Toward Computerizing a Causal Modeling Approach to Strategic Problem Framing

NOTES AND COMMENTS

Toward Computerizing a Causal Modeling Approach to- Strategic Problem Framing* Timothy J. Heintz College of Business Administration, Marquene University, Milwaukee, WI 53233

William Acar Department of Administrative Sciences, Graduate School of Management, Kent State University, Kent, OH 44242

ABSTRACT This paper proposes an object-oriented approach to the development of interactive software for the pu'pose of managerial problem solving. A prototype is being developed using CSM causal mapping to represent each manager's perceptions of the relationships between key variables of a fm's strategic situation. This paper suggests the design of GDSS that would enable a group of managers to discuss, learn from each other, and possibly develop consensus about decisions or their causes. Issues involving future development are discussed. Subject Areas: Decision Support Systems, Group Decision Processes, and Strategy and Policy.

INTRODUCTION Within decision support systems research is the evolving concept of a group support system (GSS). One aspect of a GSS suggests that technology can facilitate managerial problem framing and problem-solving processes [171. These processes usually involve examining the perspectives of multiple participants or stakeholders. This prospect raises serious questions on the use of computer technology in supporting the ill-structured decision-making processes associated with high-level managerial meetings. Following Ackoff [3], we advocate developing an understanding of a problem situation and formulating a strategy to attack a problem based on the recognition of causes as opposed to their end effects. We suggest the use of a combination of a causal mapping method and a computing approach called object-oriented programming (OOP)to analyze complex decision-making situations. By examining their causal maps, groups can form an understanding through agreement on the underlying assumptions associated with a problem situation. A recent study by Lind and Zmud [25] showed empirically that information frequency speeds up convergence of ideas or interests and that convergence affects innovativeness positively. Their study also cited communications channel richness as a predictor of convergence. Causal mapping presents a graphical basis for communicating and arguing logically about problem situations. OOP provides a computer implementation approach that makes it easier to handle the complexities of building causal models. *The encouragement and constructive criticisms of Dr. Leo Charalambides of the University of Hartford, Dr. Arun Rai of Southern Illinois University, Dr.G. Jay Weinroth of Kent State University, two anonymous referees, and an associate editor are gratefully acknowledged.

1220

19921

Heintz and Acar

1221

This paper describes our progress in building such a GSS. We demonstrate how OOP assisted with problems in conceptualizing and implementing the prototype system. We conclude with an evaluation of where our work will go from here, including a discussion on the need to incorporate more intelligence in the system, and to gain empirical evidenw on the feasibility and effectiveness of the suggested approach.

RELATED WORK Group Support Systems A GSS is part of what many call groupware. A relevant subfield of a GSS is group decision support systems (GDSS) [33]. Used primarily as research tools, most GDSS facilitate the collection of ideas and forming of consensus among individuals in face-to-face meeting situations. These systems provide support for agenda formulation, brainstorming on problems, and consensus formulation via a voting mechanism. DeSanctis and Gallupe [121 developed a three-level categorization for GDSS. Level-1 systems are those that enable the entry, sharing, and group evaluation of ideas; Level-2 systems provide specific tools to support decision making; and Level-3 systems facilitate the decision-making process by acting as an intelligent mediator. Current empirical research focuses primarily on Level- 1, decision-room environments. While an early study by Joyner and Tunstall [23] suggested that the computer itself has no significant effect on the quality of the group decision, other studies have suggested that GDSS applied to situations involving more complex tasks [18], or larger, more experienced groups [32] do impact decision quality positively. Other studies cite similar results for experienced groups [21], or when subjects use the technology over time [25] [39]. There are also studies suggesting that computer support may decrease satisfaction with the decisions produced and reduce the number of ideas generated [8] [18]. However, group systems research [32] indicates the opposite, reporting very positive results using GDSS to support a brainstorming activity. GSS applications incorporating Level-2 and Level-3 concepts include support for the negotiation process [20] and an electronic mail system [26] that uses an intelligent filtering mechanism. Also, both the SAMM and Groupsystems environments [14] provide a stakeholder analysis tool [2] [4] [15]. This later effort reflects an attempt at using the computer to perform a type of situation modeling as advocated in this discussion. The results of the efforts undertaken in the area of GDSS have been generally modest. Lacking is an understanding on how specifically the computer can augment group processes. It is not surprising that groups using a system which simply transmits ideas and tabulates votes perceive the computer as providing limited benefit. Our discussion here focuses on ways to build the Level-2 and Level-3 systems.

