model.
Journal of the Association for Information Systems, 9(12), 748-802.
Burton-Jones, A., Wand, Y., & Weber, R. (2009). Guidelines for empirical evaluations of
conceptual modeling grammars. Journal of the Association for Information Systems,
10(6), 495-532.
Burton-Jones, A., & Weber, R. (1999). Understanding relationships with attributes in entity-
relationship diagrams. Paper presented at the Proceedings of the 20th International
Conference on Information Systems, Atlanta.
Burton-Jones, A., & Weber, R. (2014). Building conceptual modeling on the foundation of
ontology. In H. Topi & A. Tucker (Eds.), Computing handbook: information systems
and information technology (3rd ed., pp. 15.11-15.24). Boca Raton, FL, United States:
CRC Press.
Byström, K., & Järvelin, K. (1995). Task complexity affects information seeking and use.
Information Processing & Management, 31(2), 191-213.
Calder, B. J., Phillips, L. W., & Tybout, A. M. (1981). Designing research for application.
Journal of Consumer Research, 8(2), 197-207.
Campbell, D. J. (1988). Task complexity: A review and analysis. Academy of Management
Review, 13(1), 40-52.
Chan, H. C., Tan, C.-H., & Teo, H.-H. (2014). Data Modeling: An Ontological Perspective of
Pointers. Journal of Database Management, 25(4), 17-37.
Chan, H. C., Wei, K. K., & Siau, K. L. (1993). User-database interface: The effect of
abstraction levels on query performance. MIS Quarterly, 17(4), 441-464.
Chang, H. C. (2010). A new perspective on Twitter hashtag use: Diffusion of innovation
theory. Proceedings of the American Society for Information Science and Technology,
47(1), 1-4.
Checkland, P. B., & Scholes, J. (2000). Soft systems methodology in action: Including a 30-
year retrospective. BASINGSTOKE: Macmillan Press.
Chen, H., Chiang, R. H., & Storey, V. C. (2012). Business intelligence and analytics: From
big data to big impact. MIS Quarterly, 36(4), 1165-1188.
Chen, P. P.-S. (1976). The entity-relationship model—toward a unified view of data. ACM
Transactions on Database Systems (TODS), 1(1), 9-36.
Chen, P. P.-S. (2006). Suggested research directions for a new frontier–active conceptual
modeling. In D. W. Embley, A. Olivé, & S. Ram (Eds.), Proceedings of International
Conference on Conceptual Modeling (pp. 1-4). Heidelberg: Springer.
160 Bibliography
Chen, W., & Hirschheim, R. (2004). A paradigmatic and methodological examination of
information systems research from 1991 to 2001. Information Systems Journal, 14(3),
197-235.
Chisholm, R. M. (1996). A realistic theory of categories: An essay on ontology. Cambridge,
UK: Cambridge University Press.
Chua, C. E. H., Chiang, R. H., & Lim, E.-P. (2002). On conceptual micro-object modeling.
Journal of Database Management, 13(3), 1-16.
Chua, C. E. H., Storey, V. C., & Chiang, R. H. (2012). Knowledge representation: a
conceptual modeling approach. Journal of Database Management, 23(1), 1-30.
Cierniak, G., Scheiter, K., & Gerjets, P. (2009). Explaining the split-attention effect: Is the
reduction of extraneous cognitive load accompanied by an increase in germane
cognitive load? Computers in Human Behavior, 25(2), 315-324.
Clarke, R., Burton-Jones, A., & Weber, R. (2016). On the Ontological Quality and Logical
Quality of Conceptual-Modeling Grammars: The Need for a Dual Perspective.
Information Systems Research, 27(2), 365-382.
Clegg, B., & Shaw, D. (2008). Using process‐oriented holonic (PrOH) modelling to increase
understanding of information systems. Information Systems Journal, 18(5), 447-477.
Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Hillsdale,
NJ: Lawrence Earlbaum Associates.
Compeau, D., Higgins, C. A., & Huff, S. (1999). Social cognitive theory and individual
reactions to computing technology: A longitudinal study. MIS Quarterly, 23(2), 145-
158.
Compeau, D., Marcolin, B., Kelley, H., & Higgins, C. A. (2012). Research commentary—
Generalizability of information systems research using student subjects—A reflection
on our practices and recommendations for future research. Information Systems
Research, 23(4), 1093-1109.
Currim, F., & Ram, S. (2012). Modeling spatial and temporal set-based constraints during
conceptual database design. Information Systems Research, 23(1), 109-128.
D'aubeterre, F., Singh, R., & Iyer, L. (2008). Secure activity resource coordination: empirical
evidence of enhanced security awareness in designing secure business processes.
European Journal of Information Systems, 17(5), 528-542.
da Silva, A. R. (2015). Model-driven engineering: A survey supported by the unified
conceptual model. Computer Languages, Systems & Structures, 43, 139-155.
Dahanayake, A., Sol, H., & Stojanovic, Z. (2003). Methodology evaluation framework for
component-based system development. Journal of Database Management, 14(1), 1-
26.
Darke, P., & Shanks, G. (1997). User Viewpoint Modelling: Understanding and Representing
User Viewpoints During Requirements Definition. Information Systems Journal 7(3),
213-219.
Davern, M., Shaft, T., & Te'eni, D. (2012a). Cognition matters: Enduring questions in
cognitive IS research. Journal of the Association for Information Systems, 13(4), 273-
314.
Davern, M., Shaft, T., & Te'eni, D. (2012b). More Enduring Questions in Cognitive IS
Research: A Reply. Journal of the Association for Information Systems, 13(12), 1012-
1016.
Bibliography 161
Davies, I., Green, P., Rosemann, M., Indulska, M., & Gallo, S. (2006). How do practitioners
use conceptual modeling in practice? Data & Knowledge Engineering, 58(3), 358-
380.
Davis, F. D. (1989). Perceived usefulness, perceived ease of use, and user acceptance of
information technology. MIS Quarterly, 13(3), 319-340.
de Albuquerque, J. P., & Christ, M. (2015). The tension between business process modelling
and flexibility: Revealing multiple dimensions with a sociomaterial approach. The
Journal of Strategic Information Systems, 24(3), 189-202.
de Sousa Saraiva, J., & da Silva, A. R. (2010). Evaluation of mde tools from a metamodeling
perspective. Journal of Database Management, 19(4), 21-46.
Delfmann, P., Herwig, S., Lis, Ł., & Stein, A. (2015). Supporting distributed conceptual
modelling through naming conventions-a tool-based linguistic approach. Enterprise
Modelling and Information Systems Architectures, 4(2), 3-19.
Denning, P. J. (2003). Great principles of computing. Communications of the ACM, 46(11),
15-20.
Dietz, G., & Juhrisch, M. (2012). Negotiating language barriers–a methodology for cross-
organisational conceptual modelling. European Journal of Information Systems, 21(3),
229-254.
Dobing, B., & Parsons, J. (2008). Dimensions of UML diagram use: a survey of practitioners.
Journal of Database Management, 19(1), 1-18.
Domingos, P. (1999). The role of Occam's razor in knowledge discovery. Data Mining and
Knowledge Discovery, 3(4), 409-425.
Dominguez, E., Rubio, A. L., & Zapata, M. A. (2002). Dynamic semantics of UML state
machines: a metamodeling perspective. Journal of Database Management, 13(4), 20-
38.
Dori, D., & Goodman, M. (1996). From object-process analysis to object-process design.
Annals of Software Engineering, 2(1), 25-50.
Dreiling, A., Rosemann, M., van der Aalst, W. M. P., Heuser, L., & Schulz, K. (2006).
Model-based software configuration: patterns and languages. European Journal of
Information Systems, 15(6), 583-600.
Dubé, L., & Paré, G. (2003). Rigor in information systems positivist case research: current
practices, trends, and recommendations. MIS Quarterly, 24(4), 597-636.
Dunn, C. L., Gerard, G. J., & Grabski, S. V. (2011). Diagrammatic Attention Management
and the Effect of Conceptual Model Structure on Cardinality Validation. Journal of
the Association for Information Systems, 12(8), 585-605.
Dussart, A., Aubert, B. A., & Patry, M. (2004). An evaluation of inter-organizational
workflow modelling formalisms. Journal of Database Management, 15(2), 74-104.
Dwivedi, Y. K., Wastell, D., Laumer, S., Henriksen, H. Z., Myers, M. D., Bunker, D., . . .
Srivastava, S. C. (2015). Research on information systems failures and successes:
Status update and future directions. Information Systems Frontiers, 17(1), 143-157.
Eisenhardt, K. M. (1989). Building theories from case study research. Academy of
Management Review, 14(4), 532-550.
Elliott, A. C., & Woodward, W. A. (2007). Statistical analysis quick reference guidebook:
With SPSS examples. Thousand Oaks: Sage.
Erickson, J., Lyytinen, K., & Siau, K. L. (2005). Agile modeling, agile software development,
and extreme programming: the state of research. Journal of Database Management,
16(4), 88-100.
162 Bibliography
Evermann, J. (2005). Towards a cognitive foundation for knowledge representation.
Information Systems Journal, 15(2), 147-178.
Evermann, J. (2008). Theories of meaning in schema matching: A review. Journal of
Database Management, 19(3), 55-82.
Evermann, J., & Wand, Y. (2001). Towards ontologically based semantics for UML
constructs. In Proceedings of International Conference on Conceptual Modeling (ER
2001) (pp. 354-367). Yokohama, Japan: Springer.
Evermann, J., & Wand, Y. (2005). Ontology based object-oriented domain modelling:
fundamental concepts. Requirements Engineering, 10(2), 146-160.
Evermann, J., & Wand, Y. (2006). Ontological modeling rules for UML: An empirical
assessment. The Journal of Computer Information Systems, 46(5), 14-19.
Evermann, J., & Wand, Y. (2009). Ontology based object-oriented domain modeling:
Representing behavior. Journal of Database Management, 20(1), 48-77.
Fernández-Medina, E., Trujillo, J. C., & Piattini, M. (2007). Model-driven multidimensional
modeling of secure data warehouses. European Journal of Information Systems, 16(4),
374-389.
Fettke, P. (2009). How conceptual modeling is used. Communications of the Association for
Information Systems, 25(1), 571-592.
Fettke, P., & Loos, P. (2005). Ontological Analysis of Reference Models. In P. F. Green & M.
Rosemann (Eds.), Business Systems Analysis with Ontologies (pp. 56-81). Hershey,
Pennsylvania: Idea Group.
Fickinger, T., & Recker, J. (2013). Construct Redundancy In Process Modelling Grammars:
Improving The Explanatory Power Of Ontological Analysis. In Proceedings of the
21st European Conference on Information Systems (pp. 1-12). Utrecht, The
Netherlands: Association for Information Systems.
Figl, K. (2017). Comprehension of Procedural Visual Business Process Models–A Literature
Review. Business & Information Systems Engineering, 59(1), 41-67.
Figl, K., & Laue, R. (2015). Influence factors for local comprehensibility of process models.
International Journal of Human-Computer Studies, 82, 96-110.
Figl, K., Mendling, J., & Strembeck, M. (2013). The influence of notational deficiencies on
process model comprehension. Journal of the Association for Information Systems,
14(6), 312-338.
Figl, K., & Recker, J. (2014). Exploring cognitive style and task-specific preferences for
process representations. Requirements Engineering, 21(1), 63-85.
Figl, K., Recker, J., & Mendling, J. (2013). A study on the effects of routing symbol design
on process model comprehension. Decision Support Systems, 54(2), 1104-1118.
Fonseca, F., & Martin, J. (2007). Learning the differences between ontologies and conceptual
schemas through ontology-driven information systems. Journal of the Association for
Information Systems, 8(2), 129-142.
Fornell, C., & Larcker, D. F. (1981). Evaluating structural equation models with unobservable
variables and measurement error. Journal of Marketing Research, 18(1), 39-50.
