Belief revision in a framework for default reasoning

Abstract

We present an approach which provides a common solution to problems recently addressed in two different research areas: nonmonotonic reasoning and theory revision. We define a framework for default reasoning based on the notion of preferred maximal consistent subsets of the premises. Contrary to other formalizations of default reasoning, this framework is able to handle also unanticipated inconsistencies. This makes it possible to handle revisions of default theories by simply adding the new information. Contractions require the introduction of constraints, i.e. formulas used to determine preferred maximal consistent subsets but not used to determine the derivable formulas. Both, revisions and contractions, are totally incremental, i.e. old information is never forgotten and may be recovered after additional changes. Moreover, the order of changes is — in a certain sense — unimportant.