Main Theorem and converse — c.e.d.s are exactly the action distributions of Bayes-rational information systems
ProvedAumann1987.BayesRational.ced_iff_bayes_rational_systemLet be an -person game with action sets and payoffs , and let be a real function on action -tuples . Then is a correlated equilibrium distribution of if and only if there is an information system — a finite set of states, a common prior on , an information partition of for each player, and action functions measurable with respect to — in which every player is Bayes rational at every state and
This is the two-sided form announced in the paper's introduction: if each player maximizes expected utility given his information at every state, under a common prior, the chosen actions form a correlated equilibrium, and conversely every correlated equilibrium arises in this way.
Formalization Note. State spaces and probability spaces range over Type (universe 0), which loses nothing for finite sets. Bayes rationality at a state whose information cell has probability is vacuous ( in Lean); the prior need not have full support.
import Mathlib import Definitions.Def_agt_games import Definitions.Def_Aumann1987_BayesRational_CorrelatedEquilibrium import Definitions.Def_Aumann1987_BayesRational_InformationSystem
namespace Aumann1987.BayesRational
/-- **Main Theorem and its converse** (Aumann 1987, Econometrica 55, Main Theorem, Sect. 3, p. 7,
PDF p. 8, and its converse, Sect. 4d, p. 11, PDF p. 12; announced in Sect. 1, p. 2, PDF p. 3):
"Thus under Bayesian rationality, the set of all information systems corresponds precisely to the
set of all correlated equilibria."
A function `Q` on action `n`-tuples is a correlated equilibrium distribution of the game with
payoffs `h` if and only if there is an information system — a finite set `Ω` of states, a common
prior `p` on `Ω`, an information partition `𝒫ⁱ` for each player, and action functions `sⁱ`
measurable w.r.t. `𝒫ⁱ` — in which every player is Bayes rational at every state and the
distribution of the action `n`-tuple `s` is `Q`.
**Formalization Note.** Finite sets of states and finite probability spaces range over `Type`
(universe 0), which loses nothing for finite sets. Bayes rationality at a state whose cell has
probability `0` is vacuous (`condExp` divides by `0` and returns `0`); the paper's argument
never uses such states. The prior is not required to have full support. -/
theorem ced_iff_bayes_rational_system {ι : Type*} [Fintype ι] [DecidableEq ι] {S : ι → Type*}
(h : ι → (∀ i, S i) → ℝ) (Q : (∀ i, S i) → ℝ) :
IsCED h Q ↔
∃ (Ω : Type) (_ : Fintype Ω) (I : InformationSystem Ω S),
(∀ i ω, IsBayesRationalAt h I.p I.P I.s i ω) ∧ distr I.p I.s = Q := by sorry
end Aumann1987.BayesRational
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.