Francis, J. J., Johnston, M., Robertson, C., Glidewell, L., Entwistle, V., Eccles, M. P., &
Grimshaw, J. M. (2010). What is an adequate sample size? Operationalising data
saturation for theory-based interview studies. Psychology and Health, 25(10), 1229-
1245.
Bibliography 163
Garrigós, I., Pardillo, J., Mazón, J.-N., Zubcoff, J., Trujillo, J. C., & Romero, R. (2012). A
Conceptual modeling personalization framework for OLAP. Journal of Database
Management, 23(4), 1-16.
Gehlert, A., & Esswein, W. (2005). Ontological Analysis Reconsidered - A Formal Approach.
In M.-A. Sicilia, S. Sánchez-Alonso, E. García-Barriocanal, & J. J. Cuadrado-Gallego
(Eds.), Ontology, Conceptualizations and Epistemology for Software and Systems
Engineering, Proceedings of the ONTOSE 2005 International Workshop Madrid,
Spain: University of Alcalá.
Geiger, J. (2010). Why Do We Model? Information Management 20(6), 32-33.
Gemino, A., & Parker, D. (2009). Use case diagrams in support of use case modeling:
Deriving understanding from the picture. Journal of Database Management, 20(1), 1-
24.
Gemino, A., & Wand, Y. (2003). Evaluating modeling techniques based on models of
learning. Communications of the ACM, 46(10), 79-84.
Gemino, A., & Wand, Y. (2004). A framework for empirical evaluation of conceptual
modeling techniques. Requirements Engineering, 9(4), 248-260.
Gemino, A., & Wand, Y. (2005). Complexity and clarity in conceptual modeling: comparison
of mandatory and optional properties. Data & Knowledge Engineering, 55(3), 301-
326.
Genero, M., Fernández-Saez, A. M., Nelson, H. J., Poels, G., & Piattini, M. (2011). Research
review: a systematic literature review on the quality of UML models. Journal of
Database Management, 22(3), 46-70.
Glass, R. L. (2013). The Queensland Health Payroll Debacle. Information Systems
Management, 30(1), 89-90.
Goodchild, M. F. (2012). Twenty years of progress: GIScience in 2010. Journal of Spatial
Information Science, 2010(1), 3-20.
Green, P. (1996). An ontological analysis of information systems analysis and design (ISAD)
grammars in upper case tools. (PhD), The University of Queensland, Brisbane,
Australia.
Green, P., & Rosemann, M. (1999). An Ontological Analysis of Integrated Process
Modelling. In M. Jarke & A. Oberweis (Eds.), Advanced Information Systems
Engineering: 11th International Conference, CAiSE"99 Heidelberg, Germany, June
14—18, 1999 Proceedings (pp. 225-240). Berlin, Heidelberg: Springer Berlin
Heidelberg.
Green, P., & Rosemann, M. (2004). Applying ontologies to business and systems modelling
techniques and perspectives: Lessons learned. Journal of Database Management,
15(2), 105-117.
Green, P., Rosemann, M., Indulska, M., & Manning, C. (2007). Candidate interoperability
standards: An ontological overlap analysis. Data & Knowledge Engineering, 62(2),
274-291.
Green, P., Rosemann, M., Indulska, M., & Recker, J. (2011). Complementary use of modeling
grammars. Scandinavian Journal of Information Systems, 23(1), 59-86.
Grossman, M., Aronson, J. E., & McCarthy, R. V. (2005). Does UML make the grade?
Insights from the software development community. Information and Software
Technology, 47(6), 383-397.
Guest, G., Bunce, A., & Johnson, L. (2006). How many interviews are enough? An
experiment with data saturation and variability. Field Methods, 18(1), 59-82.
164 Bibliography
Guizzardi, G. (2005). Ontological foundations for structural conceptual models (Vol. 015).
Enschede, The Netherlands: Telematica Instituut.
Hadar, I., & Soffer, P. (2006). Variations in conceptual modeling: classification and
ontological analysis. Journal of the Association for Information Systems, 7(8), 568-
592.
Halpin, T. A. (2002). Metaschemas for ER, ORM and UML data models: A comparison.
Journal of Database Management, 13(2), 20-30.
Henderson-Sellers, B. (2002). The use of subtypes and stereotypes in the UML model.
Journal of Database Management, 13(2), 43-50.
Hirschheim, R., Klein, H. K., & Lyytinen, K. (1994). Information Systems Development and
Data Modeling: Conceptual and Philosophical Foundations. . Cambridge,
Massachusetts: Cambridge University Press.
Hoffer, J., Ramesh, V., & Topi, H. (2012). Modern database management (11th ed.). Boston,
MA: Prentice Hall Press.
Hoyningen-Huene, P. (2013). Systematicity: The nature of science. Oxford, UK: Oxford
University Press.
Hutchinson, J., Whittle, J., Rouncefield, M., & Kristoffersen, S. (2011). Empirical Assessment
of MDE in Industry. Paper presented at the 33rd International Conference on Software
Engineering (ICSE), Honolulu, HI, USA.
Indulska, M., Recker, J., Rosemann, M., & Green, P. (2009). Business process modeling:
Current issues and future challenges. Paper presented at the Advanced Information
Systems Engineering, Heidelberg, Berlin.
Irwin, G., & Turk, D. (2005). An ontological analysis of use case modeling grammar. Journal
of the Association for Information Systems, 6(1), 1-36.
Jabbari Sabegh, M. A., Lukyanenko, R., Recker, J. C., Samuel, B., & Castellanos, A. (2017).
Conceptual modeling research in information systems: What we now know and what
we still do not know. Paper presented at the Symposium on Research in Systems
Analysis and Design (AIS SIGSAND 2017).
Jabbari Sabegh, M. A., & Recker, J. (2017). Combined Use of Conceptual Models in Practice:
An Exploratory Study. Journal of Database Management, 28(2), 56-88.
Jabbari Sabegh, M. A., Recker, J., & Green, P. (2016). Designing Experiments to Test the
Theory of Combined Ontological Coverage. Paper presented at the International
Conference on Information Systems, Dublin, Ireland
Joosten, S., & Purao, S. (2002). A rigorous approach for mapping workflows to object-
oriented is models. Journal of Database Management, 13(4), 1-19.
Kaur, K., & Rani, R. (2013, 6-9 Oct. 2013). Modeling and querying data in NoSQL
databases. Paper presented at the 2013 IEEE International Conference on Big Data.
Kent, W. (1978). Data and Reality: Basic assumptions in data processing reconsidered.
North-Holland, Amsterdam: North-Holland Pub. Co.
Khatri, V., & Vessey, I. (2016). Understanding the Role of IS and Application Domain
Knowledge on Conceptual Schema Problem Solving: A Verbal Protocol Study.
Journal of the Association for Information Systems, 17(12), 759 – 803
Khatri, V., Vessey, I., Ramesh, V., Clay, P., & Park, S.-J. (2006). Understanding conceptual
schemas: Exploring the role of application and IS domain knowledge. Information
Systems Research, 17(1), 81-99.
Bibliography 165
Kim, H. M., Fox, M. S., & Sengupta, A. (2007). How to build enterprise data models to
achieve compliance to standards or regulatory requirements (and share data). Journal
of the Association for Information Systems, 8(2), 105-128.
Kim, J., Hahn, J., & Hahn, H. (2000). How do we understand a system with (so) many
diagrams? Cognitive integration processes in diagrammatic reasoning. Information
Systems Research, 11(3), 284-303.
Kirschner, P. A. (2002). Cognitive load theory: Implications of cognitive load theory on the
design of learning. Learning and Instruction, 12(1), 1-10.
Kitchenham, B., Brereton, O. P., Budgen, D., Turner, M., Bailey, J., & Linkman, S. (2009).
Systematic literature reviews in software engineering–a systematic literature review.
Information and Software Technology, 51(1), 7-15.
Koschmider, A., Song, M., & Reijers, H. A. (2010). Social software for business process
modeling. Journal of Information Technology, 25(3), 308-322.
Krishnan, M. S., Mukhopadhyay, T., & Kriebel, C. H. (2004). A decision model for software
maintenance. Information Systems Research, 15(4), 396-412.
Krogstie, J. (2001). Using a Semiotic Framework to Evaluate UML for the Development of
Models of High Quality. In Unified modeling language: Systems analysis, design and
development issues (pp. 89-106): IGI Global.
Krogstie, J., Sindre, G., & Jørgensen, H. (2006). Process models representing knowledge for
action: a revised quality framework. European Journal of Information Systems, 15(1),
91-102.
Kung, C. H., & Sølvberg, A. (1986). Activity Modeling and Behavior Modeling of
Information Systems. In T. W. Olle, H. G. Sol, & A. A. Verrijn-Stuart (Eds.),
Information Systems Design Methodologies: Improving the Practice (pp. 145-171).
Amsterdam: North-Holland.
La Rosa, M., Reijers, H. A., van der Aalst, W. M. P., Dijkman, R. M., Mendling, J., Dumas,
M., & GarcíA-BañUelos, L. (2011). APROMORE: An advanced process model
repository. Expert Systems with Applications, 38(6), 7029-7040.
Lange, C. F., Chaudron, M. R. V., & Muskens, J. (2006). In practice: UML software
architecture and design description. IEEE Software, 23(2), 40-46.
Larsen, T. J., Niederman, F., Limayem, M., & Chan, J. (2009). The role of modelling in
achieving information systems success: UML to the rescue? Information Systems
Journal, 19(1), 83-117.
Laurier, W., & Poels, G. (2012). Research note: ontology-based structuring of conceptual data
modeling patterns. Journal of Database Management, 23(3), 50-64.
Lee, A. S. (1991). Integrating positivist and interpretive approaches to organizational
research. Organization Science, 2(4), 342-365.
Lee, A. S., & Hubona, G. S. (2009). A scientific basis for rigor in information systems
research. MIS Quarterly, 33(2), 237-262.
Leonardi, P. M. (2011). When flexible routines meet flexible technologies: Affordance,
constraint, and the imbrication of human and material agencies. MIS Quarterly, 30(1),
147-167.
Liddle, S. W. (2011). Model-Driven Software Development. In D. W. Embley & B. Thalheim
(Eds.), Handbook of Conceptual Modeling: Theory, Practice, and Research
Challenges (pp. 17-54). Berlin, Heidelberg: Springer Berlin Heidelberg.
Liddle, S. W., & Embley, D. W. (2007). A Common Core for Active Conceptual Modeling
for Learning from Surprises. In P. P. Chen & L. Y. Wong (Eds.), Active Conceptual
166 Bibliography
Modeling of Learning: Next Generation Learning-Base System Development (pp. 47-
56). Berlin, Heidelberg: Springer Berlin Heidelberg.
Lincoln, Y. S., Lynham, S. A., & Guba, E. G. (2011). Paradigmatic controversies,
contradictions, and emerging confluences, revisited. In The Sage handbook of
qualitative research (4th ed., pp. 97-128). Thousand Oaks: Sage Publication.
Lindland, O. I., Sindre, G., & Solvberg, A. (1994). Understanding quality in conceptual
modeling. IEEE Software, 11(2), 42-49.
Liu, F., & Myers, M. D. (2011). An analysis of the AIS basket of top journals. Journal of
Systems and Information Technology, 13(1), 5-24.
Liu, L., Pu, C., & Ruiz, D. D. (2004). A systematic approach to flexible specification,
composition, and restructuring of workflow activities. Journal of Database
Management, 15(1), 1-40.
Loucopoulos, P., & Kadir, W. M. N. (2008). BROOD: business rules-driven object oriented
design. Journal of Database Management, 19(1), 41-73.
