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

Rule based expert systems deal with inexact reasoning through avariety of quasi-probabilistic methods, including the widely usedcertainty factors (CF) and subjective Bayesian (SB) models, versionsof which are implemented in many commercial expert systemshells. Previous research established that under certain independenceassumptions, SB and CF are ordinally compatible:when used to compute the beliefs in several hypotheses of interestunder the same set of circumstances, the hypothesis that willattain the highest posterior probability will also attain the highestcertainty factor, etc. This is very relevant to the expert systemsfield, where most inference engines and explanation facilities aredesigned to utilize the relative scales of posterior beliefs, makinglittle or no use of their absolute magnitudes. The objective ofthis research is to explore empirically whether the compatibilityof SB and CF extends to the field, where subjective degrees ofbelief and different elicitation procedures might bias the mathematicalkinship of the two belief languages. In particular, we seekto know (i) whether this bias is random or systematic; and (ii)what the bias reveals about SB and CF as two alternative meansto elicit and revise beliefs in a rule based system