Introduction: Workshop on Microanalytic Simulation and Behavior Theory

Citation:

Michael L. Ray and Clark Leavitt (1974) ,"Introduction: Workshop on Microanalytic Simulation and Behavior Theory", in NA - Advances in Consumer Research Volume 01, eds. Scott Ward and Peter Wright, Ann Abor, MI : Association for Consumer Research, Pages: 242-246.

Advances in Consumer Research Volume 1, 1974    Pages 242-246

INTRODUCTION: WORKSHOP ON MICROANALYTIC SIMULATION AND BEHAVIOR THEORY

Michael L. Ray, Stanford University

Clark Leavitt, Ohio State University

This workshop brought together people who have had experience modeling individual consumer behavior (either in flow chart or computer simulation form), developing simulation for applied purposes, measuring the steps of consumer decision making, or using behavior theory concepts in simulation work.

Microanalytic simulation is defined here in terms of attempts to model the decision process of individual consumers rather than aggregates. In recent years it has been used as the basis of models of consumer behavior, in applied settings, as a teaching device, to analyze data, to represent communication response, and as a computer representation of the active recent work on decision nets and processes in consumer behavior.

These simulations are becoming more and more representative, in whole and in part, of behavior theory itself. Built into their structures are concepts from balance, attribution and curiosity theories, perception and learning theories, multiple attribute attitude-notions, lexicographic models, utility theory, and the like

The idea for this workshop came most directly out of a two-day workshop on consumer information processing that took place prior to the ACR meetings at the University of Chicago in 1972 (Hughes and Ray, 1974). Discussions at that event revealed two general attacks on consumer information processing. Some researchers were concerned primarily with the structure of response, such as in attitude research. Others put more emphasis on causal links and the process of decision making itself. Although computer simulation itself was explicitly involved in only one paper in that workshop (Leavitt, 1974), microanalytic simulation was discussed as one tool to reconcile these two approaches. It seemed to be time to assess the role that such simulation might play in the representation of behavior theories and of consumer behavior itself.

A microanalytic simulation is "micro" because of its emphasis on individual behavior. It is "analytic" in the sense that it attempts to be explanatory or descriptive. There is an attempt to develop a causal structure for behavior. The word "simulation" is used in two ways. First a model is developed that is a simulation, representation or analogy of the hypothesized process of decision making. Second, computer simulation is often necessary to develop market forecasts of behavior.

Such simulation has been used in the behavioral sciences for some time. In fact Arnold Amstutz pointed out in his research statement for the workshop that attempts had been made in marketing at M.I.T. as early as 1958.

The most famous early example of the use of micro-analytic simulation was for the Kennedy campaign in 1960 (Pool,et al, 1965).Later simulation was used in advertising media modeling (Gensch, 1973) and in consumer behavior models (Amstutz, 1967, Nicosia, 1968). Since much of the early research was done in Massachusetts and at M.I.T., it seemed that Boston was an especially propitious place to hold this particular workshop. Many of the original researchers were participants in it.

PARTICIPANTS AND THEIR RESEARCH THRUSTS

Prior to the workshop each pre-registered participant received a set of three major papers and a number of one or two page research statements describing particular research thrusts. In all, 34 individuals prepared papers for the workshop.

The three major papers follow this introduction. They represent the old research axiom that "You must have method in order to do anything, theory to do anything interesting and application to do anything important."

Peter Wright of Illinois provided a central paper on method with "Research Orientations for Analyzing Consumer Judgment Processes." William T. Moran of Lever discussed the development of theory with simulation in his "CONSUMENOID and Methods of Psychology in Marketing." And the applied emphasis was generated by Gerald Eskin and M. Venkatesan of Iowa with their paper "A Micro-Analytic Simulation of Consumer Behavior: Economic and Behavior Theory Antecedents."

Other individual research statements could not be so easily classified into method, theory and application. These statements revealed the diversity of research in this area and the likely next steps it will take.