Lukyanenko, R., & Parsons, J. (2013). Is Traditional Conceptual Modeling Becoming
Obsolete? In W. Ng, V. C. Storey, & J. C. Trujillo (Eds.), Proceedings of 32th
International Conference Conceptual Modeling, ER 2013, Hong-Kong, China,
November 11-13, 2013. (pp. 61-73). Berlin, Heidelberg: Springer Berlin Heidelberg.
Lukyanenko, R., & Parsons, J. (2014). Using Field Experimentation to Understand
Information Quality in User-generated Content. Paper presented at the
CodeCon@MIT.
Lukyanenko, R., Parsons, J., & Samuel, B. M. (2015). Do We Need an Instance-Based
Conceptual Modeling Grammar? Paper presented at the Symposium on Research in
Systems Analysis and Design (AIS SIGSAND 2015).
Lukyanenko, R., Parsons, J., & Wiersma, Y. F. (2014). The IQ of the Crowd: Understanding
and Improving Information Quality in Structured User-Generated Content.
Information Systems Research, 25(4), 669-689.
Lukyanenko, R., Wiersma, Y., Huber, B., Parsons, J., Wachinger, G., & Meldt, R. (2017).
Representing Crowd Knowledge: Guidelines for Conceptual Modeling of User-
generated Content. Journal of the Association for Information Systems, 18(4), 297-
339.
Lyytinen, K. (1987). Two views of information modeling. Information & Management, 12(1),
9-19.
Ma, Z. (2005). A conceptual design methodology for fuzzy relational databases. Journal of
Database Management, 16(2), 66-83.
Maes, A., & Poels, G. (2007). Evaluating quality of conceptual modelling scripts based on
user perceptions. Data & Knowledge Engineering, 63(3), 701-724.
Marcus, N., Cooper, M., & Sweller, J. (1996). Understanding instructions. Journal of
Educational Psychology, 88(1), 49-63.
Masri, K., Parker, D., & Gemino, A. (2008). Using iconic graphics in entity-relationship
diagrams: the impact on understanding. Journal of Database Management, 19(3), 22-
41.
Maxwell, J. A. (1992). Understanding and validity in qualitative research. Harvard
Educational Review, 62(3), 279-301.
Mayer, R. E. (1983). Can you repeat that? Qualitative effects of repetition and advance
organizers on learning from science prose. Journal of Educational Psychology, 75(1),
40-49.
Bibliography 167
Mayer, R. E. (1989). Models for understanding. Review of Educational Research, 59(1), 43-
64.
Mayer, R. E. (2005). Cognitive theory of multimedia learning. In R. E. Mayer (Ed.), The
Cambridge Handbook of Multimedia Learning (pp. 31-48). New York: Cambridge
University Press.
Mayer, R. E. (2009). Multimedia learning (2nd ed.). United States of America: Cambridge
university press.
Mendling, J., Reijers, H. A., & Cardoso, J. (2007). What Makes Process Models
Understandable? In G. Alonso, P. Dadam, & M. Rosemann (Eds.), Business Process
Management: 5th International Conference, BPM 2007, Brisbane, Australia,
September 24-28, 2007. Proceedings (pp. 48-63). Berlin, Heidelberg: Springer Berlin
Heidelberg.
Mendling, J., Reijers, H. A., & Recker, J. (2010). Activity labeling in process modeling:
Empirical insights and recommendations. Information Systems, 35(4), 467-482.
Mendling, J., Reijers, H. A., & van der Aalst, W. M. P. (2010). Seven process modeling
guidelines (7PMG). Information and Software Technology, 52(2), 127-136.
Mendling, J., & Strembeck, M. (2008). Influence Factors of Understanding Business Process
Models. In W. Abramowicz & D. Fensel (Eds.), Business Information Systems: 11th
International Conference, BIS 2008, Innsbruck, Austria, May 5-7, 2008. Proceedings
(pp. 142-153). Berlin, Heidelberg: Springer Berlin Heidelberg.
Merschbrock, C., & Munkvold, B. E. (2012). A Research Review on Building Information
Modeling in Construction―An Area Ripe for IS Research. Communications of the
Association for Information Systems, 31(10), 207-228.
Mertens, D. M. (2014). Research and evaluation in education and psychology: Integrating
diversity with quantitative, qualitative, and mixed methods (4th ed.). Los Angeles:
Sage Publications.
Milton, S. K., & Kazmierczak, E. (2004). An ontology of data modelling languages: a study
using a common-sense realistic ontology. Journal of Database Management, 15(2),
19-38.
Milton, S. K., Rajapakse, J., & Weber, R. (2012). Ontological clarity, cognitive engagement,
and conceptual model quality evaluation: An experimental investigation. Journal of
the Association for Information Systems, 13(9), 657-693.
Mingers, J. (2001). Combining IS research methods: towards a pluralist methodology.
Information Systems Research, 12(3), 240-259.
Mitra, S., Bagchi, A., & Bandyopadhyay, A. K. (2007). Design of a data model for social
network applications. Journal of Database Management, 18(4), 51-79.
Montero, S., Díaz, P., & Aedo, I. (2007). From requirements to implementations: a model-
driven approach for web development. European Journal of Information Systems,
16(4), 407-419.
Moody, D. L. (2000). Building links between IS research and professional practice:
improving the relevance and impact of IS research. Paper presented at the Proceedings
of the twenty first International Conference on Information systems.
Moody, D. L. (2005). Theoretical and practical issues in evaluating the quality of conceptual
models: current state and future directions. Data & Knowledge Engineering, 55(3),
243-276.
168 Bibliography
Moody, D. L. (2009). The “physics” of notations: toward a scientific basis for constructing
visual notations in software engineering. IEEE Transactions on Software Engineering,
35(6), 756-779.
Mussbacher, G., Amyot, D., Breu, R., Bruel, J.-M., Cheng, B. H. C., Collet, P., . . . Whittle, J.
(2014). The Relevance of Model-Driven Engineering Thirty Years from Now. In J.
Dingel, W. Schulte, I. Ramos, S. Abrahão, & E. Insfran (Eds.), Model-Driven
Engineering Languages and Systems: 17th International Conference, MODELS 2014,
Valencia, Spain, September 28 – October 3, 2014. Proceedings (pp. 183-200). Cham:
Springer International Publishing.
Myers, M. D., & Newman, M. (2007). The qualitative interview in IS research: Examining
the craft. Information and Organization, 17(1), 2-26.
Nesbit, J. C., & Adesope, O. O. (2006). Learning with concept and knowledge maps: A meta-
analysis. Review of Educational Research, 76(3), 413-448.
Neuman, W. L. (2006). Social research methods: Qualitative and quantitative approaches
(6th ed.). Boston: Pearson/Allyn and Bacon.
Oei, J. L. H., Van Hemmen, L. J. G. T., Falkenberg, E. D., & Brinkkemper, S. (1992). The
meta model hierarchy: a framework for information systems concepts and techniques.
Retrieved from The Netherlands:
Offir, B., Lev, Y., & Bezalel, R. (2008). Surface and deep learning processes in distance
education: Synchronous versus asynchronous systems. Computers & Education,
51(3), 1172-1183.
Olle, T. W., Sol, H. G., & Verrijn-Stuart, A. A. (1982). Information System Design
Methodologies: A Comparative Review: Proceedings of the IFIP WG 8.1 Working
Conference on Comparative Review of Information Systems Design Methodologies
North-Holland, Amsterdam, The Netherlands.
OMG. (2011a). Business Process Model and Notation (BPMN) Version 2.0. Retrieved from
http://www.omg.org/spec/BPMN/2.0
OMG. (2011b). Unified Modeling Language (UML), Version 2.4.1. Retrieved from
http://www.omg.org/spec/UML/2.4.1/
Opdahl, A. L., & Henderson-Sellers, B. (2002). Ontological evaluation of the UML using the
Bunge–Wand–Weber model. Software and Systems Modeling, 1(1), 43-67.
Opdahl, A. L., & Henderson-Sellers, B. (2004). A template for defining enterprise modelling
constructs. Journal of Database Management, 15(2), 39-73.
Orlikowski, W. J. (1993). CASE tools as organizational change: Investigating incremental
and radical changes in systems development. MIS Quarterly, 17(3), 309-340.
Paas, F. G., Tuovinen, J. E., Tabbers, H., & Van Gerven, P. W. (2003). Cognitive load
measurement as a means to advance cognitive load theory. Educational Psychologist,
38(1), 63-71.
Pallant, J. (2007). SPSS survival manual: a step by step guide to data analysis using SPSS for
Windows (version 15) (3rd ed.). England: McGraw-Hill Education.
Palvia, P., Jacks, T., & Brown, W. (2015). Critical issues in EHR implementation: provider
and vendor perspectives. Communications of the Association for Information Systems,
36(36), 706-726.
Palvia, P., Leary, D., Mao, E., Midha, V., Pinjani, P., & Salam, A. (2004). Research
methodologies in MIS: an update. Communications of the Association for Information
Systems, 14(24), 526-542.
Bibliography 169
Pardillo, J., Mazón, J.-N., & Trujillo, J. C. (2011). An MDA Approach and QVT
Transformations for the Integrated Development of Goal-Oriented Data Warehouses
and Data Marts. Journal of Database Management, 22(1), 43-68.
Paré, G., Trudel, M.-C., Jaana, M., & Kitsiou, S. (2015). Synthesizing information systems
knowledge: A typology of literature reviews. Information & Management, 52(2), 183-
199.
Parsons, J. (2002). Effects of local versus global schema diagrams on verification and
communication in conceptual data modeling. Journal of Management Information
Systems, 19(3), 155-183.
Parsons, J. (2011). An experimental study of the effects of representing property precedence
on the comprehension of conceptual schemas. Journal of the Association for
Information Systems, 12(6), 401-422.
Parsons, J., & Cole, L. (2005). What do the pictures mean? Guidelines for experimental
evaluation of representation fidelity in diagrammatical conceptual modeling
techniques. Data & Knowledge Engineering, 55(3), 327-342.
Parsons, J., & Wand, Y. (2008). Using cognitive principles to guide classification in
information systems modeling. MIS Quarterly, 30(4), 839-868.
Parsons, J., & Wand, Y. (2013). Extending Classification Principles from Information
Modeling to Other Disciplines. Journal of the Association for Information Systems,
14(3), 245-273.
Parsons, J., & Wand, Y. (2014). A foundation for open information environments. Paper
presented at the European Conference on Information Systems, Tel Aviv, Israel.
Petre, M. (2013). UML in practice. Paper presented at the Proceedings of the 2013
International Conference on Software Engineering, San Francisco, California.
Petri, C. A. (1962). Fundamentals of a Theory of Asynchronous Information Flow. In C. M.
Popplewell (Ed.), IFIP Congress 62: Information Processing (pp. 386-390). Munich,
Germany: North-Holland.
Pettit, R. G. (2008). Increasing Confidence in Concurrent Software through Architectural
Analysis. In F. Kordon & T. Vardanega (Eds.), Reliable Software Technologies – Ada-
Europe 2008: 13th Ada-Europe International Conference on Reliable Software
Technologies, Venice, Italy, June 16-20, 2008. (pp. 199-210). Berlin, Heidelberg:
Springer.
Pettit, R. G., & Street, J. A. (2005). Lessons Learned Applying UML in the Design of Mission
Critical Software. In N. Jardim Nunes, B. Selic, A. Rodrigues da Silva, & A. Toval
Alvarez (Eds.), UML Modeling Languages and Applications: < <UML> > 2004
Satellite Activities, Lisbon, Portugal, October 11-15, 2004, Revised Selected Papers
(pp. 129-137). Berlin, Heidelberg: Springer.
Poels, G. (2011). Understanding business domain models: The effect of recognizing resource-
event-agent conceptual modeling structures. Journal of Database Management, 22(1),
69-101.
