Theorem 2 — in pride raises the optimal profit, guilt lowers it iff (corrected: demand )
ProvedPrideGuilt.theorem_2Fix every parameter except the social pride and guilt , and let be the manufacturer's optimal profit over prices and rewards with nonnegative demand. Assume the standing assumptions, , and , and write . Then has both partial derivatives at , and:
- ;
- if , and if ; moreover holds iff ;
- if , and if ; moreover holds iff .
So in the region of moderately high difficulty of remanufacturing where pride outweighs guilt, stronger pride never hurts the manufacturer, while stronger guilt hurts exactly when the optimal recycling rate is below , and guilt helps more than pride exactly when that rate exceeds .
Formalization Note The partial derivatives are the derivatives of the one-variable sections at and at ; their existence is part of the conclusion. Each "i.e." equivalence is stated once as an iff; its negation is the "i.e." of the second line.
Differences from the source:
- correction: the optimum of (5) (and Π⋆(f) = max_p Π(p,f)) is an unconstrained max over (p, f) → supremum over (p, f) ∈ ℝ² with demand D(p, f) ≥ 0; reason: as printed the max counts a negative demand times a negative margin as profit: it is +∞ for unrestricted f, and with b0 = 1, b1 = 0.5, cn = 1, cu = 0.5, ce = 3, IP = IG = 0, γP = 4, γG = 1 (a point of Ω4) the max over f ≤ c_r is 1.5625 at demand −1.25, independent of γG, contradicting Theorem 2(ii); the paper's footnote 4 (p.13) and p.12 require feasible demand
- reading: max over p, f with no stated domain → p and f range over all of ℝ; reason: (5) puts no bound on p or f; f* is negative at some points of Ω4 (−0.5 in the example above), so a sign restriction would falsify Proposition 5
- generalization: the discussion of §4 is restricted to the part of Ω4 where p* ≥ 0 (p.13) → all of Ω4, no p* ≥ 0 hypothesis; reason: Theorem 2 holds on all of Ω4 (Π* = K²/(4b0) with K = 1 − b1IG − b0cn + b0a²/(4s) > 0, a = γP − (c_r − Δc) − (b1/b0)γG > 0, s = γP − γG > 0); it specializes to the source's statement on the p* ≥ 0 part
- encoding: ∂Π*/∂γP and ∂Π*/∂γG → deriv of the one-variable sections x ↦ Π*(x, γG) and y ↦ Π*(γP, y), with their differentiability asserted; reason: partial derivatives of a function of two variables; differentiability rules out Lean's junk value deriv = 0
- encoding: τ*(f), the evolutionarily stable rate of Lemma 1 → the Lemma 1 formula tauStar, including its printed values at the endpoints and at τ^0 = τ_m(f); reason: the manufacturer's problem is built on the Lemma 1 formula (p.11); in Ω4 the endpoint values are the limits of τ_m, so the supremum does not depend on them
import Mathlib import Definitions.Def_PrideGuilt_Standing import Definitions.Def_PrideGuilt_optProfit import Definitions.Def_PrideGuilt_InOmega4 import Definitions.Def_PrideGuilt_tauMStar
namespace PrideGuilt
/-- **Theorem 2 (impact of pride and guilt on the optimal profit in `Ω₄`), corrected: demand
`D ≥ 0` in (5).**
Huang–Nguyen–Tseng–Chen–Kirshner, *Pride or Guilt?*, M&SOM 26(6) (2024), doi:10.1287/msom.2023.0721, Theorem 2, p.18 (PDF 19).
Fix every parameter except `γ_P, γ_G`, assume the standing assumptions, `γ_P, γ_G > 0` and
`(γ_P, γ_G) ∈ Ω₄`, and write `d = c_r − Δc`. The optimal profit `Π*(γ_P, γ_G)` has partial
derivatives in `γ_P` and in `γ_G` at this point, and
(i) `∂Π*/∂γ_P ≥ 0`;
(ii) `∂Π*/∂γ_G < 0` if `(b₀ − 2b₁)γ_P + b₁γ_G < b₀d`, and `∂Π*/∂γ_G ≥ 0` if
`(b₀ − 2b₁)γ_P + b₁γ_G ≥ b₀d`; the first condition holds iff `τ*_m < b₁/b₀`;
(iii) `∂Π*/∂γ_P ≥ ∂Π*/∂γ_G` if `b₀γ_G − b₁γ_P ≤ b₀d`, and `∂Π*/∂γ_P < ∂Π*/∂γ_G` if
`b₀γ_G − b₁γ_P > b₀d`; the first condition holds iff `τ*_m ≤ (b₀ + b₁)/(2b₀)`.
**Formalization Note.** `Π*` is `optProfit`, the supremum of `Π(p, f)` over `(p, f) ∈ ℝ²` with
`D(p, f) ≥ 0` (correction; as printed, (5) is maximized at a negative demand and the theorem fails
at `b₀ = 1`, `b₁ = 0.5`, `c_n = 1`, `c_u = 0.5`, `c_e = 3`, `I_P = I_G = 0`, `γ_P = 4`, `γ_G = 1`).
The partial derivatives are `deriv` of the one-variable sections `x ↦ Π*(x, γ_G)` and
`y ↦ Π*(γ_P, y)`, whose differentiability is part of the conclusion. Each "i.e." equivalence is
one `↔`; its negation is the "i.e." of the second line. No `p* ≥ 0` restriction is assumed. -/
theorem theorem_2 (P : Params) (hP : Standing P) (γP γG : ℝ) (hγP : 0 < γP) (hγG : 0 < γG)
(hΩ : InOmega4 P γP γG) :
DifferentiableAt ℝ (fun x => optProfit P x γG) γP ∧
DifferentiableAt ℝ (fun y => optProfit P γP y) γG ∧
0 ≤ deriv (fun x => optProfit P x γG) γP ∧
((P.b0 - 2 * P.b1) * γP + P.b1 * γG < P.b0 * (cr P - dc P) →
deriv (fun y => optProfit P γP y) γG < 0) ∧
(P.b0 * (cr P - dc P) ≤ (P.b0 - 2 * P.b1) * γP + P.b1 * γG →
0 ≤ deriv (fun y => optProfit P γP y) γG) ∧
((P.b0 - 2 * P.b1) * γP + P.b1 * γG < P.b0 * (cr P - dc P) ↔
tauMStar P γP γG < P.b1 / P.b0) ∧
(P.b0 * γG - P.b1 * γP ≤ P.b0 * (cr P - dc P) →
deriv (fun y => optProfit P γP y) γG ≤ deriv (fun x => optProfit P x γG) γP) ∧
(P.b0 * (cr P - dc P) < P.b0 * γG - P.b1 * γP →
deriv (fun x => optProfit P x γG) γP < deriv (fun y => optProfit P γP y) γG) ∧
(P.b0 * γG - P.b1 * γP ≤ P.b0 * (cr P - dc P) ↔
tauMStar P γP γG ≤ (P.b0 + P.b1) / (2 * P.b0)) := by sorry
end PrideGuilt
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.