On the combination of uncertain or imprecise pieces of information in rule-based systems—A discussion in the framework of possibility theory

AbstractThis paper is devoted to a general discussion of the combination of distinct imprecise or uncertain pieces of information pertaining to the same logical or numerical variable in rule-based expert systems using numerical approximate reasoning techniques. The reasons that the results of this combination may be meaningless are considered in detail. In particular, the presence of implicit default assumptions in the condition part of a rule is discussed and coped with. Moreover, the paper studies why the conclusions obtained by the most specific rule must be definitely preferred to conclusions derived using more general rules. Specificity ordering is defined within the framework of possibility theory