On the soundness and safety of expert systems.

Abstract

The problems of developing sound and safe expert systems are discussed, with particular reference to medicine. The concepts, notations, methods, results and technologies which have emerged from the study of mathematical logic as a computational paradigm offer many benefits for improving the quality of expert systems. Logic programming offers a better discipline for design, specification and implementation than ad hoc development methodologies. When logic programming is combined with software engineering methods, such as a software development life-cycle, the probability of routinely developing large-scale yet efficient and sound applications will be increased. However, although soundness is a necessary property of any technology it is not sufficient for assuring safety. Established methods for improved software safety are discussed, and a number of approaches to improving the safety of medical expert systems is identified. The possibility of introducing an appropriately extended life-cycle, and the potential benefits of a formal theory of safety are discussed.