信念修正中的一种新型评价算子研究

本文考察了评价算子,即刻画认知主体如何借助评价依据来理性评价新信息的算子,它放弃新信息中部分不可信信息,但是与传统收缩算子不同,它不指定收缩的信息而只要求评价结果与评价依据相一致。S.O.Hansson认为以单个句子作为新信息的非优先信念修正具有三种模式:决策–修正(Decision-Revision)、整合选择(Integrated Choice)和膨胀–巩固(Expansion-Consolidation)。评价算子单独刻画了非优先复合修正决策–修正模式的决策模块,对于修正模块这里不作探讨。我们先按AGM信念修正模式对评价算子进行了公理性刻画,接着通过不一致核心集和一致保留集分别构造了核心评价算子和部分交评价算子,然后分别证明了这两种算子与公理性刻画之间的表达定理,最后对评价算子和其他相关工作进行了比较。

A New Evaluative Operator in Belief Revision

In this paper we explore evaluative operators, i.e., operators in which epistemic agent evaluates new information base on agent’s core beliefs. It deletes some informa-tion from new information, but different from traditional contraction, it doesn’t specify information to be deleted, but only makes the output consistent with core beliefs. S. O. Hansson thought that non-prioritized belief revision by one sentence has three kind of schemes: Decision-Revision, Integrated Choice and Expansion-Consolidation. Evalua-tive operators only characterize Decision module of Decision-Revision Scheme in non-prioritized multiple revision. First, we characterize this kind of operators by some pos-tulates in the way of AGM scheme, and then we construct kernel evaluative operator by inconsistent kernel set and partial meet evaluative operator by consistent remainder set, and then we prove representation theorem of these two functional constructions. At last we compare evaluative operators with related work in the literature.