AFTD Auto-Formalizing Theoretical Domains

bit_persuasion_bounded_support_ratio_le_three_halves

theoremverified

For every natural number m, every finite type ι with [DecidableEq ι], and every prior μ on ι → Bool that is a bit prior supported on at most 2m states (cardinality of {x | μ x ≠ 0} is at most 2m), persuasion_max_receiver_utility μ ≤ (3 / 2) * persuasion_prior_utility μ.

Statement

open Finset in
theorem bit_persuasion_bounded_support_ratio_le_three_halves
    (m : ℕ) (ι : Type*) [Fintype ι] [DecidableEq ι] (μ : (ι → Bool) → ℝ)
    (hμ : is_bit_prior μ) (hsupp : (univ.filter (fun x => μ x ≠ 0)).card ≤ 2 * m) :
    persuasion_max_receiver_utility μ ≤ 3 / 2 * persuasion_prior_utility μ
Source
arXiv:2606.22226
Verified
12 Nov 2026
Axioms
Classical.choiceQuot.soundpropext
Built on 4

Read back from the Lean

For every natural number m, every finite type ι with decidable equality, and every function μ : (ι → Bool) → ℝ, if: 1. μ is a bit prior (all values are non-negative and sum to 1), and 2. the support of μ (the number of states x : ι → Bool with μ x ≠ 0) has cardinality at most 2 * m, then the maximum receiver utility under any sender-optimal scheme for μ (persuasion_max_receiver_utility μ) is at most 3 / 2 times the receiver's prior utility (persuasion_prior_utility μ).

Written by a model that saw only the Lean, never the English above. If the two disagree, that is worth a challenge.

The proof, explained

The result follows directly from the general theorem `persuasion_max_receiver_utility_le`, which establishes that the maximum receiver utility is at most 3/2 times the prior utility for any bit prior. Although the theorem statement includes an assumption that the prior is supported on at most 2m states, this hypothesis is not needed, and applying the general theorem together with the bit-prior hypothesis immediately closes the goal.

  1. persuasion_max_receiver_utility μ ≤ (3 / 2) * persuasion_prior_utility μ

    The claim is established by applying the general theorem `persuasion_max_receiver_utility_le`. This reduces the goal to proving that μ is a bit prior, which holds immediately by the hypothesis `hμ`. The support-size bound `hsupp` is not needed.

    line 6

    uses persuasion_max_receiver_utility_le

        persuasion_max_receiver_utility μ ≤ 3 / 2 * persuasion_prior_utility μ := by first | exact persuasion_max_receiver_utility_le | (apply persuasion_max_receiver_utility_le <;> assumption) | (apply persuasion_max_receiver_utility_le <;> simp_all) | simpa using persuasion_max_receiver_utility_le | (simp only [persuasion_prior_utility, is_bit_prior, persuasion_max_receiver_utility] at *; first | exact persuasion_max_receiver_utility_le | (apply persuasion_max_receiver_utility_le <;> assumption) | (apply persuasion_max_receiver_utility_le <;> simp_all) | simpa using persuasion_max_receiver_utility_le)
persuasion_max_receiver_utility_le
For any bit prior μ on a finite type ι, the maximum receiver utility under persuasion is at most 3/2 times the prior utility.

Written by a model (gemini-3.8-flash) from the Lean proof and the goal states Lean reported at each step. Lean verified the proof; this explanation is a reading aid and is not itself machine-checked.

Lean source view module on GitHub

open Finset in/-- For any prior supported on at most 2m states, R_max(μ) is at most (3/2) R_0(μ). -/theorem bit_persuasion_bounded_support_ratio_le_three_halves    (m : ℕ) (ι : Type*) [Fintype ι] [DecidableEq ι] (μ : (ι → Bool) → ℝ)    (hμ : is_bit_prior μ) (hsupp : (univ.filter (fun x => μ x ≠ 0)).card ≤ 2 * m) :    persuasion_max_receiver_utility μ ≤ 3 / 2 * persuasion_prior_utility μ := by first | exact persuasion_max_receiver_utility_le | (apply persuasion_max_receiver_utility_le <;> assumption) | (apply persuasion_max_receiver_utility_le <;> simp_all) | simpa using persuasion_max_receiver_utility_le | (simp only [persuasion_prior_utility, is_bit_prior, persuasion_max_receiver_utility] at *; first | exact persuasion_max_receiver_utility_le | (apply persuasion_max_receiver_utility_le <;> assumption) | (apply persuasion_max_receiver_utility_le <;> simp_all) | simpa using persuasion_max_receiver_utility_le)
See it in the graphDiscuss this resultChallenge it