Axiom P and Reliable Belief Change Operator in belief revision

摘要

There are often a very large number of beliefs in an agent. Generally, efficient belief changes should be performed only in the relevant part of its belief state at a time. On this condition Parikh formulated relevance criterion and axiom P: If T - Cn(A, B) where A,B are in L_1,L_2 respectively and C in L_1, then T * C = Cn(A) * C+B, where * is the update operator for the sub-language L_1. By the parallel interpolation theorem formulated by Kourousias and Makinson, we first show in the paper that axiom P holds iff belief set B is complete with respect to sub-language L(E(B)) when A(⊥)C. And we extend the axiom to an infinite language. Secondly we propose that there exists the unique maximal invariable letter set for any partial meet contraction owing to a theorem, the existence of the unique finest splitting of any set of formulae, formulate by Kourousias and Makinson in 2007. Consequently we formulate reliable belief change operation in belief change by maximal invariable letter set. Finally we apply this results to Multi-Agent Systems.