Poels, G., Maes, A., Gailly, F., & Paemeleire, R. (2011). The pragmatic quality of Resources‐Events‐Agents diagrams: an experimental evaluation. Information Systems Journal,
21(1), 63-89.
Poppe, E., Brown, R., Recker, J., Johnson, D., & Vanderfeesten, I. (2017). Design and
Evaluation of Virtual Environments Mechanisms to Support Remote Collaboration on
Complex Process Diagrams. Information Systems, 66, 59-81.
170 Bibliography
Power, D. (2011). Data modeling: The lingua franca of a successful MDM effort. Information
Management, 21(1), 28-29.
Purao, S., Storey, V. C., & Han, T. (2003). Improving analysis pattern reuse in conceptual
design: Augmenting automated processes with supervised learning. Information
Systems Research, 14(3), 269-290.
Rai, A. (2017). Diversity of design science contributions. MIS Quarterly, 41(11), iii–xviii.
Ramesh, B., Cao, L., & Baskerville, R. (2010). Agile requirements engineering practices and
challenges: an empirical study. Information Systems Journal, 20(5), 449-480.
Ramesh, V., & Browne, G. J. (1999). Expressing casual relationships in conceptual database
schemas. Journal of Systems and Software, 4, 225-232.
Recker, J. (2007). A socio-pragmatic constructionist framework for understanding quality in
process modelling. Australasian Journal of Information Systems, 14(2), 43-63.
Recker, J. (2010a). Continued use of process modeling grammars: the impact of individual
difference factors. European Journal of Information Systems, 19(1), 76-92.
Recker, J. (2010b). Explaining usage of process modeling grammars: Comparing three
theoretical models in the study of two grammars. Information & Management, 47(5),
316-324.
Recker, J. (2012). “Modeling with tools is easier, believe me”—The effects of tool
functionality on modeling grammar usage beliefs. Information Systems, 37(3), 213-
226.
Recker, J. (2013). Empirical investigation of the usefulness of Gateway constructs in process
models. European Journal of Information Systems, 22(6), 673-689.
Recker, J. (2014). Theorizing the Use of Multiple Conceptual Models in Combination Paper
presented at the 25th Australasian Conference on Information Systems 8th-10th Dec
2014, , Auckland New Zealand.
Recker, J. (2015). Research on conceptual modelling: less known knowns and more unknown
unknowns, please. Paper presented at the 11th Asia-Pacific Conference on Conceptual
Modelling, Sydney, Australia.
Recker, J., & Dreiling, A. (2011). The effects of content presentation format and user
characteristics on novice developers’ understanding of process models.
Communications of the Association for Information Systems, 28(6), 65-84.
Recker, J., Indulska, M., Rosemann, M., & Green, P. (2010). The ontological deficiencies of
process modeling in practice. European Journal of Information Systems, 19(5), 501-
525.
Recker, J., Mendling, J., & Hahn, C. (2013). How collaborative technology supports cognitive
processes in collaborative process modeling: A capabilities-gains-outcome model.
Information Systems, 38(8), 1031-1045.
Recker, J., Reijers, H. A., & van de Wouw, S. G. (2014). Process model comprehension: the
effects of cognitive abilities, learning style, and strategy. Communications of the
Association for Information Systems, 34(9), 199-222.
Recker, J., & Rosemann, M. (2010a). A measurement instrument for process modeling
research: development, test and procedural model. Scandinavian Journal of
Information Systems, 22(2), 3-30.
Recker, J., & Rosemann, M. (2010b). The measurement of perceived ontological deficiencies
of conceptual modeling grammars. Data & Knowledge Engineering, 69(5), 516-532.
Recker, J., Rosemann, M., Green, P., & Indulska, M. (2011). Do ontological deficiencies in
modeling grammars matter? MIS Quarterly, 35(1), 57-79.
Bibliography 171
Recker, J., Rosemann, M., Indulska, M., & Green, P. (2009). Business process modeling-a
comparative analysis. Journal of the Association for Information Systems, 10(4), 333-
363.
Reijers, H. A., Freytag, T., Mendling, J., & Eckleder, A. (2011). Syntax highlighting in
business process models. Decision Support Systems, 51(3), 339-349.
Reinhartz-Berger, I., & Sturm, A. (2008). Enhancing UML models: a domain analysis
approach. Journal of Database Management, 19(1), 74-94.
Rittgen, P. (2006). A language-mapping approach to action-oriented development of
information systems. European Journal of Information Systems, 15(1), 70-81.
Rosemann, M. (2006). Potential pitfalls of process modeling: part A. Business Process
Management Journal, 12(2), 249-254.
Rosemann, M., & Green, P. (2002). Developing a meta model for the Bunge–Wand–Weber
ontological constructs. Information Systems, 27(2), 75-91.
Rosemann, M., Green, P., & Indulska, M. (2004). A reference methodology for conducting
ontological analyses. In H. Lu, W. Chu, P. Atzeni, S. Zhou, & T. W. Ling (Eds.),
Conceptual Modeling–ER 2004 (pp. 110-121). Shanghai, China: Springer.
Rosemann, M., Recker, J., Green, P., & Indulska, M. (2009). Using ontology for the
representational analysis of process modelling techniques. International Journal of
Business Process Integration and Management, 4(4), 251-265.
Rowe, F. (2014). What literature review is not: diversity, boundaries and recommendations.
European Journal of Information Systems, 23(3), 241-255.
Rumbaugh, J., Jacobson, I., & Booch, G. (2004). The Unified Modeling Language Reference
Manual Massachusetts: Addison-Wesley.
Saghafi, A., & Wand, Y. (2014). Do Ontological Guidelines Improve Understandability of
Conceptual Models? A Meta-analysis of Empirical Work. Paper presented at the 47th
Hawaii International Conference on System Sciences (HICSS), 2014 Hawaii.
Samuel, B. M., Watkins III, L. A., Ehle, A., & Khatri, V. (2015). Customizing the
Representation Capabilities of Process Models: Understanding the Effects of
Perceived Modeling Impediments. IEEE Transactions on Software Engineering,
41(1), 19-39.
Scheer, A.-W., & Nüttgens, M. (2000). ARIS Architecture and Reference Models for
Business Process Management. In W. M. P. van der Aalst, J. Desel, & A. Oberweis
(Eds.), Business Process Management: Models, Techniques, and Empirical Studies
(pp. 376-389). Berlin, Heidelberg: Springer
Sein, M. K., Henfridsson, O., Purao, S., Rossi, M., & Lindgren, R. (2011). Action Design
Research. MIS Quarterly, 35(1), 37-56.
Shadish, W. R., Cook, T. D., & Campbell, D. T. (2002). Experimental and quasi-
experimental designs for generalized causal inference. Boston: Houghton Mifflin.
Shanks, G. (1997). Conceptual Data Modelling: an empirical study of expert and novice data
modellers. Australasian Journal of Information Systems, 4(2), 63-73.
Shanks, G. (2002). Guidelines for conducting positivist case study research in information
systems. Australasian Journal of Information Systems, 10(1), 76-85.
Shanks, G., Moody, D., Nuredini, J., Tobin, D., & Weber, R. (2010). Representing classes of
things and properties in general in conceptual modelling: An empirical evaluation.
Journal of Database Management, 21(2), 1-25.
172 Bibliography
Shanks, G., Tansley, E., Nuredini, J., Tobin, D., & Weber, R. (2008). Representing part-
whole relations in conceptual modeling: an empirical evaluation. MIS Quarterly,
32(3), 553-573.
Shanks, G., Tansley, E., & Weber, R. (2004). Representing composites in conceptual
modeling. Communications of the ACM, 47(7), 77-80.
Shanks, G., & Weber, R. (2012). THE HOLE IN THE WHOLE: AResponse TO ALLEN
AND MARCH. MIS Quarterly, 36(3), 965-980.
Shiu, H., & Fong, J. (2008). Reverse engineering from an XML document into an extended
DTD graph. Journal of Database Management, 19(4), 62-80.
Siau, K. L. (2004). Informational and computational equivalence in comparing information
modeling methods. Journal of Database Management, 15(1), 73-86.
Siau, K. L. (2010). An Analysis of Unified Modeling Language (UML) Graphical Constructs
Based on BWW Ontology. Journal of Database Management, 21(1), i-viii.
Siau, K. L., Erickson, J., & Lee, L. Y. (2005). Theoretical vs. practical complexity: The case
of UML. Journal of Database Management, 16(3), 40-57.
Siau, K. L., & Lee, L. Y. (2004). Are use case and class diagrams complementary in
requirements analysis? An experimental study on use case and class diagrams in
UML. Requirements Engineering, 9(4), 229-237.
Siau, K. L., & Rossi, M. (2011). Evaluation techniques for systems analysis and design
modelling methods–a review and comparative analysis. Information Systems Journal,
21(3), 249-268.
Siau, K. L., & Tan, X. (2005). Improving the quality of conceptual modeling using cognitive
mapping techniques. Data & Knowledge Engineering, 55(3), 343-365.
Silverman, D. (2006). Interpreting qualitative data: Methods for analyzing talk, text and
interaction (3rd ed.). Thousand Oaks, London: Sage Publications.
Soffer, P. (2005). Refinement equivalence in model-based reuse: Overcoming differences in
abstraction level. Journal of Database Management, 16(3), 21-39.
Soffer, P., & Hadar, I. (2007). Applying ontology-based rules to conceptual modeling: a
reflection on modeling decision making. European Journal of Information Systems,
16(5), 599-611.
Soffer, P., & Kaner, M. (2011). Complementing business process verification by validity
analysis: A theoretical and empirical evaluation. Journal of Database Management,
22(3), 1-23.
Soffer, P., Kaner, M., & Wand, Y. (2010). Assigning Ontological Meaning to Workflow Nets.
Journal of Database Management, 21(3), 1-35.
Soffer, P., Wand, Y., & Kaner, M. (2015). Conceptualizing Routing Decisions in Business
Processes: Theoretical Analysis and Empirical Testing. Journal of the Association for
Information Systems, 16(5), 345-393.
Sowa, J. F. (2000). Knowledge representation: logical, philosophical, and computational
foundations. London: Brooks/Cole, Pacific Grove.
Storey, V. C., Goldstein, R. C., & Ding, J. (2002). Common sense reasoning in automated
database design: An empirical test. Journal of Database Management, 13(1), 3-14.
Storey, V. C., & Song, I.-Y. (2017). Big data technologies and management: What conceptual
modeling can do. Data & Knowledge Engineering, 108, 50-67.
Storey, V. C., Trujillo, J. C., & Liddle, S. W. (2015). Research on conceptual modeling:
Themes, topics, and introduction to the special issue. Data & Knowledge Engineering,
98, 1-7.
Bibliography 173
Straub, D., Boudreau, M.-C., & Gefen, D. (2004). Validation guidelines for IS positivist
research. Communications of the Association for Information Systems, 13(1), 380-427.
Sun, S. X., Zhao, J. L., Nunamaker, J. F., & Sheng, O. R. L. (2006). Formulating the data-
flow perspective for business process management. Information Systems Research,
17(4), 374-391.
Sweller, J. (1994). Cognitive load theory, learning difficulty, and instructional design.
Learning and Instruction, 4(4), 295-312.
Sweller, J., & Chandler, P. (1994). Why some material is difficult to learn. Cognition and
Instruction, 12(3), 185-233.
Tabchnick, B. G., & Fidell, L. S. (2007). Using multivariate statistics (5th ed.). Boston:
Pearson Education.
Thompson, J. D. (1967). Organizations in action: Social science bases of administrative
theory. New York: McGraw-Hill.
Topi, H., & Ramesh, V. (2002). Human factors research on data modeling: a review of prior
research, an extended framework and future research directions. Journal of Database
Management, 13(2), 3-19.
