20
Knowledge Management in Organizational Planning LYNDA M. APPLEGATE, TSUNG TENG CHEN, BENN R. KONSYNSKI, and JAY F. NUNAMAKER, Jr. LYNDA M . AppLEGATE.is an Assistant Professor at the Harvard Business School. She received her Ph.D. in Business Administration from the University of Arizona with a major in Management Information Systems and a minor in Management. Prior to joining the faculty at the Harvard Business School, Dr. Applegate held a number of administrative positions. In addition, she held previous faculty appoint- ments at the University of Arizona and the University of Washington. She has consulted with a number of companies on a variety of general management issues including the development of information systems product marketing strategies and the design and implementation of integrated information systems. Dr. Applegate's primary research interests and recent publications are concerned with die design, implementation, and management of information systems based on emerging tech- nologies (e.g., knowledge-based systems, group decision support systems, and model management systems) and the design and implementation of executive infor- mation systems. TsuNG TENG CHEN is a doctoral student in Management Information Systems at the University of Arizona. His current research interests include the knowledge base management system, knowledge management in integrated development enviroment for information system, and the application of artificial intelligence techniques to information system analysis and design. BENN R. KONSYNSKI is a Professor at the Harvard Business School on leave from the University of Arizona. He received his Ph.D. in Computer Science from Purdue University. Current research interests include: knowledge-based systems in business and research, model management, inter-organizational systems, strategic planning, group decision support, dialogue design, distributed systems, and the impact of information technologies on organization design. JAY F . NUNAMAKER, JR., is Head of the Department of Management Information Systems and is a Professor of Management Information System (MIS) and Computer Science at the University of Arizona. He received a Ph.D. from Case Institute of Technology in systems engineering and operations research. He was an This paper was presented al the TSwentieth Annual Hawaii International Conference on System Scieoces, Honolulu, Januaryfr-9,1987,and is published here with permission. The paper describes work performed at the University of Arizona. The research was in part by grants from IBM and NCR corporations. Jimmiloflimatimtiil b^ormakm SyaamlSftini I9>7. Vol. 3. No. 4

Knowledge Management

Embed Size (px)

Citation preview

Page 1: Knowledge Management

Knowledge Management inOrganizational Planning

LYNDA M. APPLEGATE, TSUNG TENG CHEN,BENN R. KONSYNSKI, and JAY F. NUNAMAKER, Jr.

LYNDA M . AppLEGATE.is an Assistant Professor at the Harvard Business School.She received her Ph.D. in Business Administration from the University of Arizonawith a major in Management Information Systems and a minor in Management.Prior to joining the faculty at the Harvard Business School, Dr. Applegate held anumber of administrative positions. In addition, she held previous faculty appoint-ments at the University of Arizona and the University of Washington. She hasconsulted with a number of companies on a variety of general management issuesincluding the development of information systems product marketing strategies andthe design and implementation of integrated information systems. Dr. Applegate'sprimary research interests and recent publications are concerned with die design,implementation, and management of information systems based on emerging tech-nologies (e.g., knowledge-based systems, group decision support systems, andmodel management systems) and the design and implementation of executive infor-mation systems.

TsuNG TENG CHEN is a doctoral student in Management Information Systems at theUniversity of Arizona. His current research interests include the knowledge basemanagement system, knowledge management in integrated development enviromentfor information system, and the application of artificial intelligence techniques toinformation system analysis and design.

BENN R. KONSYNSKI is a Professor at the Harvard Business School on leave from theUniversity of Arizona. He received his Ph.D. in Computer Science from PurdueUniversity. Current research interests include: knowledge-based systems in businessand research, model management, inter-organizational systems, strategic planning,group decision support, dialogue design, distributed systems, and the impact ofinformation technologies on organization design.

JAY F. NUNAMAKER, JR., is Head of the Department of Management InformationSystems and is a Professor of Management Information System (MIS) and ComputerScience at the University of Arizona. He received a Ph.D. from Case Institute ofTechnology in systems engineering and operations research. He was an

This paper was presented al the TSwentieth Annual Hawaii International Conference onSystem Scieoces, Honolulu, January fr-9,1987, and is published here with permission.

The paper describes work performed at the University of Arizona. The research wasin part by grants from IBM and NCR corporations.

Jimmiloflimatimtiil b^ormakm SyaamlSftini I9>7. Vol. 3. No. 4

Page 2: Knowledge Management

KNOWLEDGE MANAGEMENT IN ORGAKEZATiONAL PLANNING 2 1

Associate Professor of Computer Science and Industrial Administration at PurdueUniversity. Dr. Nunamaker joined the faculty of the University of Arizona in 1974 todevelop the MIS program. Author of more than 40 papers on group decision supportsystems, the automation of software construction, performance evaluation of com'puter systems, and decision support systems for systems analysis and design, he haslectured throughout Europe, Russia, Asia, and South America. Dr. Nunamaker isChairman of the Association for Computing Machinery (ACM) Curriculum Com-mittee on Information Systems.

ABSTRACT: There is growing recognition that the ability to provide automated sup-port for unstructured decision making within organizations will require the integra-tion of knowledge-based expert system techniques and traditional decision supportsystem architectures. Several knowledge representations are applicable in the speci-fication, management, and communication of knowledge associated with organiza-tional planning, a classic ill-structured problem facing organizations.

This paper describes the requirements for knowledge management in organiza-tional planning. A knowledge-based planning system that has been implemented bythe audiors is presented. The system integrates data management, model manage-ment, and process management systems within a group decision support systemenvironment. Knowledge management tools for describing, classifying, and storingthe output of the plaiming process are described. Use of the system in the Manage-ment Information Systems (MIS) Planning and Decision Laboratory at the Universityof Arizona with a group of city planners is discussed.

