Instructor: Sam Marks
Spring 2023
T/Th, 5:30-6:45pm
Office hours: Tuesdays 4:30-5:30pm or by appointment; in Science Center 431e
This course is an introduction to decision theory, covering topics like representation theorems, iteratated games, causal and evidential decision theories, and updateless policies.
1/24: Introduction: prisoner's dilemma, iterated games, and backwards induction (notes).
1/26: Introduction cont'd: games where agents predict each other's strategies; track record, trust, and precommitments (notes).
1/31: Preference modelling: weak orderings and money pump/Dutch book arguments (notes).
2/02: Preference modelling cont'd: utility functions, lotteries, expected utility, and the von Neumann-Morgenstern axioms (notes).
2/07: Preference modelling cont'd: more vNM, uniqueness of utility functions up to positive affine transformation, violations of independence (notes).
2/09: Preference modelling cont'd: proof of vNM representation theorem, introduction to Savage representation theorem (vNM notes, Savage notes).
2/14: Preference modelling cont'd: Savage axioms (notes). Classical game theory: two player games, mixed and pure strategies, Nash equilibria (section 4.2 in KP book).
2/16: Classical game theory: games in extensive form (notes, section 6.1 in KP book), iterated games (section 6.4 in KP book).
2/21: Classical game theory: evolutionarily stable equilibria (section 7.1 in KP book). Introduction to decision theory: overview of CDT and EDT (notes).
2/23: Introduction to Bayes nets (notes), Bayes' theorem (notes).
2/28: Conditional independence, DAGs, Markovian parents, and Markov compatibility (sections 1.1.5-1.2.2 of Causality).
3/02: d-separation, observational equivalence, inference in Bayes nets (sections 1.2.3-1.2.4 of Causality).
3/07: Efficient inference in Bayes nets, structure learning, Markov blankets (notes).
3/09: Some interesting games: ultimatum game, information cascade, dollar auction, 2/3 the average, Keynsian beauty contest.
3/21: The do operator and causal Bayes nets (section 1.3.1 of Causality, notes). Causal and evidential decision theories.
3/23: CDT and EDT: examples, smoking lesion (page 2 here).
3/28: Analysis of examples: psychopath button and Newcomb's problem (notes).
3/30: Updatelessness and analysis of XOR blackmail (notes).
4/04: Stable matching and Gale-Shapley (section 10.2 in KP book). Introduction to auctions (notes).
4/06: More auction theory: uniform private values, revenue equivalence (section 14.2 in KP book). Cake cutting (section 11.1 in KP book).
4/11: Predictions, scoring rules, and prediction markets (reference).
4/13: Sponsored search auctions: GSP (reference) and VCG (notes) mechanisms.