Iterated Belief Revision

Iterated belief revision studies how rational agents should update their beliefs through multiple cycles of evidence acquisition, where each revision may depend on the history of previous revisions. The naive application of single-step AGM revision operators produces pathological results: an agent may oscillate between accepting and rejecting the same proposition, or fail to learn from accumulating evidence because each revision wipes the slate clean. The Darwiche-Pearl postulates and subsequent work by Nayak and others attempt to constrain iterated operators so that revision histories exhibit stability and responsiveness — but no consensus framework has emerged. The problem is deeply connected to belief revision, non-monotonic logic, and the question of whether rationality should be defined synchronically or diachronically. A full theory may require abandoning the assumption that the agent's epistemic state at any moment can be represented as a simple belief set, and instead modeling it as a dynamic epistemic structure that carries its own history as constitutive content.