Bridging merging prioritized information in possibilistic logic and inconsistency handling methods

Abstract

Exception handling, belief revision and merging information coming from multiple sources are important problems in AI. These problems are related in the sense that they all rely on resolving conflicts. This paper investigates relationship between merging prioritized information, belief revision and inconsistency handling methods. Syntactic merging in possibilistic logic, which is semantically meaningful in the framework of possibility theory, is presented here as a process which combines prioritized knowledge bases into a new prioritized knowledge base. Several classes of merging operators are distinguished according to whether the bases are independent, conflicting, etc. We show that merging multiple sources information provides a compilation of the lexicographical approach, an inconsistency handling method known to have desirable properties and used in diagnosis problems. Surprisingly, we show that this compilation recovers also a strategy for belief revision called disjunctive adjustment.