A Logic for Games with Imperfect Information

Speaker

Vaishak Belle
Title A Logic for Games with Imperfect Information
When 08.10.2008, 15.00 Uhr
Where Lecture Room Informatik 7
Abstract While the importance of imperfect information games has been long recognized, and efficient equilibria solvers proposed thus, an effective model to reason about knowledge and action in these rich games remains at large. More precisely, we want to reason that after Alice reads her card, Bob does not know what she has but he knows that she knows. A semantic model is introduced that establishes this distinction between de re and de dicto knowledge, treats in an intuitive manner the interplay between knowledge and action,

and

addresses only-knowing in games with imperfect information. Poker is selected for study, and it is shown here to bring out various properties of interest and sample multi-agent reasoning in a dynamic setting. The logic closely fashions a modal situation calculus variant, making action theories amenable to regression-based reasoning.