Group Announcements: Logic and Games

Thomas Agotnes (University of Bergen) will give a talk on epistemix logic and game theory.

Everyone welcome.

Public announcement logic is used to reason about the epistemic pre- and post-conditions of actions in the form of public announcements. In the talk I will present Group Announcement Logic (GAL), an extension of public announcement logic with constructs (well known from coalition logic) of the form phi, where G is a group of agents. In GAL, the meaning of phi is that there exists an announcement that the members of G can jointly and truthfully make, and after that announcement is made public phi will be true. After introducing GAL I will discuss how it can be used to express properties such as "there is a sequence of truthful public announcements by agents in G, after which phi is true"; the distinction between "agent i knows *that* phi can be achieved by a public announcement" and "agent i knows *how* phi can be achieved by a public announcement"; and meta-logical properties such as axiomatisation, expressivity and the complexity of the model checking problem. In a second part of the talk I will also briefly discuss the question of which announcements the agents actually will make, assuming that they are rational. I consider situations where each agent has a goal in the form of a (typically epistemic) formula that she would like to become true (e.g., Bob learns my secret without Cat learning it). What will each agent announce, assuming common knowledge of the situation? The truth of the goal formula after the agent's announcement depends on the announcements made by other agents, hence we have a game theoretic scenario.