M4: Extensive-Form Games
Extensive-form games model strategic interaction that unfolds over time: game trees, information sets, and strategies capture sequential moves and imperfect information. We develop subgame perfect equilibrium and backward induction, then move to imperfect-information games via the sequence form and behavioral strategies. An LLM agent's observe → plan → implement → evaluate loop can induce an extensive-form game—but only once we fix what is externally observed, what remains private, and which nodes belong to chance or another strategic player.
Module preview
Extensive-form games model strategic interaction that unfolds over time: game trees, information sets, and strategies capture sequential moves and imperfect information. We develop subgame perfect equilibrium and backward induction, then move to imperfect-information games via the sequence form and behavioral strategies. An LLM agent's observe → plan → implement → evaluate loop can induce an extensive-form game—but only once we fix what is externally observed, what remains private, and which nodes belong to chance or another strategic player.
Lectures and materials
L10: Extensive-Form Games
Game trees, information sets, and strategies.
L11: Extensive-Form Games (cont.)
Subgame perfect equilibrium and backward induction.
L12: Extensive-Form Games (cont.)
Imperfect information, sequence form, and behavioral strategies.
Programming assignment
PA3: CFR and Kuhn Poker · Due November 9 at 11:59 PM CDT