Comprehensive Situation Mapping (CSM) Causal mapping is a problem-solving approach that traces its philosophical roots to the work of Churchman on the design of inquiring systems [9], its theoretical roots to Ackoff and Emery’s work on adaptive and interactive social systems [5], its processual (process implementation) roots to Mason [28] and Mason and

1222

Decision Sciences

[Vol. 23

Mitroff [29], and its methodological roots to Axelrod [6] and Jones and Eden’s [22] formulation of influence diagramming (ID) methods for problem framing and problem solving. In proposing this approach, we argue that participants bring to a meeting or problem-solving situation differing mental models of the problem situation [7]. These cognitive models consist of personal constructs (which we symbolically represent as variables or factors) and linkages between them. Influence diagramming [13] [lS] [16] [27] is an implementation of cognitive modeling involving a diagramming approach in which variables or nodes are connected by a single type of linkage: a unidirectional link represented by an arrow connecting two nodes. The intensities of the influence relationships are ignored in ID--only their directions and signs (+ or -) are represented. Acar’s [13 comprehensive situation mapping (CSM) brand of causal mapping extends the ID concepts by assigning a strength as well as a direction to the links. There are also different types of causal links that model the traditional distinction between necessary and sufficient causality [3]. Nodes are either instrumental or goal variables. Instrumental variables may be those controlled by the environment (called the triggers of change), by an actor (called the levers), or they may be any intermediate or midnodes of the causal graph (endogenous variables). Figure 1 provides a simplified illustration of CSM applied to a marketing example. It shows how CSM allows for causal links and for modeling their quantitative parameters. Since CSM includes quantitative information (i.e., intensities and time lags), it provides additional power beyond ID. CSM enables a simulation-like computational power called the forward analysis which determines the effect of changing selected levers and triggers within the causal network. For the case illustrated in Figure 1, this helps a user formulate statements of the form: “If number of calls increases by x percent, then we would expect earnings to increase by y percent.” CSM thus aids the decision maker in generating alternative decision scenarios. The argumentation literature [31] [37] deals with the way decision makers attempt to convince each other of the cause-and-effect structure underlying their beliefs in the reality or inevitability of something. For individuals with differing points of view to reach a common understanding, they must understand each other’s mental models through the recognition of differences in either the variables or the linkages. Examination of these differences enables decision makers to understand the underlying assumptions associated with a coparticipant’s position. The use of the causal map in this regard moves the participants away from a mode of averaging out feelings and preconceptions and into a mode of recognizing the existence and influence of causal links. Both ID and CSM lend themselves to what Mitroff and Emshoff [30] called assumptional analysis and Acar [l] called the backward analysis of a complex business situation. Acar suggested that the assumptions a person brings to a situation can be discovered by examining the relationships implied by the causal map. As a result, a dialectic process for negotiation among multiple participants evolves through the resolving of differences among alternative maps. As described in the dialectical inquiry literature [3] [28] [30], such a process allows hidden assumptions to surface and meaningful discussion to begin. Dialectical inquiry develops consensus through explanation and negotiation rather than impulsive voting or a power play.

19921

Heintz and Acar

1223

Figure 1: Illustrative causal map for a marketing application (from [l]).

