A comparison of two evidential reasoning schemes

摘要

Gordon and Shortliffe [2] advocate the use of Dempster-Shafer (D-S) theory in the evidence-gathering process. It is stated that they are unaware of any formal model which could allow inexact reasoning at whatever level of abstraction. Pearl [3] later shows how evidential reasoning can be conducted in the same hypothesis space using a Bayesian model. The purpose of this note is to examine the difference between these two schemes, and to point out certain inconsistencies of this Bayesian model with the motives behind the use of the D-S model.