Proposition 2.3, proof — (2.2) holds iff it holds conditionally on each possible suggestion
ProvedAumann1987.TwoPerson.devCond_iff_condPayoffcorrelated-equilibriumgame-theoryp2o-batch-pfp1ap2o-gran-per-chapterp2o-plan-paperp2o-v1
Let be a distribution on and let be the payoffs. A suggestion to player 1 is possible if , and then is player 1's conditional expected payoff from playing ; likewise for a possible suggestion to player 2 (). Then:
- player 1's equilibrium condition (2.2) written on (no increases above ) holds if and only if
- player 2's condition holds if and only if for every possible and every .
This is the first step of the proof of Proposition 2.3: the equilibrium condition is equivalent to obedience being optimal after each suggestion that occurs with positive probability.
Preamble
import Mathlib import Definitions.Def_Aumann1987_TwoPerson_Model open Finset
Formal statement
namespace Aumann1987.TwoPerson
/-- **Conditioning on possible suggestions** (Aumann 1987, Proposition 2.3, proof, p. 6, PDF p. 7:
"For (2.2) it is necessary and sufficient that for each i, the expectation still obeys the
inequality when conditioned on each possible value of fⁱ"). For a distribution `p`:
player 1's condition (2.2) on `p` holds iff for every possible suggestion `j` (`∑_k p_{jk} > 0`)
and every action `q`, `H¹(q|j) ≤ H¹(j|j)`; and player 2's holds iff for every possible
suggestion `k` (`∑_j p_{jk} > 0`) and every `r`, `H²(r|k) ≤ H²(k|k)`.
Formalization Note: impossible suggestions are excluded by the hypothesis `0 < ∑ …`, as in the
paper, so the convention `x / 0 = 0` in `condPayoff₁`, `condPayoff₂` is never used. -/
theorem devCond_iff_condPayoff {S₁ S₂ : Type*} [Fintype S₁] [Fintype S₂]
(h₁ h₂ : S₁ → S₂ → ℝ) (p : S₁ → S₂ → ℝ) (hp : IsDistribution p) :
(DevCond₁ h₁ p ↔
∀ j : S₁, 0 < ∑ k, p j k → ∀ q : S₁, condPayoff₁ h₁ p q j ≤ condPayoff₁ h₁ p j j) ∧
(DevCond₂ h₂ p ↔
∀ k : S₂, 0 < ∑ j, p j k → ∀ r : S₂, condPayoff₂ h₂ p r k ≤ condPayoff₂ h₂ p k k) := by sorry
end Aumann1987.TwoPerson
Source
Aumann, Correlated Equilibrium as an Expression of Bayesian Rationality, Econometrica 55 (1987), DOI 10.2307/1911154, p. 6 (PDF p. 7), Proposition 2.3, proof (conditioning on each possible suggestion)
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.