KEY:

=* -->Half

Full Channel Channel - - - -> Restriction ( 1 year) Time Lag

In their new book, Eden and Radford [17] set out to establish the linkage between problem formulation approaches, cognitive modeling, influence diagramming (ID) and a GSS. One of the most developed systems in use is Eden's [16] SODA technique. His cognitive mapping variety is a form of ID buttressed by extensive software. It suffers from its reliance on ID rather than a richer causal mapping method. In addition, the SODA software produces output too abundant and undifferentiated for strategic implementation. The approach suggested by this discussion produces a richer system and one easier for decision makers to understand. Object-Oriented Programming OOP [ 111 [38] views programming as a process in which the programmer creates special data structures called objects. These objects form a conceptual representation of the physical process that the programmer tries to encode on the computer. Each object consists of attributes that have values indicating the state of the object. Objects also contain procedures called methods. These methods act as operators that can transform the state of the object or related objects. Methods are executed by objects sending messages to other objects. This process of embedding both data and code within autonomous objects provides a great deal of flexibility in managing highly complex software systems. An OOP environment will often include a number of predefined objects for handling such tasks as data representation, user interface, and input-output operations. OOP

1224

Decision Sciences

[Vol. 23

systems are also typically implemented within sophisticated graphically-oriented computing environments. Within such environments, general object classes are usually available to manage a very rich, interactive windowing system with mousedriven terminal input and output. Other object classes often support multiprocessing, disk and printer output, and a variety of data types that allow easy manipulation of strings, arrays, and sets of objects. Implementation of CSM within a GSS necessitates a development approach that facilitates the generation of highly complex software. The use of interactive graphics and a need to coordinate the activities or maps of many individuals requires the use of software protocols that provide for easy communications between modules, access to graphical code libraries, and a rich information representation scheme. OOP provides such capabilities. Others, including the intelligent E-mail efforts [24]and hypertext-based systems [lo] rely heavily on OOP concepts within their implementation. A PROTOTYPE CAUSAL MAPPING IMPLEMENTATION

The prototype causal mapping system is composed of five major segments as follows: 1. Factor Identifier. This segment helps scope out the problem domain by identifying the critical factors or variables associated with a situation or discussion issue. 2. Situation Mapper. This segment assists in developing the causal maps. 3. Scenario Simulator. This segment helps test implications of varying assumptions by propagating change values through a causal map; essentially, it projects values for goals and subgoals at some future time. 4. Assumption Surfacer. This segment examines differences between causal maps with the purpose of enabling discussion and understanding these differences. 5. Consensus Formulator. This segment enables the formulation of new maps for replacing existing ones to represent the consensus of a given number individuals or groups. We envision a process in which participants interactively move back and forth among these phases. Individuals initially form their own causal maps through separate analyses of the problem situation. As part of the processdof generating a map, new variables or factors may surface. Thus we allow for the possibility to go back to factor identification. Similarly, the scenario simulation and the assumption surfacing modules may motivate individuals to go back and modify their maps before the consensus formulation step takes place. In formulating maps we use the factors established in the factor identification. The initially complex task of defining the problem and properly framing it within its domain can be a stumbling block [16] [17] [28].The CSM tool would handle this by having the participants agree on a set of public factor definitions. These represent a commonly agreed-on terminology. If new factors surface during the process of map formulation, the user can define these as private, or alternatively request that they be registered as public terms. The latter case requires agreement by other members of the group. The map drawing process depends heavily on the interactive graphics capability provided by OOP environments. Figure 2 illustrates the interface generated for out