Because of space limitations it has not been possible to publish here all of the research statements that were received for the workshop. But they are reviewed below in terms of topics and trends. And the statements themselves can be obtained by sending three dollars to the first author of this introduction.

Management Science Applications

In addition to the major Eskin-Venkatesan paper there were four research statements which reflected actual application of varieties of microanalytic simulation to decision making situations.

The statement by Amstutz of M.I.T. has already been mentioned. He reviewed the value of these models and presented a general structure of market-oriented simulations. This structure included a communication process, a communication response process (similar to multiple attribute models), a brand choice process, a product response and two forms of feedback: use satisfaction and word of mouth.

John D. C. Little of M.I.T. is most representative of the "decision calculus" type of model (Little, 1970). His research statement described his BRANDAID model.

Both Dudley Ruch of Pillsbury and William Etter of General Mills described models that mirror the Amstutz and Little approaches. They also represent a main trend that seems to be developing in applied simulation. That trend is toward developing models that can deal with small, well-defined problem situations. Once the model is working it can be expanded to deal with new factors.

Stochastic, Econometric, Macro, Aggregate Approaches

A group of participants in the workshop represented something of a polar opposite to the main assumptions of the workshop. These individuals put less emphasis on micro and deterministic modeling and more on macro and stochastic approaches.

The line between these researchers and other researchers is often blurred, but it is quite sharp in the statement by Frank Bass of Purdue who says "Although it is heresy, in some circles, honesty compels one to question the fundamental premise that all behavior is caused."

Darryl G. Clarke of Harvard feels even micro models must be subjected to aggregate testing and development. Luis Dominquez of Indiana, on the other hand, questions the use of causal econometric approaches. Similarly, Tanniru Rao of Wisconsin has made some shift from macro, stochastic models to more micro-level attempts. He has found the shift to be frustrating.

It appears that stochastic components will be essential to simulation efforts in consumer behavior.

Behavior Theory Representation

Perhaps the group of papers and research statements best exemplifying the workshop title was that group which saw simulation as a tool to represent general behavior theory.

The major paper by Moran (and Alan Renda of Columbia) reviews an attempt by the ARF Theory and Models Committee to consolidate relevant behavior theories in a deterministic computer model of consumer behavior called CONSUMENOID.

Four research statements each emphasized particular theory applications. Richard Olshavsky of Indiana and Denis A. Lussier of Montreal suggested the use of the information processing approach of Newell and Simon (1972) in consumer behavior research. Nancy Harries of Michigan State discussed the territorial imperative (Ardrey, 1970), external-internal control (Rotter, 1971) and concrete-abstract thinking (Schroder, et al., 1967, 1973). David Wilson of Penn State covered certainty, risk perception and subjective expected utility models. Gerrit Wolf of Yale reviewed work in oligopoly and matrix gaming and considered how they might apply to consumer behavior.

It seems that such ideas from outside consumer behavior have potential application which might be facilitated by simulation.

Multiple Attribute Models and Multidimensional Scaling

In Amstutz's general simulation structure he included a level for the communication response. It is this level of most microanalytic simulations that best represents attitude structure. And attitude structure work in recent years has been primarily work on multiple attribute models and multidimensional scaling.

The former were represented in this workshop in research statements by Ollie Ahtola of Florida, Joel B. Cohen of National Analysts, Leon B. Kaplan of duPont, and Richard Lutz of UCLA. There appears to be some movement to broaden multiattribute models to consider the decision process more explicitly. Some broadening was exemplified in another statement by Roger M. Heeler and Harold Posner of York University.

Two other papers -- by Marshall Greenberg of National Analysts and Adrian Ryan of Western Ontario -- concentrated on multidimensional scaling. Both attempt to relate the scaling to choice behavior.

