On the logic of theory change: Contraction without recovery (Q1387802)

Of the AGM postulates for belief contraction, the most controversial has been recovery: \(K\subseteq Cn(K- x\cup\{x\})\). The author's purpose is to define a class of operations on belief sets (understood to be closed under logical consequence) that fail recovery but satisfy the other basic AGM postulates for contraction. His starting point is the observation that an operation -- satisfies recovery iff for all \(K,x\): \(K-x\lvdash\neg x\vee y\) for all \(y\in K\setminus K-x\). His basic idea is to fail recovery by getting rid of \(\neg x\vee y\) for at least one such \(y\). This is done by intersecting \(K-x\) with \(K-(\neg x\vee y)\) for some \(y\) picked out by a choice function. The author shows that this operation, which he calls `semicontraction', does indeed fail recovery but satisfies the other five basic postulates, and shows how it is related to AGM revision.