19921

Heintz and Acar

1225

prototype. Here each participant has a window associated with it that provides menus for selecting analysis options. This window also displays instructions and information, such as messages from other participants. A second window exists for display and building individual maps. To add or change nodes or links, the user uses a pointing device such as a mouse to select, position, or move the item. If the user wishes to specify or examine more detailed information concerning a node or link, he or she can open additional windows. The graphical interface suggested by Figure 2 demonstrates how an OOP environment provides a natural means of presenting causal maps. Application-defined objects representing such entities as participants, nodes and links, coexist with system-provided objects such as windows, menus, and various graphical entities. Figure 3 provides a simplified illustration of the objects used in generating causal maps. In this diagram object classes are shown within the boxes. Although a single object class may have many instances, the diagram only presents a single box for the entire class of objects. The arrows within the diagram designate references within an object to other objects. Not shown in Figure 3 are object classes used to manage and drive the overall system and to generate the user interface. Within the situation analysis phase, a participant creates a map by opening a window and assigning a label to a map or map subsets. If desired, the user can open and reposition additional map windows. The system does not require all users to examine the same maps. This allows participants to argue alternative perspectives while displaying on the screen only the level of detail and the logical information deemed most appropriate. Each instance of a map contains a series of link objects that make up the map. Each link has two attributes indicating the cause and effect variables associated within the link. This representation illustrates the power of using an OOP approach. Where two or more links have a common cause or effect, they share the same (rather than a copy) node object, thus maintaining the referential integrity of these nodes. Further, we associate the methods that draw a node or link on the map with the individual node or link objects. The map object has its own procedure for drawing the entire map which sends messages that invoke the node and link draw methods when needed. It is in the node rather than the factor object that information is maintained on where to display a variable. This feature enables the definition factors to be independent of the specification of the variable as being part of a specific map. Within scenario simulation, menu options allow specification of change values (positive or negative) for each lever or trigger node. These changes are then propagated through the network of causal links to generate change values for the goal nodes. The scenario object references all the input and generated values. We maintain node value objects independently of the nodes to allow maintenance of alternative values for differing scenarios at different times. Individual scenario objects contain methods to accept input values and display all their values in textual form. The link and node objects handle value propagation by receiving and transmitting transmit value messages. This latter case implements constraint propagation concepts commonly found in the A1 literature [36]. Within assumption analysis, the system allows pairwise comparison of several causal maps and systematically examines differences between them. Two participants in this process establish point (a selected map) and counterpoint (the comparative

1226

Decision Sciences

[Vol. 23

Figure 2: Illustrative user interface for a prototype causal mapping system.

Remove Node

map) positions as represented by each individual’s (or group’s) causal map. An assumption object contains a textual narrative and a set of rules specifying the conditions under which the computer system should present the assumption for consideration. For instance, one type of assumption may involve inclusion or exclusion of a particular node within the counterpoint map. The map object itself responds to a message requesting a list of the nodes contained in itself but not in the comparative map. The assumption object for this case would have a rule stating that if the set of nodes generated by this comparison message is not empty, then present assumption with a list of the generated nodes. Similarly, assumption objects exist for detecting such items as the existence or type of links, strength of links, or the type of nodes (lever, trigger, or goal). This assumptional or backward analysis attempts to resolve issues through understanding of alternative perspectives. It accomplishes this by having decision makers temporarily shift away from the scenarios implied by their own causal maps and having them focus on map differences. It is through negotiation or resolution of these assumptions that new perspectives develop and consensus emerges. The consensus maps emerge through a dialectic process of group members discussing each other’s maps, reviewing scenarios to test their implications and amending them accordingly until differences are either negotiated away or accepted. Consensus may not always be reached; in which case, it is through understanding of varying assumptions that tolerance for differing positions is obtained. Our system implements consensus formulation as simply a user-directed, cutand-paste operation for building new maps from existing ones. Since new maps

Heintz and Acar

19921

1227

Figure 3: Object classes with attributes that establish links between objects used by the CSM System.

I

Participant

I

factor

KEY:

Reference to single object instance .--) Reference to a collection of object instances

may represent views of a group of individuals, we define a group object as a subclass of participant. A group behaves identically to a participant except for allowing shared ownership of maps.

