Formal specification of a rule-based expert system

Presents the initial steps in the design of a rule-based expert system. The purpose of the paper is to show how general, desirable properties of a knowledge-based system can be formalized and explored at an early stage, before details such as those of knowledge representation and uncertainty management cloud the picture. The method used is that of mathematical specification. First an abstract model of inference is presented, then the inference task is decomposed into component procedures. A simple but general theory of dependence between procedures is developed. This enables circularity in a knowledge-base to be avoided and consistency to be maintained. Several more concrete instantiations of the theory are given to illustrate the range of systems that it covers