Issues in Rule Base Development

Abstract

An expert system is an efficient human‐computer system designed to automatically produce decisions that normally require the application of both factual and heuristic information by a human expert. Expert systems do not add inherent validity to the decision‐making process. They currently provide, however, the avoidance of errors in solution procedures arising from inexperience, fatigue, overconfidence, and other human factors. Information in expert systems is normally represented by production rules. In the development of an expert system two primary concerns are the development of the set of rules and the logic by which the information is processed to obtain a decision. Several approaches to the elicitation of rules, the logical structuring of rules and uncertainty, and calibration of expert systems are reviewed. Decisions resulting from the application of expert systems are very dependent on the logic structure of the system used. Little evidence exists to guide expert system developers in the selection of these approaches.