An analysis of four uncertainty calculi

Abstract

An important issue faced by contemporary artificial intelligence workers is how to deal with uncertain information. Four of the more prominent calculi-probability theory (especially the Bayesian approach), the Dempster-Shafer theory, fuzzy set theory, and the MYCIN and EMYCIN calculi-are examined. Particular attention is paid to the underlying assumptions of these calculi and to their computational complexities. Each of the four calculi has a different perspective in uncertainty, and each manipulates uncertain information in a different way. Despite what some authors have claimed, there does not seem to be one calculus that is the best for all situations. Each of the calculi has its strong points; the main disadvantage seen in all of the calculi is that they compute aggregate numbers, but keep no record of divergence in opinions.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>