IMPLEMENTATION PROBLEMS AND SOLUTIONS The language used in developing the prototype discussed here was Digitalk’s version of Smalltalk [35]. The use of standard Smalltalk objects reduced the programming effort significantly. We were able to experiment with different user interfaces and incrementally expand the functionality of the system. The prototype provided a generally good demonstration of the technical feasibility of developing this type of system. Our experience with the prototype surfaced a number of implementation issues. These include difficulties with operating the system over multiple environments or operating systems, the necessity of moving the prototype into a network environment, and the potential need for high-resolution graphics. The most recent versions of the Smalltalk use the Graphical User Interface (GUI) existing within Windows 3.0, OS2 Presentation Manager, and the Apple Macintosh. A standardized GUI allows easier integration of new applications and expansion of existing ones and may not limit the use of the software to a single hardware platform. Since many GUI implementations are based on object-oriented concepts, its use naturally fits with the philosophy of our overall development approach. However, even with the advent of GUIs, we found problems in moving code from one operating environment to another. We implemented various versions of

1228

Decision Sciences

[Vol. 23

our prototype in Smalltalk’s own windowing system, Windows 3.0,and OS/2 Presentation Manager. In converting from one environment to another, we encountered major compatibility problems and conversion difficulties. We need to continue to investigate alternative implementation environments. One possibility is using an and perhaps the UNIX-based, object-oriented version of the C language (C++) X-Windows GUI. Eventually GUIs will become more standardized. Also, the availability of more powerful OOP-based, systems development tools will reduce these problems. “he implementation of the software within a multiuser environment is paramount among the technical concerns. Conceptually, the object-oriented approach lends itself to the design of multiprocessing and multiuser environments. Although language extensions to Smalltalk provide support for multiprocessing and communications, today’s network operating systems fall short of providing high-level, network-based, application services for a GSS. In a recent paper, Heintz [19] suggested the use of an object-oriented software environment for managing GSS applications and communicating among GSS participants. We need an environment where multiple-computer, communications protocols already exist and developers could simply “plug in” GSS tools as needed. There is a need for improved graphics. Large maps could necessitate higher resolution than is currently provided within the IBM PC environment. The capability for scrolling a large map within a window, resizing a window or zooming a window to occupy a full screen provided by the Smalltalk system helps in this regard; however, as maps become larger, added capabilities for scaling and zooming on map segments may be necessary. There are also alternative workstation-type machines that provide high-resolution graphics capability, but at a higher cost. Also, moving applications to such machines raises the multiple-platform issue addressed earlier. SUMMARY This discussion presented our approach to using a causal mapping tool within a GSS for facilitating the problem framing process that decision-making groups undertake. We propose the use of an object-oriented development and programming approach to implement the tool. A prototype of this tool has been developed and implemented on a single machine using Smalltalk and Windows 3.0. A parallel effort is examining the concept of developing an object-based, GSS architecture for implementation of this and other GSS tools [19]. At present the majority of features described here have been reimplemented with this architecture. Work still pending includes moving and adapting the system to a computer network, and general testing, refining, and extending of already implemented features. Although there are other tool imp!ementations that attempt to have users map cause-and-effect relationships or more carefully structure the arguments they present, none provide as comprehensive support of a dialectically-based decision-making process. We offer the first computerized rendering of Churchman’s [9] dialectical inquiry and Mason and Mitroff‘s [29] and Acar’s [2] assumptional analysis. Although there are technical issues to resolve, our main concern is the incorporation of more intelligence into the system. Our prototype system does take an active tole in detecting assumptions, but there is much more to accomplish. We are considering ways to have the computer more actively assist in scenario development

