The use of an expert system to simulate group decision-making in a stochastic environment

This thesis presents an expert system model developed to simulate group decision makers engaged in a strategy and policy making task whereby, on a quarterly basis, they evaluate strategic scenarios and then execute a set of interrelated decisions. Can the expert system model make meaningful decisions? Can the expert system "learn" from previously executed decisions? Can the expert system compete? Answers to these questions are formulated through the specification of an operationally deed expert system model based on the decision making processes of a panel of experts.;A "management game" computer program implemented in an undergraduate university is used as the setting for the strategic task. The game environment offers the opportunity to test the expert system model in a stochastic decision making environment and at the same time to have some control over the limits of the experimental domain. In addition, the management game incorporates a number of hypothetical relationships that represent a conceptualized view of how a company and its environment might behave.;The development process of the expert system simulation of group decision makers was pursued by: (1) Using a group of management faculty members who are both game administrators and real world business consultants as the panel of experts. They were given a set of thirty possible variables and relationships that are indigenous to the game. They then rated the importance of the variables, listed the interdependences and relationships affecting these variables, and estimated the likelihood that the relationships would influence different decisions and chose a fixed strategy. (2) Incorporating the responses from the expert panel into the knowledge base of the expert system model along with decisions and results for a hypothetical company during four years of operation; a standard synthetic history included in the management game by its authors. (3) Validating and then testing the decision making performance of the expert system in a "live" environment as it competes with five other student teams of decision makers while playing the management game extending over a simulated three year period.;This research, as a first step, will attempt to give some insight into the process of group decision making and to present an operational basis for further research.