CERTAINTY FACTOR THEORY AND ITS IMPLEMENTATION IN AN INFORMATION SYSTEM-ORIENTED EXPERT SYSTEM SHELL

This article discusses some topics related to the MYCIN model for uncertainly handling in expert systems. Some previously identified problems in this model should be considered as evidence to support the point of view that the probabilistic interpretation for measure of belief (MB) and measure of disbelief (MD) given by Shortliffe is inappropriate. In connection to this, some results achieved by Heckerman are also reviewed and one of his arguments is questioned. According to this argument, the original interpretation for certainty factors implies a noncommutative method for evidence combination, and therefore, this interpretation should be abandoned. Problems similar to those identified by Adams in the original interpretation for certainty factors exist also in the interpretation recommended by Heckerman. Then, in the context of previous results stating that there is an isomorphic mapping from the evidence combination scheme for the likelihood ratio to the evidence combination function adopted in EMYCIN and its successors, we point out that the evidence combination scheme for the likelihood ratio suffers from a strong restriction. We remark upon the complaint that the MYCIN model cannot always function as well as one has expected. Implementation of the certainty factor mechanism in an information system-oriented expert system shell with some substantial changes is described.