KEY WORDS AND PHRASES: Knowledge base, knowledge management, knowledge-based management systems, model management, organizational planning, groupdecision support systems, planning systems.

Few companies today would say that they are happy withthe way they plan for an increasingly fluid and turbulent

business environment.Pierre Wack, Royal Dutch Shell

1. Introduction

Traditionally, organizational planning has been tiased on forecasts. This ap-proach is appropriate if the assumption can be made that the future business environ-ment will remain relatively st^le. Many organizations, however, face a busiiKssenvironment that is complex and unstable. These organizations are becoming awarethat traditional forecasting tools are inadequate for capturing the utM:ertain assump-tions upon which current strategic planning decisions are based [23]. Surfacing ideasand assumptions upon which plans are develcqied and providing an oj^wrtunity fordebate and analysis of those issues and assumjrtions by a number of experts are usefulmethods for effective planning [23]. The ability to link isstus and assumptions toplanning decisions atid to store that infra-mation for future reference has also beenidentified as necessary to provicte the flexibility and adaptability necessary for

Page 3: Knowledge Management

22 APPLEGATE, CHEN, KONSYNSKI, AND NUNAMAKER

effective strategic planning in times of rapid environmental change [1].Recently there has been a call to expand the focus of decision support systems to

include support for complex, unstructured decisions within organizations [9, 18].There is growing recognition that the ability to provide support for this class ofmanagement decisions will require the integration of knowledge-based expert sys-tem techniqties with traditional decision support system architectures [8]. Organiza-tion planning, as described above, is a classic ill-structured and relevant problemfacing organizations and provides an excellent domain for the study of knowledge-based information system support of complex, unstructured management processes.This paper analyzes the characteristics of the planning process in an attempt todetermine the knowledge requirements that can be used to guide the design ofinformation systems for support of organizational planning. The design of a knowl-edge-based management support system for organizational planning, the PLEXSYS

Planning System, is described. This system is designed to provide informationsystem support throughout the planning process. Internal and external data manage-ment systems and a model management system that provides access to both qualita-tive and quantitative planning models are integrated with a process managementsystem that controls access to the models and data based on the class of organizationplanning process (e.g., strategic planning, information systems planning). Theinformation generated by the planners during a planning session is integrated usingsession analysis tools and is stored in the system as an instance of a plaiuiing framerepresenting a scenario. A domain-independent knowledge management system isused to design, implement, and maintain the domain-dependent PLEXSYS PlanningSystem. A case study illustrating the use of the system by a group of city anduniversity planners is presented.

2. Knowledge Requirements for Organizational Planning

A C K O F F [ 1 ] H A S defined organizational planning as ' 'the design of a desired futureand of effective ways of bringing it about." Planning is a type of decision-makingprocess but not all decision making is planning. Ackoff [1] has identified severalimportant distinctions that characterize planning decisions.

1. Planning is done in advance of taking action and can, therefore, be described asanticipatory decision making.

2. Planning is required when the desired future state involves a number of interre-lated, interdependent decisions—a system of decisions.

3. These systems of decisions are too large to handle all at once. This means thatthe planning process itself must be planned. This involves dividing the planningprocess into stages or phases that can be performed sequentially by one planninggroup, can be performed simultaneously by a number of planning groups, or caninvolve a combination of the two.

4. ITiese systems of decisions involve decisions that must be made in light of priordecisions. As a Fesult, the planning problem or task cannot be subdivided into

Page 4: Knowledge Management

KNOWLEDGE MANAGEMENT IN ORGANIZATIONAL PLANNING 2 3

independent subsets and planning must be carried out before action is required.The planning process has been described as a complex, unstructured process [1,

14]. The unstructured nature of the planning process derives from the anticipatorynature of the planning problems. Planning problems or tasks are often based on acritical set of uncertain, and often conflicting, assumptions and often involve novelsituations with no predetermined response pattern. The complexity of the planningprocess is primarily due to the interrelated nature of the system of planning deci-sions . Individual decisions within a system of decisions required for planning may besimple or complex. But when all decisions must be considered in light of all otherdecisions, much greater complexity results.

In an attempt to provide a more structured approach to planning and to reducecomplexity, a number of planning models have been proposed. The planning modelsinclude both structural models, designed to provide a framework for the planningprocess, and process models, designed to define a sequence of steps (stages) in theplanning process. Examples of planning model classifications include (1) Anthony'sorganizational triangle that classifies organizational processes as operational controlprocesses, management control processes, and strategic planning processes [2]; (2)Ackoffs strategic planning and tactical planning classifications [1]; (3) Hofer andSchendel's corporate planning and business planning classes [11]; and (4) Ham-brick's functional planning class [10]. Planning process models include: (1) Witte's[24] four-stage model (information gathering, alternative development, alternativeevaluation, and choice); (2) Mason and Mitroff s [14] five-stage model (problemsensing, problem defining, solution derivation, implementation, monitoring, andevaluation) and Nutt's [17] five-stage model (formulation, conceptualization, detail-ing, evaluation, and implementation).

A number of decision aids and tools have been developed to assist planners inaccomplishing specific planning tasks [19]. These include tools for gathering infor-mation and generating ideas and alternatives, for synthesizing information, ideas,and alternatives, for analyzing information, ideas, and alternatives, and for choos-ing among competing alternatives.

The characteristics of the planning process, discussed above, can be used to helpguide the design of an automated system to support organizational planning pro-cesses. A summary of the requirements for information system support of organiza-tional planning is presented below.

