A model of reasoning about knowledge (Q1286688)

Historically, modal logic was introduced to describe the modalities ``necessary'' and ``possible''. Later on, it turned out that the same formalisms (with a similar semantics) can be used to describe a different pair of modalities corresponding to knowledge and belief. Traditional modal logics, however, describe knowledge and belief of a \textit{single} intelligent agent. To describe knowledge adequately, we must take into consideration that different agents may know different things and believe in different things. To describe the reasoning about such multi-agent knowledge, researchers from Artificial Intelligence (AI) have formulated modal logics with several modalities corresponding to different agents; for a latest survey of their formalisms, see, e.g., \textit{R. Fagin, J. Halpern, Y. Moses} and \textit{M. Y. Vardi}, Reasoning about knowledge (1995; Zbl 0839.68095). All these formalisms, however, deal with the situation in which the knowledge and belief sets of different agents may be different, but the language which different agents use to describe their knowledge and belief is the same. Of course, we can easily assume that the agents' languages may be different, but we can always translate each statement from every language to every other language. In practice, some agents may use languages in which some statements cannot be adequately translated into the language used by the others; e.g., one of the agents may use a primitive language while another agent may be a highly intelligent agent using a universal language which cannot be adequately translated into the primitive language of the first agent. The authors generalize the AI's multi-modal formalism so that it covers such non-translatability as well. They describe the Kripke semantics for this new logic, and provide a system of axioms which they show to be complete with respect to this semantics.