Possibilities versus Fuzzy Probabilities--Two Alternative Decision Aids.

Abstract : In this paper we look at the two major approaches that have been proposed using fuzzy set theory to help in the analysis of decision trees, one based on possibility theory, and one on a fuzzy extension of probability theory. We conclude that the first approach requires an operational definition of possibility in order to be sufficiently compelling to justify its use in preference to standard decision analysis. We suggest that this work should prompt us to consider the potential of using different connectives in different decision contexts. It is stressed that the second approach is an extension of the standard paradigm, rather than an alternative, as has sometimes been assumed, and we suggest that it is more promising than the first approach. We show how the extended theory may be used to help understand and deal with inconsistent probability (or utility) assessments, and we introduce the concept of the 'value of perfect coherence' as an aid to calculate the value of a decision analysis.