Anticipatory Nature of Planning Decisions

* tools for elicitation of assumptions on which decisions, models, and information

are based* capability to link assumptions to planning decisions and decision analysis models* knowledge representation mechanisms that enable the storage of the planning

decision history within the computer and maintain the links among the decision,models, assumptions, and data used to support the decision

* capability to retrieve the planning decision history and make changes in themodels, data, and/or assumptions based on new information or new decisions

Page 5: Knowledge Management

24 APPLEGATE, CHEN, KONSYNSKI, AND NUNAMAKER

InterdqKndent, Interrelated System of Planning Decisions

* capability to link specific decision histories to form a system of decisions thatdefines a planning process

* knowledge representation mechamsms that enable the storage of the system ofplanning decisions making up a specific planning process and that maintain thelinks among the specific planning decisions

Complexity of the Planning Process

* capability to decompose the planning process into subprocesses (phases)* flexibility to enable sequential or simultaneous processing of each phase of the

planning process* ability to move back and forth from one plaiming process phase to another

Unique Nature of Many Platming Problems or Tasks

* knowledge representation that enables categorization of planning problems ortasks and provides a normative framework for approaching problems or taskswithin each category

* capability to modify the normative framework based on the unique characteristicsand demands of the specific planning problem or task

Nature of Planning Information

* knowledge representation that enables the storage of quantitative and qualitativeinformation within the same knowledge management system structure

* access to multiple automated stores of internal and external data and the ability todownload that information into a temporary rapid access storage structure for useduring real-time interactive planning sessions

* capability for a number of planners to access the system and conduct real-timeinteractive planning sessions and share particular areas of expertise

Nature of Planning Information Analysis

* capability for storage and rapid access to qualitative and quantitative planningdecision aids

* knowledge representation that permits die abstraction of planning decision aidsand storage as unique fragments that can be accessed independently or as a group

* knowledge management tools that allow a knowledge base designer to create newplanning decision aids and modify/delete existing ones

* capability for all d»:ision aids to be run as stand-alone or networked versions tosupport individual and group decision making

The PLEXSYS Planning system, described in the next section of this paper, isdesigned to meet the need for providing informatimi systems su[^r t throughout the

Page 6: Knowledge Management

KNOWLEDGE MANAGEMENT IN OKGANIZATIONAL PLANNING 2 5

BBRERKOUTROOM

BRERKOUTROOM

BBRERK-OUTROOM

B

mi SCREEN

CD[D Dtsnitv

SVST£M

03inQ]

XQQX

UIOEODISK

MICROS

BRERKOUTROOM

LOBBV

Figure 1. Management Information Systems (MIS) Planning and Decision Laboratory

planning process. It integrates internal and external information management sys-tems and a model management system providing both qualitative and quantitativeplanning models within a group decision support system focus. A domain-indepen-dent knowledge management system is used to design, implement, and maintain thedomain-dependent PLEXSYS knowledge base. This research draws on the PLEXSYS

knowledge base design described by Konsynski, Kotteman, Nunamaker, and Stott[12] and used by Mclntyre, Konsynski, and Nunamaker [15]. The model manage-ment system design for strategic planning described in [4, 5] provides a frameworkfor the management of models within the PLEXSYS Planning System.

3. Knowledge Management in the PLEXSYS Planning System

THE PLEXSYS PLANNING SYSTEM provides an integrated workbench of planningand decision models that can be used by organization planners to assist in theunstructured process of organization planning. The system is available for use byplanners in the Management Information Systems (MIS) Planning and EtecisionLaboratory. This laboratory provides a research facility for the study of the planningand decision process while top executives from a variety of organizations use thefacili^ for organizaticHial planning and for solving cam^Aex, unstriKtured decisionproblems. A diagram of the laboratory is jwesented in Figure 1.

A large U-shaped table is equipped with networked microcomputers that arerecessed into the table to facilitate interaction among participants. A microcomputer

Page 7: Knowledge Management

2 6 APPLEGATE, CHEN, KONSYNSKI, AND NUNAMAKER

attached to a large screen projection system is also on the network, which permitsdisplay of work done at individual workstations or of aggregated information fromthe total group. Breakout rooms are equipped with microcomputers that arenetworked to the microcomputers at the main conference table. The output fromthese small group sessions can also be displayed on the large screen projector forpresentation to the group and can be updated and integrated with the planning sessionresults.

The participants interact with a variety of automated and manual platining andproblem-solving models that provide support throughout the planning and decisionprocess. Four general classes of planning and problem-solving models are imple-mented in the system: (1) idea and information generation models, (2) idea andinfortnation structuring and analysis models, and (3) models for choosing amongcompeting alternatives.

The output from all three classes of models serves as input to a knowledge basethat provides a mechanism for representing and storing the planning knowledgeusing a variety of knowledge representation techniques (including frames, semanticinheritance networks, and production rules). The knowledge base design allows formultiple planning and decision process representations at any time. These represen-tations can change dynamically as new knowledge is added to the system.

3.1 Knowledge Elicitation in the PLEXSYS Planning System

classes of knowledge are represented within the PLEXSYS Planning System.These are: (1) planner knowledge and (2) facilitator knowledge.

Planner knowledge is elicited through the use of qualitative and quantitativeplanning models stored within the model management component of the system.Figure 2 lists the models for elicitation of planner knowledge that are currentlyavailable (or planned for implementation) in the PLEXSYS Planning System. In mostcases, both group and individual versions of the model are available.

Facilitator knowledge is elicited through the process management system. Theprocess management system is a collection of software tools that allow the facilitatorto specify the goals of a planning session, to de&ie the necessary session activitiesthat must be accomplished to reach those goals, to run the plantiing session, and toanalyze the output of a session and prepare reports for the planners. The processmanagement system accesses a library of network management software tools toallow for the group decision support focus that is crucial to the planning process. Inaddition, the process manager controls all access to and storage of planning knowl-edge in the knowledge base as planning scenarios. This latter function will bedescribed in detail in the discussion on knowledge representation and storage.Figure 3 presents an overview of the elicitation of planner and facilitator knowledgethrough the PLEXSYS Planning System.