19921

Heintz and Acar

1229

and evaluation, in the prioritization of differing assumptions, and in the use of causal maps themselves as a knowledge source. A more intelligent system could facilitate generation of causal maps by suggesting the use of certain generic linkages. These suggestions would depend on the problem context as described by a partial map and the factors or goals specified by the participant. One area that we have begun to focus on involves developing inferencing schemes that use rule-base concepts in combination with a repository of causal linkages. We also feel that case-based reasoning approaches [34] may suggest ways to use experience with existing maps to adapt and refine a knowledgebase built around representing causal linkages and that the OOP approach provides a means of integrating the knowledge component within the GSS. The ultimate test of this system will come with experience in real case situations. In some instances,manually applying the CSM technique becomes a time-consuming and difficult process. A computerized system could facilitate this process. As such, the proposed system might prove valuable in the implementation of more networked organizations and to those organizations attempting to install interactive planning [4]. [Received: December 31, 1990. Accepted: April 7, 1992.1 REFERENCES Acar, W. Toward a theory of problem formulation and the planning of change: Causal mapping and dialectical debate in situation formulation. Unpublished doctoral dissertation, University of Pennsylvania. Acar, W. Toward expanding dialectical inquiry: Design requirements of a comprehensive dialectical inquiry system. Proceedings of the Annual Meeting of The Decision Sciences Institute. Atlanta, GA: Decision Sciences Institute, 1987. Ackoff, R. L. Scient$c rrrethod. New York: Wiley, 1962. Ackoff, R. L. Creating the corporatefuture. New York: Wiley, 1981. Ackoff, R. L., & Emery, F. E. On purposeful systems. Chicago: Aldine-Atherton, 1972. Axelrod, R. (Ed.). Structure of decision. Princeton, NJ: Princeton University Press, 1976. Bannister, D., & Fransella, F. Inquiring nran: The theory ofpersonal constructs. London: Penguin, 1971.

Cass, K., Heintz, T. J., & Kaiser, K. M. An investigation of satisfaction when using a voicesynchronous GDSS in dispersed meetings. Inforrrration and Management, in press. Churchman, C. W. The design of inquiring systems. New York Basic Books, 1971. Conklin, J. A hypertext: An introduction and survey. Computer, September 1987, pp. 17-41. Cox, B., & Hunt, B. Objects, icons, and software-IC's. BYTE, August 1986, pp. 161-176. DeSanctis. G., & Gallupe, R. B. A foundation for group decision support systems design. Management Science, 1987, 33(5), 589-609. [131 Diffenbach, J. Influence diagrams for complex strategic issues. Strategic Managenrent Journal, 1982, 3, 133-146. [I41 Easton, A., Vogel, D. R., & Nunamaker, J. F. Stakeholder identification and assumption surfacing in small groups: An experimental study. Proceedings of the Twenty-Second Annual Hawaii International Conference on Systettr Sciences, 1989. [I51 Eden, C. Operational research and organization development. Hurrran Relations, 1978, 31(S), 657-674. [16] Eden, C. Using cognitive mapping for strategic options development and analysis (SODA). In

J. Rosenhead (Ed.), Rational analysis for a problenrotic world: Problem structuring rtrethodsfor cornplexity, uncertainty, and conflict. Winchester, UK: John Wiley, 1989. [17] Eden, C., & Radford, J. Tackling of strategic problerns, the role of group decision srrpport. London: Sage, 1990. [18] Gallupe, R. B., DeSanctis, G. L., & Dickson, G. W. Computer-based support for group problem finding: An experimental investigation. MIS Quarterly, 1988, 12(2), 277-296. [19] Heintz, T. 1. An object-oriented architecture for the design of group decision support systems. Paper presented at the National TIMSlORSA Conference, Philadelphia, PA, October 1990.