Trujillo, J. C., Luján-Mora, S., & Song, I.-Y. (2004). Applying UML and XML for designing
and interchanging information for data warehouses and OLAP applications. Journal of
Database Management, 15(1), 41-72.
Urquhart, C. (2013). Grounded theory for qualitative research: A practical guide. Thousand
Oaks, California: Sage Publication.
van der Aalst, W. M. P. (2011). Process Mining Discovery, Conformance and Enhancement
of Business Processes. Heidelberg, Germany: Springer.
van der Aalst, W. M. P., & ter Hofstede, A. H. M. (2005). YAWL: yet another workflow
language. Information Systems, 30(4), 245-275.
VanderMeer, D., & Dutta, K. (2009). Applying learner-centered design principles to UML
sequence diagrams. Journal of Database Management, 20(1), 25-47.
Venkatesh, V., Brown, S. A., & Bala, H. (2013). Bridging the qualitative-quantitative divide:
Guidelines for conducting mixed methods research in information systems. MIS
Quarterly, 37(1), 21-54.
Venkatesh, V., Brown, S. A., & Sullivan, Y. W. (2016). Guidelines for conducting mixed-
methods research: An extension and illustration. Journal of the Association for
Information Systems, 17(7), 435-494.
Veres, C., & Mansson, G. (2005). Cognition and Modeling: Foundations for Research and
Practice. Journal of Information Technology Theory and Application, 7(1), 93-104.
Vergara, N. M., Linero, J. M. T., & Moreno, A. V. (2007). Model-driven component
adaptation in the context of Web Engineering. European Journal of Information
Systems, 16(4), 448-459.
Vessey, I., Ramesh, V., & Glass, R. L. (2002). Research in information systems: An empirical
study of diversity in the discipline and its journals. Journal of Management
Information Systems, 19(2), 129-174.
Vidyasankar, K., & Vossen, G. (2013). Multi-level modeling of web service compositions
with transactional properties. Journal of Database Management, 22(2), 1-31.
Wagelaar, D., & Van Der Straeten, R. (2007). Platform ontologies for the model-driven
architecture. European Journal of Information Systems, 16(4), 362-373.
Walsham, G. (2006). Doing interpretive research. European Journal of Information Systems,
15(3), 320-330.
174 Bibliography
Wand, Y., & Weber, R. (1989). An ontological evaluation of systems analysis and design
methods. In E. D. Falkenberg & P. Lindgren (Eds.), Information Systems Concepts:
An In-depth Analysis: proceedings of the IFIP TC 8/WG 8.1 Working Conference on
Information System Concepts: an in-depth analysis Namur, Belgium, 18-20 October,
1989 (pp. 79-107). Amsterdam, The Netherlands: Elsevier.
Wand, Y., & Weber, R. (1990a). Mario Bunge's Ontology as a formal foundation for
information systems concepts. In P. Weingartner & G. J. W. Dorn (Eds.), Studies on
Mario Bunge's Treatise (pp. 123-149). Atlanta: Rodopi.
Wand, Y., & Weber, R. (1990b). An ontological model of an information system. IEEE
Transactions on Software Engineering, 16(11), 1282-1292.
Wand, Y., & Weber, R. (1993). On the ontological expressiveness of information systems
analysis and design grammars. Information Systems Journal, 3(4), 217-237.
Wand, Y., & Weber, R. (1995). On the deep structure of information systems. Information
Systems Journal, 5(3), 203-223.
Wand, Y., & Weber, R. (2002). Research commentary: information systems and conceptual
modeling—a research agenda. Information Systems Research, 13(4), 363-376.
Wand, Y., & Weber, R. (2004). Reflection: Ontology in information systems. Journal of
Database Management, 15(2), I-VI.
Wand, Y., Woo, C., & Wand, O. (2008). Role and Request Based Conceptual Modeling – A
Methodology and a CASE Tool. In Q. Li, S. Spaccapietra, E. Yu, & A. Olivé (Eds.),
Conceptual Modeling - ER 2008: 27th International Conference on Conceptual
Modeling, Barcelona, Spain, October 20-24, 2008. Proceedings (pp. 540-541). Berlin,
Heidelberg: Springer Berlin Heidelberg.
Weber, R. (1996). Are attributes entities? A study of database designers' memory structures.
Information Systems Research, 7(2), 137-162.
Weber, R. (1997). Ontological foundations of information systems. Melbourne, Australia:
Coopers & Lybrand and the Accounting Association of Australia and New Zealand
Melbourne.
Weber, R. (2003). Conceptual modelling and ontology: Possibilities and pitfalls. Journal of
Database Management, 14(3), 1-20.
Weber, R., & Zhang, Y. (1996). An analytical evaluation of NIAM's grammar for conceptual
schema diagrams. Information Systems Journal, 6(2), 147-170.
Webster, J., & Watson, R. T. (2002). Analyzing the past to prepare for the future: Writing a
literature review. MIS Quarterly, 26(2), xiii-xxiii.
Whiteley, D. (2013). An Introduction to Information Systems London, UK: Palgrave
Macmillan.
Whittle, J., Hutchinson, J., & Rouncefield, M. (2014). The state of practice in model-driven
engineering. IEEE Software, 31(3), 79-85.
Wyssusek, B. (2006). On ontological foundations of conceptual modelling. Scandinavian
Journal of Information Systems, 18(1), 63-80.
Yin, R. K. (2009). Case study research: Design and methods (4th ed.). Thousand Okas,
California: Sage publications.
Zhang, H., Kishore, R., & Ramesh, R. (2007). Semantics of the MibML conceptual modeling
grammar: an ontological analysis using the Bunge-Wang-Weber framework. Journal
of Database Management, 18(1), 1-19.
Bibliography 175
Zhang, H., Liu, L., & Li, T. (2011). Designing IT systems according to environmental
settings: A strategic analysis framework. The Journal of Strategic Information
Systems, 20(1), 80-95.
Zhang, J. (1997). The Nature of External Representations in Problem Solving. Cognitive
Science, 21, 179-217.
zur Muehlen, M., & Indulska, M. (2010). Modeling languages for business processes and
business rules: A representational analysis. Information Systems, 35(4), 379-390.
zur Muehlen, M., & Recker, J. (2008). How much language is enough? Theoretical and
practical use of the business process modeling notation. In International Conference
on Advanced Information Systems Engineering (pp. 465-479). Berlin Heidelberg:
Springer.
zur Muehlen, M., Recker, J., & Indulska, M. (2007). Sometimes less is more: are process
modeling languages overly complex? Paper presented at the Eleventh International
IEEE EDOC Conference Workshop Annapolis, MD, USA.
176 Appendices
Appendices
Appendix A
Literature Review Process
Table A. 1. Search Results
Sea
rch
Ter
m:
Con
ceptu
al M
odel
ling
Sea
rch
Ter
m: “C
once
ptu
al M
odel
*”
Sea
rch
Ter
m:
Con
ceptu
al m
odel
ling g
ram
mar
Sea
rch
Ter
m: “o
nto
logy”
Sel
ecte
d P
aper
s
Dat
abas
e
Journals # # # # #
MIS Quarterly 451 100 18 84 6 AISeL
Information Systems Research 149 71 9 14 18 Informs
Journal of the association for information systems 316 71 23 72 18 AISeL
European journal of Information Systems 321 111 20 56 19 Springer
Journal of Management Information Systems 278 39 242 17 6 Jstor
Information Systems Journal 240 63 7 42 8 Wiley Online
Library
Journal of Information Technology 181 45 4 33 2 Springer
Journal of Strategic Information Systems 148 36 1 11 3 Informs
Journal of Database Management 122 74 16 61 53 Science
Direct
Appendices 177
Table A. 2 Summary of Coding Scheme
What is the objective of the paper?
Focus of Study
Which research opportunity/ies the study intended to address?
Grammar Which research opportunity
related to conceptual modelling
grammar addressed by the
study?
Script Which research opportunity
related to conceptual modelling
script addressed by the study?
Method
Which research opportunity
related to conceptual modelling
method addressed by the study?
Context
Which research opportunity
related to conceptual modelling
context addressed by the study?
Type of Grammar What type of conceptual modelling grammar studied in the paper?
Example: BPMN, UML, ER
Type of Study
Empirical Study Which research method or approach has been
used?
Case study
Action Research
Survey
Experiment
Interview
Panel
Delphi Study
Design Science
Non-empirical study Which research method or approach has
been used?
Literature Review
Commentary
Theoretical
Empirical Evidence
if the paper is an empirical study
What are the quality and quantity of the empirical evidence
(considering the type of sample)?
What are the measurement items and materials?
(e.g. material: questionnaire, interview protocol, case description,
interface, developed diagrams, etc.)