Page 8: Knowledge Management

KNOWLEDGE MANAGEMENT IN ORGANIZATIONAL PLANNING 27

INFORMATION GATHERING / IDEA GENERATION

ELECTRONIC BRAINSTORMINGSTAKEHOLDER IDENTIFICATION AND ASSUMPTION SUHFACING

ENVIRONMENTAL ASSUMPTION SURFACINGPRODUCT / MARKET ANALYSISINDUSTRY STRUCTURE ANALYSIS

ORGANIZATION CAPABILITY ANALYSIS

INFORMATION / IDEA STRUCTURING AND ANALYSIS

ELECTRONIC BRAINSTORMING IDEA ANALYSISASSUMPTION ANALYSIS

FINANCIAL RATIO ANALYSISCASH FLOW ANALYSISPORTFOLIO ANALYSISFORECASTING MODELS

STATISTICAL MODELINGDETERMINISTIC MODELING

CAPITAL BUDGETINGCOST/BENEFIT ANALYSIS

CHOICE

CALLABLE VOTING MODELSMULTIATTRIBUTE DECISION-MAKING

Figure 2. Models for Eticitation of Planner Knowledge

3.2 Knowledge Base Definition

The planning and facilitator knowledge is stored within the PLBXSYS Planning Sys-tem Knowledge Base, The knowledge base is defined by the knowledge base man-ager using the PLEXSYS Knowledge Base Description Language, PLEXL, PLEXL is afortnal process description language based on the Problem Statement Language/Problem Statement Analyzer (PSL/PSA) language for describing an information sys-tem [22]. The PLEXL language is used to specify the domain for a specific knowledgebase by defining the semantic network of terms and expressions and the frames thatare used to chunk knowledge and promote efficient storage and search.

Four levels of terms and expressions are used to fully describe a process in thePLEXL language. These are: (1) physical level terms and expressions, (2) axiomaticlevel terms and expressions, (3) median level terms and expressions, and (4) in-stance level terms and expressions. An in-depth discussion of the language con-structs that form the background for the PLEXL language can be found in [ 13,21 ]. Adetailed description of the syntax and semantics for the PLEXL language and the

Page 9: Knowledge Management

28 APPLEGATE, CHEN, KONSYNSKI, AND NUNAMAKER

PLEXSYS PROCESS MWAGEMEKT BVSTEH

SESSION QUERV

( I >

(5)relevant infoon process.

models K dataPLANNING SESSIONFACILiTATOR

(6)

select Kinitializemodels

select

reports

PLfiNNERS-

runmodels

NETWORKCONTROLtlBRORV

MODELMGMTSVSTEh

DftTAnSMTSYSTEM

Z)modelesuIts

Query facilitator forsession info fc pass toknowledge base.

SESSION INITIALIZATION

Query knowledge base forrelevant info on processmodels and data. Dis-play for facilitator.

(2)

Ct)relevant infoon process.

Interactive selection i,initialiiation of modelsby facilitator.

(9)

SESSION CONTROLLER

Run planningmodels.Allow facilitator tointeractively change &update session.Interactive selection Qfnetwork management tools

REPORT GENERATOR

Prepare reports forplanners & facilitator

(15)integrated model

results

SESSION INTEGRATORS& ftNALVZERS

Analyze planning sessionresults in light of currentknowledge in KB.Integrate info obtainedfrom planning models S. KB.

<T3)descriptive/normative model

DESCRIPTIVEPHOCESS MODEL

NORMAT1VEPROCESS MODEL

Figure 3. Elicitation of Planner and Facilitator Knowledge in the PLEXSYS Planning System

PLEXA analyzers used to ensure the integrity, consistency, and completeness of the

language can be found in [3].The physical level provi^s the domain independent structure that initialiros and

maintains the j*ysicid level of the knowledge base. Creation of the physical level ofthe PLEXSYS Knowlalge Base begins by naming the Knowledge Base GeneratOT tool.This initializes the domain-independent knowledge base network and terms by

Page 10: Knowledge Management

KNOWLEDGE MANAGEMENT IN ORGANIZATIONAL PLANNING 2 9

generating the root, record stnicture, and link mechanisms of the netvrork.At the physical level, the knowledge base is strucOired as a B^-tree. Each node in

the network is a variant record that contains a number of variable lengtii fields thatcontain path information, including inheritance properties and attribute properties,for the knowledge stored within the network. IVvo basic link mechanisms are avail-able. The "ISA" link enables the definition of the inheritance properties betweenterms (construction of the semantic inheritance network). The "HAS" link enablesthe definition of the attribute properties of terms (consbmction of the frame repre-sentations). The HAS link can be expressed through a number of relationship terms.The choice of these terms is dependent on the domain of the knowledge base.Examples of relationship values are "CREATES," "USES," and "UPDATES."The basic structure of the knowledge base, created with the Knowledge Base Gener-ator, places only the minimal restrictions necessary to configure the semantic inheri-tance network, frame, and production rule knowledge representations that are avail-able for use by the knowledge manager in creating specific knowledge bases. At thispoint the knowledge base is still domain-independent.

Four PLEXL Editors are currently available to etiable the creation and update ofdomain-specific knowledge bases within the domain-independent PLEXSYS Knowl-edge Base. These are the: (1) Term Editor, (2) Expression Editor, (3) Median Editor,and (4) Instance Editor. A Rule Editor is in the process of development and will notbe discussed in this paper. All four editors use the PLEXL language to define theknowledge base terms and expressions.

Tbe Tbrm Editor is used to define the abstract terms (axiomatic level and highlevel median) for the knowledge base. These terms are then litiked to form expres-sions using the Expression Editor. These expressions create the basic frame andinheritance network stnicture that serves to organize and define the most abstractlevel of planning knowledge. This knowledge base definition must be elegantenough to provide efficient search of the network but semantically rich and completeenough to fully describe the knowledge domain for the knowledge base.

