Principal-agent problem: A cognitive map approach

Abstract

As software agents become more and more intelligent, it becomes more and more difficult for human principals to understand and control them. This is a well-known principal-agent problem. There is, thus, need for a tool that can bridge the gap between human principals and software agents. In this paper, we discuss a new approach based on cognitive map to understand and control the knowledge of software agents. We propose a hierarchical framework to construct cognitive map from the rule base of software agents with the help of some guidelines. We applied the cognitive map approach to the famous Pursuit problem and demonstrated how learning can takes place with the help of cognitive maps. In this study, we found that cognitive map could be a promising tool for understanding and controlling intelligent software agents. © 2002 Elsevier Science B.V. All rights reserved.