(e.g., measurement items: accuracy, time, and perceived ease of
use)
and availability of measurement items and materials
178 Appendices
Table A. 3. Papers Addressed each opportunities proposed by Wand and Weber (2002)
Res
earc
h F
ram
ework
Ele
men
t
Research Opportunities Reference
Co
nce
ptu
al-M
od
elli
ng
Gra
mm
ars
G1
Evaluating ontologies based on
empirical testing of their
predictions
G2
Evaluating grammars for
ontological expressiveness
(Irwin & Turk, 2005); (Recker et al., 2009);
(Recker et al., 2010); (Green & Rosemann, 2004);
(Milton & Kazmierczak, 2004); (Hong Zhang et al.,
2007); (Dussart, Aubert, & Patry, 2004);
G3
Assigning ontological meaning to
constructs of design grammars
and generating ontologically
motivated modelling rules
(Allen & March, 2006); (Rittgen, 2006); (Bera et
al., 2010); (Evermann & Wand, 2009); (Laurier &
Poels, 2012); (Opdahl & Henderson-Sellers, 2004);
(Soffer & Kaner, 2011); (Soffer et al., 2010);
G4
Resolving outstanding ontological problems that impact conceptual
modelling—e.g., nature of the
part-of relationship
(Shanks & Weber, 2012); (Shanks et al., 2008)
G5
Empirically testing predicted
strengths and weaknesses in new
and existing grammars based on
their ontological expressiveness
G6
Determining which combinations
of grammars best support users who undertake conceptual-
modelling work
G7
Empirically testing the predicted
implications of construct deficit
and overload in grammars
(Allen & March, 2012); (Recker et al., 2011);
(Shanks et al., 2008); (Bera et al., 2014); (Bowen et
al., 2009); (Bowen et al., 2006); (Milton et al.,
2012); (Parsons, 2011); (Recker et al., 2010);
(Shanks et al., 2010)
G.O
ther
(Clarke et al., 2016); (Burton-Jones et al., 2009);
(Figl, Mendling, et al., 2013); (Recker, 2010a);
(Dobing & Parsons, 2008); (VanderMeer & Dutta,
2009); (L. Liu, Pu, & Ruiz, 2004); (Dussart et al.,
2004)
Conce
ptu
al-
Model
ling
Met
hods
M1
Evaluating how well different
methods allow users to elicit and model critical domain knowledge
(Burton-Jones & Meso, 2006);(Purao et al., 2003);
(Dunn et al., 2011); (Soffer et al., 2015); (Recker, 2013); (Poels et al., 2011); (Siau & Rossi, 2011);
(Poels, 2011); (de Sousa Saraiva & da Silva, 2010)
Appendices 179
M2
Developing procedures to assist
users of a grammar in identifying
and classifying phenomena
according to the grammar's
constructs
(Parsons & Wand, 2008); (Currim & Ram,
2012);(Hadar & Soffer, 2006);(Parsons & Wand,
2013); (Ågerfalk & Eriksson, 2004); (Dietz &
Juhrisch, 2012); (Soffer & Hadar, 2007); (Clegg &
Shaw, 2008); (Evermann, 2005); (Athenikos &
Song, 2013); (Batra, 2012); (Bera et al., 2010);
(Chua, Chiang, & Lim, 2002); (Dominguez, Rubio,
& Zapata, 2002); (Loucopoulos & Kadir,
2008);(Ma, 2005); (Opdahl & Henderson-Sellers,
2004); (Reinhartz-Berger & Sturm, 2008); (Batra,
2012)
M3
Determining the beliefs and
values that underlie different methods and evaluating the
consequences of these beliefs and
values for practice
(Hadar & Soffer, 2006)
M.O
ther
(Clarke et al., 2016);(Lukyanenko & Parsons,
2014); (Rittgen, 2006);(Allen & March, 2003); (An
et al., 2010); (Batra, 2008); (Pardillo et al., 2011);
(Storey et al., 2002); (H. M. Kim, Fox, & Sengupta,
2007); (Joosten & Purao, 2002)
Co
nce
ptu
al-m
odel
lin
g S
crip
ts
S1
Evaluating competing scripts
generated via the same grammar
to describe some phenomenon
(Allen & March, 2012); (Allen & March, 2006);
(Parsons & Wand, 2008); (Shanks et al., 2008);
(Bera et al., 2014); (Bowen et al., 2009); (Burton-
Jones & Meso, 2006); (Bowen et al., 2006);
(Burton-Jones & Meso, 2008); (Dunn et al., 2011); (Milton et al., 2012); (Parsons, 2011); (Recker,
2013); (Parsons, 2002); (Poels et al., 2011); (Bera,
2012); (Masri et al., 2008); (Poels et al., 2011);
(Reinhartz-Berger & Sturm, 2008); (Shanks et al.,
2010); (Soffer & Kaner, 2011);
S2
Evaluating competing scripts
generated via different grammars
to describe the same phenomenon
(Khatri et al., 2006); (Figl, Mendling, et al., 2013);
(Chan, Tan, & Teo, 2014)
S3
Evaluating different combinations
of scripts to determine which
combination best supports the
task at hand
S4
Developing theory to predict and
understand how humans use
scripts to accomplish various tasks
(Burton-Jones & Meso, 2008); (Parsons & Wand,
2013); (Krogstie et al., 2006); (Bera, 2012);(Bera et
al., 2010); (Siau, 2004)
S.O
ther
(Sun, Zhao, Nunamaker, & Sheng, 2006); (Burton-
Jones & Meso, 2008); (Bandara et al.,
2005);(Gemino & Parker, 2009); (Evermann,
2008);(Genero, Fernández-Saez, Nelson, Poels, &
Piattini, 2011); (Halpin, 2002)
Conce
pt
ual
-
Model
li
ng
Conte
xt
Indiv
idu
al
dif
fere
n
ces
Tas
k
Soci
al
agen
da
CI1
Development of knowledge-based
tools to support conceptual
modelling
(Purao et al., 2003);(Koschmider et al., 2010);
(Loucopoulos & Kadir, 2008); (Pardillo et al.,
2011); (Storey et al., 2002);
180 Appendices
CI2
Predicting which cognitive and
personality variables bear on a
user's ability to undertake
conceptual- modelling work
(Allen & March, 2006); (Bera et al., 2014); (Khatri
et al., 2006); (Davern et al., 2012b); (Browne &
Parsons, 2012); (Figl, Mendling, et al., 2013);
(Recker, 2010a); (Recker et al., 2010); (Soffer &
Hadar, 2007); (Evermann, 2005); (Green &
Rosemann, 2004); (Masri et al., 2008); (Shanks et
al., 2010); (VanderMeer & Dutta, 2009);
CI3
Predicting and testing empirically
which social skills affect the
outcomes of conceptual
modelling tasks
(Topi & Ramesh, 2002)
CI.
Oth
er (Bandara et al., 2005)
CT
Evaluating the strengths and
weaknesses of conceptual
modelling grammars, methods,
and scripts in the context of
different tasks
(Khatri et al., 2006); (Recker, 2010a); (Recker et
al., 2010); (Green & Rosemann, 2004);
CT
.Oth
er (Bandara et al., 2005); (Recker et al., 2010); (Soffer
& Hadar, 2007); (Topi & Ramesh, 2002);
(VanderMeer & Dutta, 2009)
CS
1 Understanding which values and
beliefs underlie conceptual-
modelling work in practice
(Hadar & Soffer, 2006); (Recker et al., 2010);
(Larsen et al., 2009);
CS
2
Determining the costs and
benefits of adopting different
values and beliefs when
undertaking conceptual-modelling
work
CS
3
Articulating detailed conceptual-modelling procedures that are
congruent with different beliefs
and values
CS
4
Understanding how existing
conceptual modelling grammars
and methods facilitate conceptual-
modelling work under different
values and beliefs
CS
.Oth
er (Koschmider et al., 2010); (He Zhang et al., 2011)
Appendices 181
Oth
ers
(Fonseca & Martin, 2007); (Chua et al., 2012);
(Garrigós et al., 2012); (Trujillo et al., 2004); (de
Albuquerque & Christ, 2015); (Fernández-Medina
et al., 2007); (D'aubeterre et al., 2008); (Dreiling et
al., 2006); (Montero, Díaz, & Aedo, 2007);
(Vergara et al., 2007); (Wagelaar & Van Der
Straeten, 2007); (H. M. Kim et al., 2007);
(Henderson-Sellers, 2002); (Soffer, 2005); ;
(Vidyasankar & Vossen, 2013);(Dahanayake, Sol,
& Stojanovic, 2003); (Dussart et al., 2004); (Shiu &
Fong, 2008); (Wand & Weber, 2004); (Mitra,
Bagchi, & Bandyopadhyay, 2007); (Weber, 2003)
182 Appendices
Appendix B
Appendix: Semi-structured Interview Protocol
Commencing Interviews
<State appreciation for time taken to participate> <Introduce myself>
<Explain the purposes of interview>
<Check if Consent to audio record, before recording>
<Begin>
Explanations of terms:
1. Conceptual models: all kinds of graphical representation of the real world created using modeling
methods such as ER diagrams, Class diagrams, business process models, activity diagrams or others.
2. Domain: A set of things in the real-world that a model represents, for example, the process of reserving
a flight ticket, or the accounts department in an organization.
3. System: An organized, purposeful structure that consists of a set of procedures created to carry out
specific functions and activities, for example online booking of flight ticket. 4. Interpretation: reading models by humans to understand its content, e.g., how different components of
the domain/system work or influence one another such as processes, data, associations, relationships.
5. Combined use: using the models that you have access to them at the same time or together, e.g.
simultaneously or in parallel.
Part A: Demographic information (background and experience)
A1: what is your current role in “organization”? Can you describe your responsibilities and tasks in general?
___________________________________________________________________________
A2: Over your working life how many years have you been working in business?
___________________________________________________________________________ A3: Over your working life how many years have you been working in IT-related business?
___________________________________________________________________________
A4: Over your working life how many years have you been working in system analysis and design?
___________________________________________________________________________
A5: Over your working life how many year you have been involved in conceptual modelling?
___________________________________________________________________________
A6: How long have you used conceptual models?
If you never used (Exit interview)
___________________________________________________________________________
A7: The training you have received in conceptual modelling includes:
(Specify level/ type / duration/ other details) (When undertook) 1. Formal training courses
2. Internal training courses
3. University courses
4. On the Job / mentor
5. Learnt by myself (how?)
6. None.
How about specific model you are using? (UML)
__________________________________________________________________________________________
____________________________________________________________
A8: In general, your primary purpose of using conceptual models is:
1. To understand system processes 2. To understand requirements of the system/project
3. Detect possible mistakes/problems in system development processes
4. System planning/design
Appendices 183
5. Communication
6. Other, please specify
__________________________________________________________________________________________
______________________________________________________
A9: Generally, the organizational level of models you use is:
1. High level (general level model, less detailed) (GIVE EXAMPLES)
2. Medium level (medium detailed models)
3. Low level (very detailed models)
A10: In particular, do you use conceptual models for the following purposes? (Multiple answers possible) Please score the purpose of use of conceptual models
Purpose
Score
Never Rarely Sometimes Often Always
1 Database design and management
2 Business process documentation
3 Improvement of internal business processes
4 Software development
5 Improvement of collaborative business processes
6 Workflow management
7 Design of enterprise architecture
8 Change management
9 Knowledge management
10 End user training
11 Software configuration
12 Software selection
13 Certification/quality management
14 Human resource management
15 Activity-based costing
16 Auditing
17 Simulation
18 Schedule and manage development process
Other, please specify and score:
----------------------------------------------------------------------------------------------------------------
A11: How many members are in the team you are working with?
___________________________________________________________________________
A12: How do the members cooperate in the project that you are involved? (How models involve in this
cooperation?)
Everyone does specific part of the project
Each part of the project is done separately and finally combined together
The entire project is done collectively
A13: Describe the project in which you are using conceptual models
__________________________________________________________________________________________
__________________________________________________________________________________________
___________________________
Part B: The use of combinations of conceptual models
B1: How are you involved in conceptual modelling?
1. You just design them (Exit Interview)
2. You just use them 3. Both
184 Appendices
B2: What is your role in the project you are involved in?
__________________________________________________________________________________
B3: Is there any requirement in your organization to use models; for example, the organization requires you to
use specific types of models?
Yes, please specify
No, why not?
________________________________________________________________________
B4: Do you need multiple types of models?
Yes, how many?
No Go to Part D ________________________________________________________________________
B4: Are different types of conceptual models available for you to use in the project that you are involved in?
Yes, which ones?
No Go to Part D
B5: Do you use different types of models?
Yes, which ones?
No Go to Part D
B6: Why do you use more than one model? 1. There is no one model that presents all concepts need for complete representation of the domain
2. That combination of model is the standard set to be used within my organization
3. Other, please specify.
__________________________________________________________________________________
B7: What types of models do you use? (e.g. class diagrams)
__________________________________________________________________________________
B8: What leads you to use these types of models?
__________________________________________________________________________________
B9: What do these models represent?
__________________________________________________________________________________
B10: As a user, do the models completely specify the requirements and convey all relevant information about the domain?
__________________________________________________________________________________
B11: How do you deal with models that are difficult to understand?
__________________________________________________________________________________
B12: How do you deal with models that are incomplete?
__________________________________________________________________________________
B13: How do you deal with models that are not accurate?
__________________________________________________________________________________
B14: How do you get more information if you need it?
1. I look at additional models (process to next question)
2. I look at documents 3. Other (explain)
__________________________________________________________________________________
B15: How do you choose additional models? (What circumstances lead you to choose and what factors in the
models are important for decision)
__________________________________________________________________________________
Appendices 185
Part C: Familiarity with Conceptual Models
Which conceptual models are you familiar with? What is your level of familiarity (1 to 5)? (Multiple answers
possible), How frequently do you use models (1 to 7)? Number of models you have seen (#)?
Familiarity Frequency of Use
Number of Models
You have Seen
Models
Very Low
Intermediate
Mediate
Advance
Professional
Never
Rarely
Occasionally
Sometimes
Frequently
Usually
Every time
Numbers
e.g.: 5, 20, 100 ….
Class diagrams
Use case diagrams
Sequence diagrams
Activity Diagrams
State chart diagrams
Object diagrams
Component diagrams
Deployment diagrams
Package diagrams
Composite structure diagram
Profile diagram
Communication diagram
Interaction overview diagram
Timing diagram
ER Diagrams
Business Process Models (BPMN)
Event-driven Process Chain (EPC)
Data Flow Diagrams
Petri Nets
Workflow
Gantt charts
Flowchart
Application diagrams
Object-Role Models
Role activity diagrams
Role interaction diagrams
Integrated definition modelling
(IDEF)
Other, Please Specify:
C2. Is there anything else you want to tell me about how you use conceptual models for your work?
________________________________________________________________________
C3. What do you think are benefits from using conceptual models?
________________________________________________________________________
C4. What are the biggest challenges in using conceptual models?