The Median Editor is used to defme the detailed (median level) terms and expres-sions tiiat enable complete expression of the knowledge domain for the knowledgebase. The median level is used to create the deep layers of the semantic inheritancenetwork and to define the frame representations that provide detailed information onthe organization of planning knowledge within the knowledge base. This providesmultiple views of the semantic inheritance network and helps ' 'chunk'' the planningknowledge for more effective representation and efficient management.

The Instance Editor enables the system to store multiple instantiations of a medianlevel frame within the knowledge base. Knowledge at the instance level involves theinformation obtained from the planning sessira facilitator and phmners using theorganization planning tools and session management tools. The raw data from theplanning session tools is analyzed by the PLEXSYS Analyzers, PLEXA . Once analyzed,the planning session instances are loaded into the knowledge base using a batch loadprogram or the interactive Instance Editor. An overview of the PLEXSYS KjiowledgeManagement System is presetted in Figure 4.

Page 11: Knowledge Management

3 0 APPLEGATE, CHEN, KONSYNSKI, AND NUNAMAKER

PLEXSVS KNOyLEDGE MflKfleEMENT SYSTEM

i

PLEXL

KB SENERATOR

GBri«r«t«(B primitivelevel of the kbincluding b»-trBBroot, variant record& links

primitive token> inheri-tance & attribute links

(5)

TERMEDITOR

Defines axiomatic:level terms

abstract termsisynonyms hdocumentat ion

(5)

EXPRESSIONEDITOR

Defines alevel and 1st levelmedian terms andexpressions (abstractBi-net and frames)

consistency i,completenessviolations

abstract framest, semantic in-heritance nets

(7)(9)

MEDIANEDITOR

Defines median levelterms and expressions(detailed se-net t.frames >

consistency S.completenessviolat ions

detailed frames& semantic in-heritance nets

< 10): ! I 1 )consistency &completenessviolations

PLEXA

FACILITftTDR

p1anningsessioninfo2)

INBTANCEEDITOR

Interactive definitionof instance level termsand expressions

frameinstance

PLEXBYS

KN0UL£DGE

kb integrity,consistency t.completenessreports4) KB

_< 13)

BATCH I/O

Figure 4. PLEXSYS Knowledge Management System

3.3 Knowledge Representation, Storage, and Management

Three knowledge representation techniques are used in the PLEXSYS Planning Sys-tem. These are: (1) semantic inheritance networks, (2) frames, and (3) productionrules.

Seniantic inheritance networks are composed of nodes and links between thenodes. The nodes are the storage structures for the knowledge while the linksrepresent the interrelationships of the knowledge [6]. The PLBXSYS Planning System

Page 12: Knowledge Management

KNOWLEDGE MANAGEMENT IN ORGANIZATIONAL PLANNING 3 1

uses a variant record structure to describe the planning knowledge at each node of thesemantic inheritance network. This knowledge is divided into four basic categories.These four categories store knowledge about: (1) terms, (2) expressions, (3) docu-mentation, and (4) rules. Term knowledge includes the name of the term, termaddress, term class (axiomatic, median, etc.), offset of the first and last expressionsthat include the term, index information on the inheritance and attribute links, andpointers to synonyms and documentation. If the term is a relationship term (e.g.,CREATES, UPDATES) information is stored on whether this is an attribute orinheritance relation term.

Expression nodes contain information about expression length, expression ad-dress, and expression class (axiomatic, median). A pointer to the next expression inthe group is also maintained within the expression node record. Document nodescontain up to five lines of documentation on a specific term or expression. Rulenodes contain information on the rule name, rule value, number of conditions,antecedent, and action. Pointers to other terms within the semantic inheritancenetwork are also included.

Frames are knowledge structures that are used to organize and classify knowledgeabout a limited domain within a knowledge base. A frame creates a description of theobject or action in question, starting with an invariant structure common to all casesin its domain and adding certain attributes that identify a specific class of cases.When combined with a semantic inheritance network representation, as is the casewith the PLEXSYS Planning System, the frame can offer multiple views on a complexknowledge domain and, therefore, simplify search and reasoning.

There are two classes of frames implemented within the PLEXSYS KnowledgeManagement System. Descriptive frames provide a structural framework for theknowledge base. Descriptive frames help to classify planning cases and providestructural organization of the planning process. They partition the planning domaininto domains of expertise that contain the knowledge necessary to create a descrip-tion of an object in that domain. Examples of descriptive frames implemented in thePLEXSYS Planning System are process frames, model frames, and scenario frames.Some knowledge within the frame tells how to take a set of observations and create acorrespondence between those observations and the descriptive mechanism of theframe. Other knowledge allows the frame to predict some features of the descriptionafter observing others. Transformation knowledge allows the knowledge base tomaintain the description even if there are minor inconsistencies and missing data.

In addition to the descriptive frames, discussed above, the PLEXSYS PlanningSystem also incorporates action frames that control the sequence of events andmovement within and between phases in the planning process. These action framesare similar to a class of frame knowledge representations alternatively referred to asscenarios [16], scripts, or plans [20]. The action frames are used in the ProcessManagement System to control the order of presentation and miming of a planningsession and for the interactive creation of a planning session scenario.

The frame knowledge is used by the Process Management System to build anormative model of the proposed planning ^ssion using the relevant median level

Page 13: Knowledge Management

3 2 APPLEGATE, CHEN, KONSYNSKI, AND NUNAMAKER

PROCESS PRAME

r- «oc«pt«—-INPUT

PROCESS creates—OUTPUT

- h«B INTERFACE ELECTRONIC BRAINSTORMING

PROCESS MANAGER ill

i— has MODEL

liri " '

MODEL

has StJB-MODEL

l a a

- reads IKPUT FILE

-creates—OtlTPUT FILE

STRATEGICFLANKING

MOTE: This figure preoents only a sntall portion of a process andmodel frame

Figure 5. Process and Model Si-Net

frames for the specific class of organizational planning related to a given planningsession. This normative model of the planning process is compared to a descriptivemodel of the planning session that represents an instantiation of the frame based onthe information collected during the planning session. These two models are com-pared and transformation and correction rules are used to relate the normative anddescriptive models prior to entering the information into the knowledge base as aframe instance.

