The Deepening of Expert Systems

Article type: Research Article

Authors: Steels, Luc

Affiliations: Artificial Intelligence Laboratory, Free University of Brussels

Abstract: Since its conception in the early seventies through the work on DENDRAL (Feigenbaum and Lederman, 1970) and MYCIN (Shortliffe, 1976), the notion of an expert system has swept the world and captured the imagination of computer technologists and users alike. The success of the concept of an expert system can no longer be denied. It points out that there is a large demand in the computer user community for user-friendly, intelligent-looking computer systems and for systems that tackle problems requiring reasoning. After some hesitation we see solid projects emerging in many organisations and we already see a number of successful applications, even though it takes at least several man-years to develop a serious system. Despite this positive outlook, there are also some negative trends. The most troublesome one is that the notion of expert system is increasingly being trivialised leading to false user expectations and superficial results. The first part of the paper focuses on this trend. The second part reviews the counteracting trend towards “deeper” expert systems that utilise richer representations of the domain instead of the shallow heuristic rules which formed the exclusive basis of first generation expert systems.

DOI: 10.3233/AIC-1987-0104

Journal: AI Communications, vol. 0, no. 1, pp. 9-16, 1987