________________________________________________________________________
Part D: Using a Single Conceptual Model
186 Appendices
D1: which model (type) are you using?
________________________________________________________________________
D2: Why do you use a single conceptual model?
I use only one model because:
A model completely specifies the requirements and conveys all relevant information we need.
That is standard procedure set by my organization.
Other, please specify
__________________________________________________________________________________
D2: What leads you to choose this specific model? Why you did not use another type of model?
__________________________________________________________________________________
D3: What does the single model represent? __________________________________________________________________________________
D4: If you need more information, do you want to use additional models?
No (Explain how you overcome with that conflict?)
Yes (Explain why do you choose additional model?)
__________________________________________________________________________________
D5: How do you choose an additional model? What factors are important for you to choose additional model?
__________________________________________________________________________________
D6: Is there anything else you want to tell me about how you use conceptual models for your work?
________________________________________________________________________
D7: What do you think are benefits from using conceptual models?
________________________________________________________________________ D8: What are the biggest challenges in using conceptual models?
________________________________________________________________________
<Ask interviewee to suggest other respondents if possible>
<Provide opportunity for participant to comment on anything further>
<Thank the participant for their time and valuable responses>
<Note if participant would like a copy of results and obtain contact details (Email or post)>
End of Interview
Appendices 187
Appendix C
Table C. 1 Ontological Analysis of Graphical Constructs of Selected UML-Diagrams
UML-
Constructs
/elements
Shapes BWW-
constructs Rational
Cla
ss D
iagra
m
Class
User
Student ID*UserNameUserInfo
Class
Class type shows types of
things which share similar
properties in general
Member
ClassName
-className-classID
Property
Specifies attributes of a class “Attributes are the names that
we use to represent properties
of things” (e.g., ID, Name)
(Recker et al., 2009)
Relationship
Type
Bill
billNumberStudentID*billDateFineAmount
Coupling and
binding mutual
properties
Relationship types describe
the binding mutual properties
that couple two classes of
things.
Operations
Bill
-billNumber
-StudentID*
-billDate
-fineAmount
+createBill()
+updateBill()
Transformation
An operation is a behavioural
feature that may be owned by an interface. UML-operations
corresponds to BWW-
transformation (Opdahl &
Henderson-Sellers, 2002)
Cardinality 1..1
1..8
State law
Cardinality constraints
represent a state law that
constrains the values of a
binding mutual property to
certain conditions.
Inheritance
Association:
represents
generalization
Hereditary
property
Representing the “kind of”
relationship between classes
(things) which indicate that
the properties of the source
class inherit from the target class.
Aggregation
association:
represents the
part-of
relationship
Composite
thing
Special kind of association
which represents a structural
relationship between a whole
and its parts. It associates a
thing to a composite thing.
(Shanks et al., 2008)
188 Appendices
Composition
association:
represents part-
of relationship
Composite
thing
A composition is a form of
aggregation, with strong
ownership and coincident
lifetime as part of the whole.
Composition association
represents that the source
class is a component of the
target class (composite thing)
(Shanks et al., 2008).
Use
-Cas
e
Use Case
Transformation
Use cases describe sets of
actions as mappings that will
change the state of the
system.
Actor
Class
Actors are roles, which
describe specific types of
things (such as humans). See
Irwin and Turk (2005, p. 13).
Association
Binding
mutual
property
Associations draw linkages
between actors and use cases
such as which role are
authoritative for carrying out
an action.
Generalization
Excess
Generalization between use
cases does not carry an ontological meaning because
they violate the “kind of”
relationship that can exist
between things (but not
between processes changes of
states. See (Irwin & Turk,
2005, p. 13)
Act
ivit
y
Activity Validate User
Transformation
Describes a change in a state
of a thing.
(Recker et al., 2009)
Control flow
Lawful
transformation
Describes the sequence of
actions and the process. A
control flow shows the flow
of control from one action to the next. (Opdahl &
Henderson-Sellers, 2002). A
transition is a relationship
between two activities
indicating the flow of the
activities.
(Booch et al., 2005)
Fork node, join
node Excess
Synchronization bar
represents splitting and
linking parallel flows of
control
Initial/final node
Event
Represent the end event and
start event in process models. (Recker et al., 2009).
The initial state and end state
represent the start and end of
Appendices 189
the control flow (Booch et
al., 2005).
Merge node
Excess
Merge node is a control node
that brings together multiple
incoming alternate flows to
accept single outgoing flow.
Decision node
Excess
The control flows coming
away from a decision node
will have guard conditions
which will allow control to
flow if the guard condition is
met. It does not carry an
ontological meaning.
Guard condition Fine=0 Fine>0
Stability
condition
Specifies the states that are
allowable under the
transformation law
(Recker et al., 2009)
Swimlane
Library Management System
Library User
Phas
e
System and Class
Partitions the activity states
on an activity diagram into
groups, each group
representing the business organization responsible for
those activities.
(Opdahl & Henderson-
Sellers, 2002, p. 51; Recker
et al., 2009)
Object b: Bill
[paid]
State
Objects in this particular
diagram show that how
activities may modify
properties of the Bill (for
example, Pay Bill will change the state of the Bill
object to paid). (Booch et al.,
2005)
Object Flow
State law
Represent the flow an object
and specifies how the state of
an object should change.
Sta
te
State Item Found
State
Represents the value of the
property function of a thing
(Opdahl & Henderson-
Sellers, 2002, p. 48; Recker
et al., 2009)
Initial/final state
Stable State
Initial State is the default
starting state for the state
machine. The final state is a
special kind of state signifying that the enclosing
region is completed.
190 Appendices
Choice
Excess
Choice pseudostate realizes a
dynamic conditional branch.
It shows that the transitions
should follow the guard
conditions. It does not carry
an ontological meaning.
Guard
Condition
[Fine=>0]/Issue Bill
[Fine=0]/Update
Status
Stability
condition
Guard conditions describe
properties that restrict the
functions of a mutual
property between things to a
lawful subset.
Transition Calculate Fine
Transformation
A transition shows the flow
of control from one state to
the next. A transition is a relationship between two
states indicating that an
object in the first state will
perform certain actions and
enter the second state when a
specified event occurs and
specified conditions are
satisfied. (Booch et al., 2005)
Staff
maintain bike register
maintain rental rates
Customer
fail to return
rent out
rent return
register new bike
record bike sale
Figure C. 1 Use Case Diagram for High Peak Bicycles
Appendices 191
Bike Class
-classCode
Bicycle
-bikeRegistrationNumber
-classCode*
-bikeMake
-bikeDateOfSale
-bikeModel
-bikeNumber
-bikeDateOfPurchase
-bikePurchasePrice
-bikeSize
-bikeSellingPrice
-bikeWrite-Off
Rental Rate
-ClassCode*
-rateStartDate
-rateDay
-rateHalf-Day
-rateLateDay
Rent Line
-rentNumber*
-bikeRegistrationNumber*
-rentPre-PaidAmount
-rentActualPayment
Customer
-customerNumber
-customerName
-customerTelephoneNumber
Rental Agreement
-rentNumber
-customerNumber*
-rentDate
-rentTimeOut
-rentExpectedTimeBack
-rentExpectedPeriod
-rentActualTimeBack
-rentActualPeriod
-rentDepositType
-rentTotalPre-Paid
-rentActualTotalPaid
-classDescription
1..1
1..*
1..1
0..*
1..1 0..*
1..4
1..1
1..1
1..*
+haveClass()
+createClass+createRate()
+readRate()+haveCustomer()
+createCustomer()
+findBike()
+changeAvailability()
+createBike()
+amendBike()
+createRentLine()
+createRent()
+returnRent()
Figure C. 2 Class Diagram for High Peak Bicycles
Idle
[Request== return]
Return Request Submitted
Rent Request Submitted
Calculate rental
Actual Rental calculated
Bicycle Status Updated
Disp
lay re
ntal
rate
Bicycle Found
Rate Accepted
Rental Agrreement Issued
Update bicycle availability
Bicycle Status Updated
Checkavailability
Assign rental agreement
[Acc
epte
d]
[Not accepted]
Request Declined
[Available]
[Req
uest
==re
nt]
[Not
avail
able
]
[Another request]
[No other request]
Pay rental
Rental Paid
Update bicycle availability
Rent Rate Displayed
Bicycle Not Found
[Another request]
[No other request]
Figure C. 3 State Machine Diagram for High Peak Bicycles
192 Appendices
High Peak Bicycles
Customer Rentline
Submit a Request Check the Request
[Ren
t B
ike]
Check Rental Type
[Ret
urn
Bik
e]
Check Rental Agreement
Calculate Actual Rental
Period
Charge the Customer
Update Rental Agreement
Return Deposit
Change Bike Status
Read Bicycle Class
Read Rent Rate
Read Bike
Inform Customer
Display the rent rate
[Available][Not available]
Check the customer
Create Customer
Create Rental Line
Change Bike/s Status
Print Rental Agreement
[New Customer]
[Existing Customer]
Choose Other Bike/s
[Choose Other Bike]
[No
othe
r ch
oice
]
[Rent RateNot Accepted]
[Rent Rate Accepted]
Sign Rental Agreement
Check Bike Availability
Pay Deposit
Make Decision
Check Rent Rate
Read Customer
Figure C. 4 Activity Diagram for High Peak Bicycles
Appendices 193
Table C. 2 Ontological Analysis of Models for High Peak Bicycles
Ontological construct Class Use
Case Activity
State
Machines
Thing Class * * * Composition Thing Property in general * Hereditary Property Binding Mutual property * * State * State Law * Stable State * History Event * External Event Transformation * * * * Lawful Transformation * Stability Condition * * Coupling *
Acts on
System * Sum out of 17 6 3 6 4
Table C. 3 Overlap Analysis of Combinations of Two Models
Diagram Type Use Case
Diagram
Class
Diagram
Activity
Diagram
State Machine
Diagram
Use Case Diagram 3 2 1
Class Diagram 6 2 1
Activity Diagram 7 10 2
State Machine Diagram 6 9 8
Table C. 4 Overlap analysis of combinations of 3 and 4 models
Use Case
Diagram
Use Case
Diagram
Use Case
Diagram Class Diagram
Use Case
Diagram
Class Diagram Class Diagram Activity
Diagram
Activity
Diagram Class Diagram
Activity
Diagram
State Machine
Diagram
State Machine
Diagram
State Machine
Diagram
Activity
Diagram
State Machine
Diagram
MOC 10 9 9 12 12
MOO 3 3 3 3 4
194 Appendices
Figure C. 5 Use Case Diagram for Library Management System
Appendices 195
User
-Student ID*
-userName
-userInfo
Undergraduate
student
-Student ID
-Name
-DoB
-Major
-Faculty
-Address
Postgraduate
student
-Student ID
-Name
-DoB
-Major
-Faculty
-Address
-Thesis Topic
-Thesis supervisor
Transaction
-transactionID
Reservation
-dateReserve
-dateAvailable
-studentID*
-bookID*
Copy
-bookID*
-copyNumber
-copyAvailability
1..1
1..8
Book
-ISBN
-bookTitle
-bookAuthor
-Year
-Edition
-Topic
-Category
-Keywords
-Description
0..*
1..1
-bookID
-bookReservation
Loan
-copyNumber*
-dateOut
-dateBack
0..*
1..1
1..1
1..*
Author
-biography
-birthdate
-name
Research
Journals
-journalTitle
-publisher
-volume
-issue
0..1
0..*-author
1..*
1..*
1..*
1..*
Collection
-Name of collection
-relevant student major(s)
-number of books in
collection
Bill
-billNumber
-StudentID*
-billDate
-fineAmount0..4
1..1
+createBill()
+updateBill()
+createUser()
+checkUser()
+cancelUser()
+payFine()
+returnBook()
+createReservation()
+getReservation()
+checkAvailableDate()
+createLoan()
+getLoan()
+completeDateBack()
+createJournal()
+createAuthor()
+createCopy()
+addBook()
+searchBook()
+displayBookDetails
+updateBookDetails()
+createCollection()
+createTransaction()
+getTransaction()
0..1
0..1
Figure C. 6 Class Diagram for Library Management System
196 Appendices
Library Management System
Library UserP
ha
se
Request an ItemCheck the Request
Calculate Fine
Check Availability
Validate User
Re
turn
Bo
ok
Request Book
Re
qu
est
Jo
urn
al A
rtic
leN
ot
Av
aila
ble
Inform User
Available or Hired
Ava
ila
ble
Request Loan H
ire
d Request Reservation
Validate User
Validate User
Make Reservation
Issue Book
Change Book Status
No
t V
alid
or
Exc
ee
d lo
an
or
bil
l
No
t V
alid
or
Exc
ee
d lo
an
or
bil
l
Submit Request
Sto
p t
he
ord
er
Continue to Order
Authorise Journal Access
Po
st-g
rad
e S
tud
en
t
Not ValidShelve Book
Fin
e =
0
Fin
e >
0
Shelve Book Issue Bill
Change User Status
Check User Status
Check User Status
Valid & No exceed loan and bill
Valid & No exceed loan and bill
b: Bill
[unpaid]
Save as Debt
Pay the Bill
Change Book Status
b: Bill
[paid]
Change User Status
Change Book Status
Figure C. 7 Activity Diagram for Library Management System
Appendices 197
Figure C. 8 State Machine Diagram for Library Management System
Table C. 5 Combined Ontological Completeness and Ontological Overlap for Library Management System
Ontological construct Class Use
Case Activity
State
Machines
Thing
Class * * *
Composition Thing *
Property in general *
Hereditary Property *
Binding Mutual property * *
State * *
State Law * *
Stable State *
History
Event *
External Event
Transformation * * * *
Lawful Transformation *
Stability Condition * *
Coupling *
Acts on
System *
Sum out of 17 8 3 8 4
198 Appendices
Table C. 6. Overlap Analysis of Combinations of Two Models
Diagram Type Use Case
Diagram
Class
Diagram
Activity
Diagram
State Machine
Diagram
Use Case Diagram 3 2 1
Class Diagram 8 3 1
Activity Diagram 9 13 3
State Machine Diagram 6 11 9
Table C. 7. Analysis of MOC and MOO
Combination Completeness Overlap
High Com, Low Overlap Class + State 11 1
Low Com, High Overlap Class + Use Case 8 3
Low Comp, Low Overlap State + Use case 6 1
High Com, High Overlap Class + Activity 13 3
Table C. 8 Measurement Instruments for the Designed Experiment to test the Theory of Combined Ontological
Coverage
Demographic Questions
What is your gender?
