Proof of Theorem 1, case (corrected: ): for every
OpenDigitalPrivacy.DataStrategy.proof_thm1_case_psi_lt_Prhodigital-privacygame-theoryoptimizationplatform-economics
Throughout, is the strength of network effects and the users' value of data (p. 6), the business's revenue model and the unit cost of data protection (p. 8). If , then
Hence the two best-response curves meet only at , and on the candidate optima are and .
Differences from the source:
- correction: ψ ≥ 0 (p. 8), so the case ψ < Pρ includes ψ = 0 → 0 < ψ < Pρ; reason: C_h divides by ψ and is undefined at ψ = 0; at ψ = 0 the profit has no maximizer (β = 0, ρ = P = 1/2: Π(1, C) = 3C/(2(C + 1)) → 3/2 unattained)
Preamble
import Mathlib import Definitions.Def_DigitalPrivacy_DataStrategy_cInv import Definitions.Def_DigitalPrivacy_DataStrategy_cH
Formal statement
namespace DigitalPrivacy.DataStrategy
/-- **Proof of Theorem 1, case `ψ < Pρ` (corrected: `ψ > 0`): `C_inv(ξ) < C_h(ξ)` for every `ξ > 0`.**
Fainmesser, Galeotti, Momot, *Digital Privacy*, author manuscript (April 2022 draft; SSRN 3459274), proof of Theorem 1 (App. A.2), pp. 29–30 (PDF 30–31), unnumbered: "Case ψ < Pρ: We first
show that Cinv(ξ) < Ch(ξ) (where Ch(ξ) is defined in Lemma 1) and they only intersect at ξ = 0
under this condition. … It is easy to check that inequality holds in this case and hence
Cinv(ξ) < Ch(ξ), ∀ξ > 0."
For `β, ρ ∈ [0, 1)`, `P ∈ [0, 1]` and `0 < ψ < Pρ`: `C_inv(ξ) < C_h(ξ)` for every `ξ > 0`.
**Formalization Note.** The page allows `ψ ≥ 0` (p. 8); `C_h` divides by `ψ`, so `ψ = 0` is
excluded. -/
theorem proof_thm1_case_psi_lt_Prho {β ρ P ψ : ℝ}
(hβ0 : 0 ≤ β) (hβ1 : β < 1) (hρ0 : 0 ≤ ρ) (hρ1 : ρ < 1)
(hP0 : 0 ≤ P) (hP1 : P ≤ 1) (hψ0 : 0 < ψ) (hψ : ψ < P * ρ) :
∀ ξ : ℝ, 0 < ξ → cInv β ρ P ξ < cH β ρ P ψ ξ := by sorry
end DigitalPrivacy.DataStrategy
Source
Fainmesser, Galeotti, Momot, Digital Privacy, author manuscript (April 2022 draft; SSRN 3459274), proof of Theorem 1 (App. A.2), pp. 29–30 (PDF 30–31), unnumbered