r.h.s. of (A.2): the head's conditional expectation under factorizes over the independent components
ProvedIncentivesInTeams.Conglomerate.cond_expectation_factorizesConsider a conglomerate model whose component weights are probability weights, a joint strategy and a value of the head's information. Suppose the event has positive probability under the product law, where is the head's information under . Then for every subunit
and for the head's component
where the left-hand sides are the conditional expectations of (3.3) (elementary conditional averages under the product law) and , are the factors of defined in the model file: conditions only subunit 's state on its own message , and conditions only the head's state on .
This is the formal content of the paper's chain for the right-hand side of (A.2), whose last step "follows by the independence of the distributions of and , ". It shows that the factorized compensation of the mission equals the paper's (3.3) wherever (3.3) is defined.
Formalization Note. Positive probability of the conditioning event is the condition under which the elementary conditional expectation of (3.3) is defined; on a null event the Lean quotient is and the identity is not claimed. The head's component is included as the paper's bracket "[the proof for is strictly analogous]" indicates.
import Mathlib import Definitions.Def_IncentivesInTeams_Conglomerate_Model
namespace IncentivesInTeams.Conglomerate
theorem cond_expectation_factorizes {ι : Type*} [Fintype ι] [DecidableEq ι] {S₀ : Type*} [Fintype S₀] {S : ι → Type*}
[∀ i, Fintype (S i)] {Z₀ : Type*} {Z M₀ M : ι → Type*} {D₀ : Type*} {D : ι → Type*}
(T : Model S₀ S Z₀ Z M₀ M D₀ D) (hT : T.WeightsOK)
(βs : JointStrategy S₀ S Z₀ Z M₀ M D₀ D) (y₀ : Z₀ × ∀ k, M k)
(hpos : 0 < T.expect (Set.indicator {s | βs.headInfo s = y₀} (fun _ => (1 : ℝ)))) :
(∀ j, T.literalSubCondExp βs j y₀ = T.subFactor βs j y₀) ∧
T.literalHeadCondExp βs y₀ = T.headFactor βs y₀ := by sorry
end IncentivesInTeams.Conglomerate
Read-back
What the Lean code literally says, in plain math · claude-opus-5-5
Setting. is a finite set of subunits. and each are finite. A state is .
A joint strategy has:
- head functions , and ;
- subunit functions , and .
Its information maps are
The model supplies:
- weights and , assumed here to be probability weights (nonnegative, each summing to ), with product ;
- payoffs and .
For real weights , , with value when the denominator is .
Theorem. Fix and a value . Assume
Then, for every ,
Also,
No strategy-set membership is assumed.
Degenerate cases. The weight hypothesis forces and every to be nonempty. The positivity hypothesis is false, so the statement is vacuous, whenever is attained by no state of positive weight. If is empty, the first claim is vacuous and only the head identity is asserted.
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.