Common-Prior Posterior Agreement
Abstract
Commonly known posteriors from a positive finite common prior agree.
Theorem 1.1 (Commonly known posterior values agree).
Proof. Machine-checked in Lean as D5/S3/ConceptDynamics/ObservationOrder/CommonPriorPosteriorAgreement.common_knowledge_posteriors_agree (✓ std3). ∎
Source. Repository-derived.
Commentary.
The world type is finite and mu is a strictly positive normalized common prior. The event E and nonempty common-knowledge cell K are finite subsets of that exact world carrier.
Each information structure is a finite partition of K. This is the restriction of the agent’s information partition to the common-knowledge cell; closure of a common-knowledge cell makes every such part a whole information cell.
The posterior on a cell C is constructed as common-prior mass of E inside C divided by common-prior mass of C. Strict positivity and nonempty partition parts make every denominator positive.
Summing the constant posterior identity over either partition gives the same event mass on K: a times mu(K) for the first agent and b times mu(K) for the second. Since mu(K) is positive, a equals b.
Repository and pinned Mathlib searches found no exact common-prior agreement theorem. The proof directly applies Mathlib’s canonical finite-partition union and disjoint-sum machinery.
References
- Truth anchor:
D5/S3/ConceptDynamics/ObservationOrder/CommonPriorPosteriorAgreement.common_knowledge_posteriors_agree