Probabilistic dynamic belief revision (Q1024132)

In this dense paper, the authors present an expressive semantic framework for the logic of belief change, combining three well-known approaches to the subject: (1) a probabilistic one using Popper-Rényi-style conditional probability functions, (2) a qualitative one using AGM structures, here formulated in terms of plausibility rankings of states, and (3) a modal one using iterable belief and knowledge operators in the object language. To some extent, the literature has already combined these pairwise, but not as a threesome. By employing probabilities (rather than merely binary discriminations) one allows levels of belief less than unqualified acceptance. By working with Popper-Rényi conditional probabilities (rather than standard ones defined as ratios of one-place probabilities), one can identify unqualified belief with having for the agent subjective probability 1 without thereby making proper belief revision (introducing something inconsistent with what was previously believed) impossible or trivial, thus opening the way for a connection between conditionalization on probability functions and AGM qualitative revision. By employing iterable epistemic operators in the object language, one can model reflexive and inter-agent beliefs up to common knowledge. All these three aspects are studied in the context of the authors' comprehensive triple framework.