DECISION MAKING AND COGNITION: AN ARTIFICIAL INTELLIGENCE PERSPECTIVE (PRODUCTION SYSTEMS, EXPERT SYSTEMS).

摘要

Psychologists have long been interested in how people make decisions. Recent work in artificial intelligence (AI) is also addressing decision making. Computer programs, known as expert systems, are being developed to make decisions and solve problems at the level of human experts. Although the goal of expert systems is to achieve a high level of performance, the process of representing and applying decision-relevant knowledge in a computer program may offer insight into the underlying mechanisms of human decision processes.;At a more general level, this research is viewed as an example of interdisciplinary work in cognitive science. It is argued that representational formalisms used to build expert systems offer a useful perspective for viewing human decision processes. Similarly, an understanding of human decision making will prove useful in developing artificially intelligent decision systems.;A theoretical perspective of human decision making is presented that emphasizes knowledge-based deciding and equates decision processes with the control structures of production systems. A computer program, ACM, has been developed to investigate knowledge-based deciding. The task of ACM is to select maneuvers in an air-to-air combat environment. ACM represents and applies knowledge of air-combat maneuvering within a production system architecture. ACM was developed with two major goals in mind: (1) to select maneuvers at a level of performance approaching that of expert fighter pilots, and (2) to investigate within a production system architecture issues relevant to human decision processes. The first of these goals was clearly met. ACM's maneuver selections agreed well with those of actual pilots. With respect to the second goal, part of the decision processes of selecting maneuvers was implemented within a conflict resolution strategy. Some support was obtained for the psychological validity of this strategy.