Examples of process and model frames defined using the PLEXL language arepresented in Figure 5. A sample instantiation of a model frame based on informationcollected during an actual planning session held in the MIS Planning and DecisionLaboratory is presented in the next section of this paper.

4. Experience with the Use of the PLEXSYS Planning System

THE MIS PLANNING AND DECISION LABORATORY opened in March 1985 withthe first planning sessions beginning in September 1985. Since that time over onehundred planners have used the facility to conduct strategic planning sessions. Table1 presents an overview of some of the groins who have used the PLEXSYS PlanningSystem.

Each of the cases listed above involved two to four planning sessions over a one- tosix-month period. Planning infonnation collected during a representative planningsession is presented to illustrate the instantiation of axiomatic and median level

Page 14: Knowledge Management

KNOWLEDGE MANAGEMENT IN ORGANIZATIONAL PLANNING 3 3

Table 1 Groups Using tiie PLEXSYS Planning System(September 1985 to June 1986)

Group

Center for Computing and Informa-tion Ttechnology

City plannersGovernment planning gr(»jp

Major computer manufacturingcompany

State university

Purpose of session

Strategic planning for organizationinformation services

City transportation planningStrategic planning for national and

international information servicesNew market development

Strategic planning at the centraladministration level

Strategic planning at the collegeieveS

Strategic planning at the depart-ment level

Number ofparticipants

16

811

6

19

42

12

114

Note: The figures for strategic planning at the college level of the state university reflectthree different planning groups conducting separate planning sessions for their organiza-tions that were then integratol with the results from the central administration and depart-ment level planning sessions^

frames for knowledge management in organization planning.The general goal of the planning session, from which the data presented in this

paper were obtained, was to analyze the issues surrounding the construction of atunnel under a major city street near a university. The proposed tunnel had createdconsiderable controversy and was a major issue to be voted on in an upcomingelection. The university became involved in this controversy because the tunnelsolution would enable the university to continue expansion of the campus to the northwhile still maintaining an integrated campus. City planner, university planners,students, and concerned citizens met in the planning laboratory to analyze thecritical issues surrounding the proposed project.

The Electronic Brainstorming model was one of the planning models chosen toelicit planning knowledge regarding the critical issues facing the city in the decisionto build the proposed tunnel. The Electronic Brainstorming procedure is presentedbelow.

Each group member sits in front of a microcomputer tiiat has theElectronic Brainstorming screen on the video display. The specific ques-tion or topic for the idea generation session is displayed at the top of thescreen. Each planner presses die Fl key and a comment window appearson the screen. The planners are instructed to type in a short idea (nomore than 5 lines long) that relates to the question at the tq) of thescreen. When finished entering the idea the planners press the Fl keyagain and their file is shipped to the server where it is exchanged with afile containing ideas ftara oCt&t group members. As the se^ion contin-

Page 15: Knowledge Management

34 APPLEGATE, CHEN, KONSYNSKl, AND NUNAMAKER

ues, the page that once contained only the question becomes filled withindividual ideas from different members of the group. The ability toscroll through the ideas is provided so that group members can read thecomments of (rther group membere before and during the entry of theiridea. The brainstorming session continues for approximately 30 to 40minutes. At the end of the session, each group member presses shift F2and the page of ideas at each terminal is shipped to the server where it isappended into a single file that is used to categorize and prioritize thegenerated ideas.

After the planners finished generating ideas using the Electronic Brainstormingmodel, they met in small groups to identify stakeholders and assumptions using theStakeholder Identification and Assumption Analysis models. During this time thefacilitator used the Electronic Brainstorming Issue Analysis model to develop cate-gories of issues and link the issues that were generated by the planners to thecategories as supporting statements. By the time the planners were finished runningthe second planning model, the categorized issues were ready for their review andrevision. (An alternate method allows each planner to develop issue categories thatare then agreed upon by the group and supporting statements are interactively linkedduring a group discussion session.)

After the issues were categorized, a callable voting program was used to allow theplanners to anonymously vote on the issue categories and, if desired, the supportingstatements. Statistical software calculated the mean and standard deviation of thevotes and displayed this information with a histogram of the individual votes.Discussion of the voting results and revision of the votes was conducted until aprioritized list of issues was identified.

The prioritized, categorized issues along with the related voting information andsupporting statements were stored in the knowledge base for future access. ThePLEXA analyzers assured that the information generated during the planning sessionwas consistent with the knowledge base definition of an Electronic Brainstormingframe. Any inconsistencies or violations of integrity rules were reported to thefacilitator and the knowledge base manager. An example of the instantiation of theElectronic Brainstorming model frame using the planning information collectedduring the previously described planning session is presented below.

ELECTRONIC BRAINSTOflMlNG(HAS) SUB_MODEL, EBS^tNITIALlZATION

(HAS) INTERFACE, FACILITATOR(HAS) NAME: FACILITATOR_NAME: NUNAMAKER(HAS) TITLE: FAC!LITATOR_TITLE: DEPT HEAD, MIS

(PERFORMS) OPERATION; 1NITIALI2E_EBS_SESSION(CREATES) FILE: SESS10N_INIT_FiLE

(CONTAINS) ELEMENTS: CASE_NAME: SPPLANSESSION_#: 1NUMBER_OF_PARTICIPANTS: 8NAME: SUSAN A.

TITLE: STUDENTNAME: ALAN H.

