Temporal interaction of information and belief (Q2464646)

This paper is a sequel to [\textit{G. Bonanno}, Synthese 147, 193--228 (2005; Zbl 1085.03012)]. The author uses time-indexed modal belief operators \(B_t\) and time-indexed modal information operators \(I_t\) (\(t \in \mathbb N\)) to describe changes of beliefs of a single agent over time. The operators \(B_t\) are normal modal operators, whereas the semantics of the \(I_t\) operators requires that a formula \(I_t \varphi\) is true at a state \(w\) in a Kripke model iff the set of states related to \(w\) is the proposition expressed by \(\varphi\). Then four increasingly strong propositional logics are defined axiomatically: \({\mathbb L}\subset {\mathbb L}_{\text{QBR}}\subset {\mathbb L}_{\text{WAGM}}\subset {\mathbb L}_{\text{SAGM}}\). Some of the axioms make use of the universal modality. The logic \(\mathbb L\) is the logic of all models, and \({\mathbb L}_{\text{QBR}}\) is the logic of all models satisfying the ``Qualitative Bayes Rule''. \({\mathbb L}_{\text{WAGM}}\) is a logic that allows one to establish a connection with the AGM (Alchourrón, Gärdenfors, Makinson) postulates for belief revision. It is shown that, assuming the agent receives consistent information, on a given \({\mathbb L}_{\text{WAGM}}\)-model one can define belief sets and a revision operator satisfying the first six AGM postulates for belief revision.