§3 p.11 and proof of Proposition 1 p.26 — (3) is equivalent to , and
ProvedStrategicObfuscation.posterior_le_priorgame-theoryinformation-economicsp2o-gran-per-chapterp2o-plan-paperp2o-strengthenedp2o-v1patent-examinationperfect-bayesian-equilibrium
Let and let be an equilibrium at . Then , and
The right side is inequality (3): the benefit of funding an uncertified proposal exceeds its cost. The bound says that an uncertified report is bad news.
Formalization Note The statement is for ; at the report can be off path, and then is free.
Differences from the source:
- generalization: Assumption 1 with ∂²π/∂σ∂e < 0 for a twice differentiable π → IsDetection: monotonicity plus decreasing differences in σ, strict whenever π(σ,·) separates the two efforts; reason: the C² class satisfies it by integrating the mixed partial, and the paper's own example on p.14, which the strict form excludes, also satisfies it
- encoding: e(α, σ) written as a function → a fixed selection E with IsEffort π g E; reason: the argmax in (1) may not be a singleton; every selection is covered and existence of a maximizer stays part of the hypothesis, as in (1)
- encoding: Bayes' rule (2) as a quotient → (2) cross-multiplied in IsEquilibrium; reason: the same equation whenever the denominator is nonzero; at σ = 1, e = 0 the report is off path and a PBE leaves Q free, where Lean's 0/0 = 0 would pin it
- reading: (3) ⇔ QV > c and Q ≤ p, for all σ → stated for σ < 1; reason: at σ = 1 with e = 0 the report is off path and Q is free, so neither claim can hold; the paper's Q* is only defined by (2) when its denominator is positive
Preamble
import Mathlib import Definitions.Def_StrategicObfuscation_IsDetection import Definitions.Def_StrategicObfuscation_IsEffort import Definitions.Def_StrategicObfuscation_IsEquilibrium open Set
Formal statement
namespace StrategicObfuscation
/-- **`Q* ≤ p`, and inequality (3) is equivalent to `Q* V > c`.**
Figueroa & Lemus, *The Benefit of the Doubt: Decision Making under Strategic Obfuscation*,
SSRN 3840423 (version of April 4, 2022), §3, eq. (3), p.11 (PDF p.11), and proof of Proposition 1,
a.1, p.26 (PDF p.26): "The benefit is larger than the cost if and only if
p(V − c)[1 − π(σ, 0)] > (1 − p)c[1 − π(σ, e∗)], (3) which is equivalent to Q∗V > c."
**Formalization Note** Stated for `σ < 1`. At `σ = 1` the report can be off path, and then `Q*` is
free. -/
theorem posterior_le_prior {V c g p : ℝ} {π : ℝ → ℝ → ℝ} {E : ℝ → ℝ → ℝ}
(hc : 0 < c) (hcV : c < V) (hg : 0 < g) (hp0 : 0 ≤ p) (hp1 : p ≤ 1)
(hπ : IsDetection π) (hE : IsEffort π g E)
{σ e α Q : ℝ} (hσ : σ ∈ Ico (0 : ℝ) 1) (heq : IsEquilibrium π V c p E σ e α Q) :
Q ≤ p ∧
(c < Q * V ↔ (1 - p) * c * (1 - π σ e) < p * (V - c) * (1 - π σ 0)) := by sorry
end StrategicObfuscation
Source
Figueroa & Lemus, The Benefit of the Doubt: Decision Making under Strategic Obfuscation, SSRN 3840423 (version of April 4, 2022), eq. (3), p.11, and proof of Proposition 1, a.1, p.26
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.