MODELS FOR REASONING UNDER UNCERTAINTY

Abstract

As applied to expert systems, two models for reasoning under uncertainty (the well-known MYCIN model and a probability-based model) are described and compared. It is proven not only that the probabilistic assumptions for the probability-based model are weaker and there fore more intuitively appealing than those for MYCIN but also that, when two rules argue for the same conclusion, the combinatoric method in the probability-based model yields a higher combined certainty than that in the MYCIN model.