1230

Decision Sciences

[Vol. 23

[20] Jarke, M. Knowledge sharing and negotiation support in multiperson decision support systems. Decision Support Systems, 1986, 2(2), 92- 102. [21] Jarvenpaa, S. L., Rao, V. S., & Huber. G.P. Computer support for meetings of groups working on unstructured problems: A field experiment. MIS Quarferly, 1988, 12(4). 645-665. [22] Jones, S., & Eden. C. O.R.in the community. Journal of the Operational Research Sociery, 1981, 32, 335-345. [23] Joyner, R., & Tunstall, K. Computer augmented organizational problem solving. Management Science, 1970. 17(4), B212-B225. [24] Lai, K.. Malone, T. W., & Yu, K. Object lens: A 'spreadsheet' for cooperative work. ACM Transactions on Office Inforination Systems, 1988, 6(4), 338-353. [25] Lind, M. R., & Zmud. R. W. The influence of a convergence in understanding between technology providers and users on information technology innovativeness. Organizational Science, 1991, 2(2). 195-217. [26] Malone, T. W., Grant, K. R.. Turbank, F. A., Brobst. S. A,, & Cohen, M. D. Intelligent infomation sharing systems. Communications of the ACM, 1987, 30(5), 390-402. [27] Maruyama, M. The second cybernetics: Deviation-amplifying mutual causal processes. American Scientist, 1963, 51, 164-179; 250-256. [28] Mason, R. 0. A dialectical approach to strategic planning. Managerrrent Science, 1969. 15(8), B403-B4 14. [29] Mason, R. 0..& Mitroff, 1. 1. Challenging strategic planning assumptions. New York: Wiley, 1981. [30] Mitroff, I. I., & Emshoff, J. R. On strategic assumption-making: A dialectical approach to policy and planning. Academy of Managenrent Review, 1979. 4( I), 1-12. [31] Mitroff, I. 1.. Mason, R. O., & Barabba, V. P. Policy as arguments-A logic for ill-structured decision problems. Management Science, 1982, 28( 12), 1391- 1404. [32] Nunamaker, J. F., Applegate, L. M., & Konsynski, B. R. Facilitating group creativity with GDSS. Journal of Management Inforination Systeins, 1987, 3(4), 5-19. [33] Pinsonneault, A,, & Kraemer, K. L. The effects of electronic meetings on group processes. European Journal of Operational Research, 1990, 46. 143-161. [34] Riesbeck, C. K., & Schank, R. C. Inside case-based reasoning. Hillsdale, NJ: Lawrence Erlbaum Associates, 1989. [35] Smalltalk Vflindows, tutorial and programnting handbook. Los Angeles, CA: Digitalk Inc., 1991.

[36] Stefik, M. Planning with constraints (MOLGEN: Part I). Artificial Intelligence, 1981, 16(2), 1 1 1-140. [37] Sycara-Cyranski, K . Problem reformulation during negotiations. Paper presented at ORSA/TIMS Joint National Meeting, New York, October 1989. [38] Thomas, D. What's in an object? BYTE, March 1989, pp. 231-240. [39] Zigurs. I. Interaction analysis in GDSS research: Description of an experience and some recommendations. Decision Support Systeins, 1989, 5(2), 233-241.

Timothy J. Heintz is Associate Professor of Business Administration at Marquette University. He holds a D.B.A in quantitative business analysis from Indiana University with minors in management information systems and operations management. His current research interests are in the application of techniques from the field of artificial intelligence to business problems. He has been focusing on applications that involve planning, coordination and communication among groups of individuals in organizations. He has published in The Transportation Research Record, The Canadian Journal of Operational Research and Itforitration Systems, The Journal of Experiential Learning and Siitrulation, Infortiintion orrd Managenrent, and others. Dr. Heintz is currently a member of The Institute of Management Sciences and the Decision Sciences Institute. William Acar is Associate Professor of Management at Kent State University. He received his Ph.D. from the Wharton School of the University of Pennsylvania. Dr. Acar has presented papers at meetings of TIMS, the Academy of Management, the Decision Sciences Institute, the Southern Management Association, the ISSS, the ASAC, and at research workshops. He has developed a method for causal mapping and has published in journals such as the Journal of Manageirrenf,System Research, Behavioral Science, Interfaces. INFOR. and the Journal of Enterprise Management.