A Comparative Analysis of the Empirical Validity of Two Rule Based Belief Languages

Abstract

Rule based expert systems deal with inexact reasoning through a variety of quasi-probabilistic methods, including the widely used certainty factors (CF) and subjective Bayesian (SB) models, versions of which are implemented in many commercial expert system shells. Previous research established that under certain independence assumptions, SB and CF are ordinally compatible: when used to compute the beliefs in several hypotheses of interest under the same set of circumstances, the hypothesis that will attain the highest posterior probability will also attain the highest certainty factor, etc. This is very relevant to the expert systems field, where most inference engines and explanation facilities are designed to utilize the relative scales of posterior beliefs, making little or no use of their absolute magnitudes. The objective of this research is to explore empirically whether the compatibility of SB and CF extends to the field, where subjective degrees of belief and different elicitation procedures might bias the mathematical kinship of the two belief languages. In particular, we seek to know (i) whether this bias is random or systematic; and (ii) what the bias reveals about SB and CF as two alternative means to elicit and revise beliefs in a rule based system.