Agreement theorems in dynamic-epistemic logic (Q692182)

This paper introduces the idea of agreement theorems, which originate with the work of Robert Aumann, and shows how they can be represented in a dynamic epistemic logic setting. Aumann's result was that, if agents have common prior beliefs, then differences in their posterior beliefs cannot be common knowledge. Since the concepts of belief and knowledge, and in particular higher-order belief and knowledge, are central to dynamic epistemic logic, it seems natural to represent agreement theorems in this context. The authors first show how an agreement theorem can be represented in the context of a static epistemic plausibility model. They prove the result that (in a well-founded model) a common belief that agents disagree on the truth of a formula implies that these agents have different prior beliefs. To express the concepts involved in the agreement theorems, however, a more expressive language than that of epistemic-doxastic logic is required. The authors show that epistemic-doxastic logic is insufficient for expressing well-foundedness and common priors, but that a hybrid logic does have sufficient expressive power. This enables them to give a syntactic proof of the agreement theorem in this latter setting. The last section of the paper considers the possibility of agents reaching agreement under different kinds of dynamic updates. They consider two different types of dynamic processes that enable agents to update their informational states: conditioning and public announcements. Both of these can be thought of as dialogues of a kind, in which agents exchange information about each others' beliefs. The main difference is in the publicity of those exchanges. In belief conditioning, agents revise their beliefs conditional upon what they learn about other agents' beliefs. But there is no assumption that agents are aware of each others' updating processes. On the other hand, in the case of public announcements, the assumption is that all agents in the group are aware that the information received is truthful, and received as such by all other agents. The consequence of this difference is that the truth value of a publicly announced formula can change its truth value as a result of the announcement. Thus even though both types of dialogue will eventually lead to a fixed point, the agreements reached may differ.