Male
Female
Are you an under- or post-graduate student?
Under-graduate
Post-graduate
Have you ever learned UML (Unified Modeling Language)?
Yes
No
Are you enrolled in a unit that teaches conceptual modeling in semester 2/2016, or have you completed such a
unit in an earlier semester?
Yes
No
Modeling Experience (Adapted from (Burton-Jones & Meso, 2008; Recker, 2010a))
Roughly, how many UML diagrams have you:
……..Created to date? None
……..Read to date? None
Familiarity with UML (Adapted from Recker (2010a))
Please rate your agreement with the following statements about your familiarity with UML on a scale from 1 (strongly disagree) to 7 (strongly agree):
(1) Overall, I am very familiar with UML.
(2) I feel very confident in understanding diagrams created with UML.
Appendices 199
(3) I feel very competent in using UML diagrams.
(4) Compared to a professional system developer or analyst, I would rate my level of experience in interpreting UML diagrams as: (a scale from 1 (Very low) to 7 (Very high))
Domain Knowledge (Bera et al., 2014; Burton-Jones & Meso, 2008)
Please rate your level of knowledge and experience on a scale from 1 (Very low) to 7 (Very high):
(1) Compared to someone who works in a bicycle rental shop, I would rate my level of knowledge of
activities in a bicycle rental shop (such as updating rent rates, preparing rental agreements,
categorizing bicycle classes) as:
(2) Compared to someone who works in a library, I would rate my level of knowledge of activities in
a library (such as shelving books, updating book status, reserving books) as:
Model Familiarity Test [the correct answers are written in bold] (Figl & Recker, 2014)
What type of model is this?
o UML Class diagram
o UML Component Diagram
o UML Object Diagram
o UML State Machine Diagram
o UML Activity Diagram
o I don’t know
What type of model is this?
o UML Component Diagram
o UML Class diagram
o UML Sequence Diagram
o UML State Machine Diagram
o UML Use Case Diagram
o I don’t know
What type of model is this?
o UML Use Case Diagram
o UML Class diagram
o UML Component Diagram
o UML Sequence Diagram
o UML Activity Diagram
o I don’t know
What type of model is this?
o UML Class diagram
o UML Sequence Diagram
o UML State Machine Diagram
o UML Communication Diagram
o UML Use Case Diagram
o I don’t know
200 Appendices
What type of model is this?
o UML State Machine Diagram
o UML Activity Diagram
o UML Timing Diagram
o UML Communication Diagram
o UML Interaction Overview Diagram
o I don’t know
What type of model is this?
o UML Component Diagram o UML Profile Diagram
o UML Package Diagram
o UML Development Diagram
o UML Object Diagram
o I don’t know
Measurement Instruments for the first phase of the experiment (model selection)
Problem Solving Questions - Upcoming task for model selection (self-developed)
1. The managers of the Bicycle rental shop are interested in improving the overall revenue. Based on the
information in the diagrams, identify at least two ways how they can do so.
2. Customer feedback indicates that client satisfaction significantly decreased over the last six months.
From the information in the diagrams, identify at least two reasons why this could be the case.
Use of Information –learning effect (Adapted from Burton-Jones and Meso (2008))
On a 1-5 scale (“Not at all” to “A great extent”), to answer the previous two problem-solving questions…:
- To what extent did you rely on the information in diagram (A)?
- To what extent did you rely on the information in diagram (B)?
- To what extent did you rely on the information in diagram (C)?
- To what extent did you rely on the information in diagram (D)?
- To what extent did you rely on your general knowledge?
- To what extent did you make assumptions?
Deselection of models (multiple answers are possible) (Self developed)
If you had an opportunity to remove a diagram/s which, in hindsight, you feel was not useful in answering the
questions, which diagram/s would you have removed?
- Diagram (A)
- Diagram (B)
- Diagram (C)
- Diagram (D)
- None of them
Measurement Instruments for the second phase of the experiment
Appendices 201
Comprehension Questions (Self developed)
Please answer the following questions about the Library based on the information in the diagrams you received:
(the correct answers highlighted in bold)
- Are there any limitations for a student to extend a loan? (Yes/No/Not Sure/There is not enough
information to know)
- Can a post-graduate student borrow more than 6 research journals at the same time? (Yes/No/Not
Sure/There is not enough information to know)
- Can a user of a library be an employee of the university? (Yes/No/Not Sure/There is not enough
information to know)
- Do students need to pay a fee to access research journals? (Yes/No/Not Sure/There is not enough
information to know)
- Can a student request books that are loaned by someone else? (Yes/No/Not Sure/There is not enough
information to know)
- Are there a limited number of copies of each book in the library? (Yes/No/Not Sure/There is not
enough information to know)
- Can an undergraduate student borrow a book and a research journal at the same time? (Yes/No/Not
Sure/There is not enough information to know)
- Do students need to pay fines for a delay in returning books? (Yes/No/Not Sure/There is not enough
information to know)
Problem Solving Questions (Self developed)
Please answer the following questions, based on your understanding, from the diagrams.
- Explain the main functionality of the library management system. Write down as many supported
activities as you can.
- Under what circumstances would a student fail to borrow an item s/he needs? Write down as many
possible reasons as you can think of.
- How can the library management system be improved to ensure that students in each major can have access to relevant books and journals they need? Write down as many possible solutions that should be
considered.
Cloze Test (Self developed) (No. of correct answers)
Please fill in the blanks with the appropriate words based on the understanding you gained about the Library
Management System.
The library management system is designed to track books and journal articles in the library of a university. It
also provides functions to …(1)… and control orders and loans, and calculate fines for …(2)… . Users of
library management system are either …(3)… or …(4)… .
The library has books in a wide variety of areas and topics. Normally, there are multiple …(5)… of a particular
book in the library. The library also provides access to research journals.
…(6)… do not have access to research journals . But, books in the library are categorised in specific …(7)… to
be sure that the library provides relevant books for students in different majors.
Users can search items using title, authors name, topic and any keywords related to their majors. In fact, …(8)…
can search items in the library, but only …(9)… can borrow books or have access to journal articles.
The results of a search for a book can be threefold: … (10) …, …(11)…, or …(12)…. If a book is on loan by someone else, a user can still …(13)… the book. There are no limitations on how many …(14)… a …(15)…
student can have access at the same time.
202 Appendices
Answers to the Cloze test
1) Monitor/manage, 2) late returns/none returns, 3) undergraduate-students, 4) postgraduate-students, 5)
copies, 6) undergraduate-students, 7) collections, 8) everybody, 9) valid users, 10) available, 11)
unavailable, 12) On loan, 13) reserve/request, 14) research journals, 15) postgraduate
Perceived Usefulness of model combinations (Adopted from (Recker & Rosemann, 2010a, 2010b) on a 1-7
scale (“Strongly disagree” to “Strongly agree”)
Please answer the following questions based on your level of agreement using a 1-7 scale from “Strongly
disagree” to “Strongly agree”
- Overall I found the combination of diagrams useful for the given tasks.
- I found the combination of diagrams useful for achieving the purpose of the given tasks.
- The combination of diagrams helped me in meeting the objectives of the given tasks.
Cognitive Load (Adapted from Cierniak et al. (2009)) on a 1-7 scale (“Not at all” to Extremely”).
Please answer the following questions based on the rates from 1- (Not at all) to 7-(Extremely).
- How difficult were the tasks for you?
- How difficult was the content of the provided diagrams for you?
- How difficult was it for you to perform the tasks with the provided diagrams?
- How much did you concentrate during the tasks?
Ease of Understanding (Burton-Jones & Meso, 2008)
Please answer the following questions based on the rates from 1- (Not at all) to 7-(Extremely).
- To what extent did you find the diagram (A) [type of diagram will vary for each group] to be complex?
- To what extent did you find the diagram (A) to be difficult to understand?
- To what extent did you find the diagram (B) [type of diagram will vary for each group] to be complex?
- To what extent did you find the diagram (B) to be difficult to understand?
Appendices 203
Table C. 9 Interrater Agreement
Value
Asymp. Std.
Errora Approx. Tb
Approx.
Sig.
Measure of Agreement Kappa
Problem-Solving .811 .037 27.208 .000
Measure of Agreement Kappa
Cloze Test .907 .027 33.037 .000
N of Valid Cases 131 a. Not assuming the null hypothesis.
b. Using the asymptotic standard error assuming the null hypothesis.
Table C. 10. Factor Loadings
Component Matrixa
Component
1
Familiarity Self Rate 1 .836
Familiarity Self Rate 2 .791
Familiarity Self Rate 3 .827
Familiarity Self Rate 4 .642
Usefulness 1 .799
Usefulness 2 .898
Usefulness 3 .884
Cognitive Load 1 .852
Cognitive Load 2 .816
Cognitive Load 3 .845
Cognitive Load 4 .480
Table C. 11 Reliability Test
Reliability Statistics
Cronbach's Alpha
Cronbach's Alpha
Based on
Standardized Items N of Items
Modeling Familiarity .775 .778 4
Perceived Usefulness .822 .825 3
Cognitive Load .806 .815 3