Specific Consumer Behavior Questions

The largest group of research statements related to specific consumer research questions. Peter Wright's major workshop paper, which attempts to review consumer research methods that might contribute to simulation, probably fits in this grouping better than any other.

The main topic of Wright's paper is the question of what research procedures can best be used to study consumer decision making. He starts from the basis of decision net analysis which James Bettman of UCLA has extensively pioneered in his research.

E. A. Pessemier of Purdue is interested in a variety of consumer research problems including the width and variety of assortments and the projection of laboratory results to the field. His colleague in the Purdue Psychology department, Jacob Jacoby, has done a series of laboratory simulations on consumer information overload.

Several researchers have done work on communication effects. Russell Haley of Haley, Overholser and Associates has participated in 88 major market studies of what segment characteristics predict response to communication. Michael Ray of Stanford and Alan Sawyer of Massachusetts have been investigating repetition response to communications in different situations. A great deal of simulation work with communication response has been done by Clark Leavitt of Ohio State and William McEwen of Connecticut. McEwen calls their work a "microsimulation of the persuasive effects of TV advertising."

DISCUSSION AND CONCLUSIONS

It is impossible to represent here all the discussion that took place over three hours among the nearly 70 people who attended the workshop. It is possible, however, to indicate the main conclusions we came to on the basis of that discussion.

First, large scale simulations of the original Amstutz-type never were valuable and could not be validated in their complete form. It seems more likely that progress in terms of application and theory development is likely to come with more limited models of the sort represented in the Eskin-Venkatesan paper. Amstutz did argue, quite forcefully, however, that the large scale model does provide a useful framework within which to develop more limited-objective simulations.

Second, it is quite difficult to test theory with simulations as they are now conceived. The development of the Eskin-Venkatesan model was totally atheoretical. Although it is possible to see several economic and behavior theory antecedents to that model, it is also possible that it could support several disparate views.

Third, the group was somewhat confused by the Moran-Renda attempt to develop a simulation entirely on a deterministic basis. The Moran paper in this publication clarifies many issues and clearly shows how such development might be possible.

Fourth, there seems to be a definite move toward studying process. This move comes out of both attitude research itself and out of attitude as represented in simulations. Wright's paper exemplifies this trend.

Fifth, simulation used alone is seen as a weak tool for studying behavior theory. It is in fact under-used at this point. Thus it should be used more in the future but primarily in combination with other methods. Wright puts emphasis on experimentation and mentions other possibilities.

REFERENCES

Amstutz, Arnold E. Computer Simulation of Competitive Market Response. Cambridge, Mass.: The M.I.T. Press, 1967.

Ardrey, Robert. The Territorial Imperative. New York: Atheneum, 1966.

Gensch, D. Mathematical Approaches to Media Planning. New York: Elvesier Publishing, 1973.

Hughes, G. D. & Ray, M. L., Buyer/Consumer Information Processing, Chapel Hill, N.C.: University of North Carolina Press, 1974.

Leavitt, C., in Hughes and Ray, 1974.

Little, J. D. C., "Models and Managers: The Concept of the Decision Calculus," Management Science, 1970.

Nicosia, Franco M. "Advertising Management, Consumer Behavior, and Simulation," Journal of Advertising Research, 9 (March 1968), 29-38.

Newell, Allen and Simon, Herbert, Human Problem Solving. New York: Prentice-Hall, 1972.

Pool, Ithiel de Sola, Abelson, Robert P., and Popkin, Samuel L., Candidates, Issues and Strategies: A Computer Simulation of the 1960 and 1964 Presidential Elections. Cambridge, Mass.: M.I.T. Press, 1965.

Rotter, J. B. "External Control and Internal Control." Psychology Today. (June, 1971), 37.

Schroder, H. M. et.al. Human Information Processing. New York: Holt, Rinehart and Winston, Inc., 1967.

Schroder, H. M. Education for Freedom. New York: John Wiley & Sons, Inc., 1973.

----------------------------------------