TITLE: CITY_PLANNER

Page 16: Knowledge Management

KNOWLEDGE MANAGEMENT IN ORGANIZATIONAL lUANNING 35

(a linked list of participants is developed until all 8 planners are identified)DATE_OF_SESSION: MAY 2. 1986EBS_QUESTION: WHAT ARE THE CRITiCAL ISSUES THATMUST BE CONSIDERED IN DEVELOPING THE PROPOSEDTUNNEL

(HAS) SUB_MODEL: EBS_IDEA_GENERATION(HAS) INTERfACE: PLANNERS(PERFORMS) OPERATIONS: IDEA_GENERATION(READS) FILE: EBS_INIT_FILE

(USES) ELEMENTS: EBS_QUESTION: (see above)(REQUESTS) ELEMENTS: PLANNER_IDEAS:

IDEA 1: Something needs to be done about the traffic problem that exists on the north side of friecampus.

IDEA 2: The expansion of the university north of Speedway is important for the growth anddevdopment of the university. With this growth it is necessary to integrate the campus.

IDEA 3: The cost of building the tunnel seems too great for the supposed benefits received.Alternative methods which view the traffic problem as city-wide problem must be pro-posed.

(a linked list of ideas is developed until all ideas are listed)(UPDATES) FILE: EBS_RAW_DATA_FILES

(CONTAINS) ELEMENTS: EBS_RAW_DATA

(HAS) SUB_MODEL: EBS_MERGE_EDIT(HAS) INTERFACE: FACILITATOR(PERFORMS) OPERATIONS: MERGE_RAW_DATA_FILES

EDIT_RAW_DATA^FILES(READS) FILE: EBS_RAW_DATA^FILES

(USES) ELEMENTS: PLANNER_IDEAS(REQUESTS) ELEMENTS: FACILITATOR_EDITS(UPDATES) FILE: EBS..._IDEA_FILE

(CONTAINS) ELEMENTS: EDITED_MERGED_IDEAS(HAS) SUB_MODEL: EDS_CATEGORIZATION

(HAS) INTERFACE: FACILITATORPLANNER

(PERFORMS) OPERATIONS: CREATE_CATEGORIESLINte_SUPPORTING_STMTS

(READS) FILE: EBS—iDEja^FILE(USES) ELEMENTS: EDITED_1DEAS

(REQUESTS) ELEMENTS: EBS_CATEG0RY1: COSTEBS_CATEG0RY1__SUPP_STMTS

The cost 0* building the tunnel seems too great for the supposed benefits received. Alternativemethods which view the traffic problem as city-wide problem must be proposed.

There is no accurate estimate of costs

The cost of the project has been estimated. Furthermore, the money has already been set asidefrom the 1984 taxes. No new taxes will be levied.

EBS_CATEG0RY2: UNIVERSITY EXPANSIONEBS_CATEGOF1Y2_SUPP_STMTS:

The expansion of the university north of Speedway is important for the growth and development ofthe university. With this growth it is necessary to integrate the campus.

Perhaps it would be appropriate to present and discuss the current master plan for universitygrowth for the next 20 years. Does the university have to sw^low up neighborfioods to accompli^its objectives?

Trying to stop the growth of the university is a fruitless task. The fact is that the university is going togrow and require additional space whether the tunnel is approved or not.

Why can't the university expand to the South?

(a linked l i ^ of idea categories and ajpporting statements is developed until all categories andsupporting statements are identified)

Page 17: Knowledge Management

36 AWLEGATE, CHEN, KONSYNSKI, AND NUNAMAKBR

(UPDATES) FtLE: EBS_CArEG_..FlLE (see stoove)(CONTAINS) ELEMENTS: EBS_CATEGORtES

EBS_SUPP_STMT

(HAS) SUB_MODEL: EBS__PRtORITIZATION(HAS) INTERFACE; FACILITATOR

PLANNER(PERFORMS) OPERATIONS: ASSIGN_INDIV_PRlORtTtES

CALCULATE_GROUP_PRtORITIESCREATE_PRIORITIZED__LtST

(READS) RLE: EBS_CATEG_F1LE(USES) ELEMENTS: EBS_CATEGORIES

CATEG__SUPP_STMT_LINKS(REQUESTS) ELEMENTS: INDIV_CATEG_RAT1NG: PLANNER1

CATEG0RY1: 8CATEG0RY2: 4CATEG0RY3: 9CATEG0RY4: 5CATEG0RY5: 3

INDIV_CATEG_RATING: PLANNER2CATEG0RY1: 6CATEGORY2: 9CATEG0RY3: 8CATEG0RY4: 4CATEGORY5: 2

(a linked list of individu^ category ratings is developed until all planner ratings are identified foreach idea category)

(UPDATES) FILE: EBS_PRIOR_FILE (see above)(CONTAINS) ELEMENTS: EBS_CATEGORIES

CATEG_SUPP_STMTStNDIV_CATEG_RATlNG!ND1V_SUPP_STMT_RATtNG (n/a)NUMBER_VOTING: 8MEAN_.CATEG1_RATING: 7.2CATEG1_STD_DEV: .8MEAN_CATEG2_RATING: 6.3CATEG2_STD_DEV: 12

(a linked list of category summary statistics is developed)(HAS) SUB_MODEL: EBS_REPORTS

(HAS) INTERFACE: FACILITATOR(PERFORMS) OPERATION: PRODUCE_EBS_REPORTS(READS) FILE: EBS_INIT_FILE

EBS_RAW_DATA_FILEEBS_IDEA_FILEEBS_CATEG_FILEEBS_PRIOR_FILE

(UPDATES) RLE: EBS_INIT_REPORTEBS_R AW_DATA_R EPORTEBS_IDEA_REPORTEBS_.CATEG_REPORTEaS_PRI0R_REPORT

The information collected using the Electronic Brainstorming model is integratedwith planning information collected using taher idea generation tools, planningtools, and quantitative analysis models. This integrated information is used to createa planning scenario that describes die assumptions and issues upon which strategicplanning decisions are made. The instantiations of the individual model frames arestored in the PLEXSYS Planning System Knowledge Base along with the instantiationsof the planning scenario frames. Access to the frames is provided by m interactivequery program that allows retrieval of the information by planning process case,model, or scenario.

