Game Theory studies rational decision making in situations of interdependent decisions, where the outcome of one's choice depends on what others decide. "Real" games like chess or Go are obvious examples, but game-theoretical models have proved useful to analyze a much broader range of phenomena, from bargaining situations, both by real and artificial agents, to conventions like driving behavior and language. In such situations, rational deliberation about what to do should take into account not only what one expects the others will do, but also what one believes about others' beliefs. Taking this intuition seriously is the trademark of contemporary epistemic game theory, a discipline that has by now a few decades of fruitful interaction between logic and economics on its record.
This course is a general introduction to epistemic game theory, with a strong accent on logical approaches to the discipline. We will start by introducing the decision-theoretic background, as well as the game-theoretical basics. We will then move to epistemic game theory proper, by presenting modern logical tools to represent information in interactive contexts, and looking in detail at the classic results in the field, both on so-called strategic form games, "matrices", and extensive form games, "trees". Towards the end of the course, we will connect with the more recent logical literature on information (dynamics), preferences and actions, showing that they offer a new perspective on the game-theoretic results.
The course should be of interest for students in philosophy, computer science (especially multi-agent systems) and linguistics (especially those interested in formal pragmatics). It will be self-contained, thus does not require previous knowledge of the logical or game- and decision-theoretical material that we will cover. We only assume a reasonable level of mathematical maturity.
We introduced the basic concepts in game and decision theory (strategic and extensive games, Nash equilibrium, iterated strict/weak dominance, maximizing expected utility).
- Eric Pacuit and Olivier Roy. Epistemic Game Theory, Stanford Encyclopedia of Philosophy (Sections 1 and 2)
- K.R. Apt (2011). A Primer on Strategic Games, in Lectures in Game Theory for Computer Scientists, Cambridge University Press, pgs. 1 - 33.
We started with a discussion of the relationship between dominance reasoning and maximizing expected utility. The main focus of the lecture was to introduce various mathematical models that describe the players' knowledge and beliefs.
- Eric Pacuit and Olivier Roy. Epistemic Game Theory, Stanford Encyclopedia of Philosophy (Sections 2 and 3)
We finished our discussion of epistemic notions (knowledge, belief, "safe" belief, common knowledge/belief). The main topic for today was a fundamental theorem of epistemic game theory: informally, assuming rationality and common belief of rationality implies that the players choose strategies that survive iterated removal of strictly dominated strategies.
- Eric Pacuit and Olivier Roy. Epistemic Game Theory, Stanford Encyclopedia of Philosophy (Section 5.1)
- Krzysztof R. Apt, Jonathan A. Zvesper, The role of monotonicity in the epistemic analysis of game theory, Games 2010, 1(4), 381-394.
We continue our discussion of epistemic characterizations of solutions concepts. The first part of the lecture focused on extensive games and backwards induction. The second part discussed the characterization of iterated weak dominance.
- Eric Pacuit and Olivier Roy. Epistemic Game Theory, Stanford Encyclopedia of Philosophy (Sections 6.1 and 6.2)
- Joe Halpern Substantive rationality and backward induction, Games and Economic Behavior 37, pp. 425-435.
A* dam Brandenburger, Amanda Friedenberg and H. Jerome Keisler, Admissibility in Games, Econometrica, Vol. 76, 2008, 307-352.
We conclude with a discussion of the Brandenburger-Keisler Paradox, Nash Equilibrium and normative vs. descriptive models of games.
- Eric Pacuit and Olivier Roy. Epistemic Game Theory, Stanford Encyclopedia of Philosophy (Section 7)
- S. Abramsky and J. Zvesper. From Lawvere to Brandenburger-Keisler: interactive forms of diagonalization and self-reference, Proceedings of LOFT 2010.