Page 18: Knowledge Management

KNOWLEDGE MANAGEMENT IN ORGANIZATIONAL PLANNING 3 7

5. Summary

T H E ABILITY TO PROVIDE information system support for unstructured decisionprocesses within organizations is rapidly becoming a reality. The PLEXSYS PlanningSystem combines advanced microcon:^»lter technology with an understanding of theknowledge requirements for organization planning to enable the elicitation, repre-sentation, storage, and management of planning information. Internal and externalorganization planning information, qualitative and quantitative planning decisionaids, and a variety of information structuring and analysis models have been inte-grated to allow support for the planning process from initial formulation of theplanning problem or task to implementation of the plan.

The PLEXSYS Knowledge Management System, a domain-independent knowledgemanagement system, is used to define and maintain the PLEXSYS Ptarming Knowl-edge Ba^. This knowledge base utilizes a variety of knowledge representations(frames, semantic inheritance networks, and production rules) to represent and storeplanning knowledge within the system. The PLEXL process description language isused to define the knowledge representations. The FLEX A Knowledge Base Analyz-ers are used to ensure consistency, completeness, and integrity of the knowledgebase.

The system was implemented in the MIS Plamung and Decision Laboratory inMarch, 1985. Since September 1985 over one hundred planners from a variety oforganizations have used the system. This has provided an excellent opportunity toanalyze and improve the system design and to study the influence of the automatedtechnology on the planning process.

REFERENCES

1. Ackoff, R. L. A Concept of Corporate Planning. New York: Wiley, 1970.2. Anthony, R. N. Plarming and Control Systems: A Framework for Analysis. Boston:

Harvard University Press, 1965. , u,- u A3. Applegate, L. M. Knowledge management in organizational planning. Unpublished

doctoral dissertation. University of Arizona, 1986.4. ARjlegate, L. M.; Klein, G.; Konsynski, B. R.; and Nunamaker, J. P. Model manage-

ment systems: Proposed representations and future designs. Proceedings ofthe SixOi Armual International Conference on Information Systems. Indianapolis, 1985.

5. Applegate, L. M.; Konsynski, B. R.; and Nunamaker, J. F. Model nmn^etnentsystems: Design for decision support. Decision Support Systems, 2, 1 (1986), 81.

6. Brachman, R. J. On the episteniological status of semantic networks. Readings mKnowledge Representation. Los Altos: Morgan Kaufmann, 1985.

7. Chen, T Design of the PLEXSYS knowledge base. Working paper. University ofArizona, 1986. _ ,.

8. Dhar, V. On the plausibUity and scope of expert systems iii inanagement. Proceedingscfthe Nineteenth Hawaii International Conference on System Sciences, 19S6.

9. Bam, J.; Henderscm, J.; Kwjsynski, B.; and Keen, P A vision for decision wpportsystems. Uiqjublished manuscript, 1984.

10. Hambrick, D. C. C^)erationalizing the concept of busineiB level strategy in research.Academy qf Managemem Journal, 26 (1980), 231.

Page 19: Knowledge Management

38 APPLEGATE, CHEN, KONSYNSKI, AND NUNAMAKES

11. Hofer, C. W., and Schendel, D. Strategy Formulation: Analytical Concepts. St. Paul:West Publishing, 1978.

12. Konsynski, B. R.; Kottcman, J. E.; Nunamaker, J. R, Jr.; and Stott, J. W. PLEXSYS-84; An integrated devel<q)tnent environment for information systenis. Journal of Manage-mem Information Systems, 1, 3 (Winter 1984-85), 64-104.

13. Kotteman, J. E. Formalisms for business information systems developnent. Unpub-lished doctoral dissertation. University of Arizona, 1984.

14. Mason, R. O., and Mitroff, I. I. Challenging Strategic Planning Assumptions. NewYork: Wiley and Sons, 1981.

15. Mclntyrc, S. C ; Konsynski, B. R.; and Nunamaker, J. F., Jr. Automating planningenvironments: Knowledge integration and model scripting. Journal of Management Informa-tion Systems, 2, 4 (Spring 1986), 49-69.

16. Minsky, M. A framework for representing knowledge. In Brachman and Levesque,eds. Readings in Knowledge Engineering. Los Altos: Morgan Kaufliman, 1985.

17. Nutt, P. C. The origins of planning ideas and their influence on the planning process.Working Paper Series. Columbus: Ohio State University, 1983.

18. Phillips, L. D. A theory of requisite decision models. Acta Psychologica, 56 (1984),29.

19. Rowe, A.; Mason, R.; and Dickel, K. Strategic Management and Business Policy.Reading, MA: Addison-Wesley, 1985.

20. Schank, R. C , and Abelson, R. Scripts, Plans, Goals and Understanding. Hillsdale,NJ: Erlbaum Associates, 1977.

21. Stott, J. W. Principles for computer-aided information systems development. Unpub-lished doctoral dissertation. University of Arizona, 1984.

22. Tfeichrow, D., and Hershey, E. A. PSL/PSA: A computer-aided technique for struc-tured documentation and analysis of information processing systems. In Couger, Colter, andKnapp, eds. Advanced System Development/Feasibility Techniques. New York: Wiley andSons, 1982.

23. Wack, P. Scenarios: Utwharted waters ahead. Harvard Business Review (September-October/November-December 1985).

24. Witte, E. Field research on complex decision making processes—the phase theorem.International Studies of Management and Organization (1972), 156-182.

Page 20: Knowledge Management