Observation 1: is U-BNIC if
ProvedBlockchainTFM.observation_1blockchainmechanism-designp2o-gran-per-chapterp2o-plan-paperp2o-strengthenedp2o-v1transaction-fee-mechanism
Let valuations be i.i.d. from a law on . Let and be two transaction fee mechanisms with the same allocation rule, and let
be their payment difference. Suppose is U-BNIC. Then is U-BNIC if, for every user and every bid ,
User 's utility under is her utility under minus , so (EC.34) makes the two mechanisms give her the same expected utility at every bid. In the paper is the dominant auxiliary mechanism of ; the observation is the U-BNIC half of Theorem 2.
Formalization Note is assumed integrable for each , so that the expectation in (EC.34) exists.
Differences from the source:
- reading: Observation 1 states no hypothesis on M; its proof ends 'As mechanism M is U-BNIC' → hypothesis IsUBNIC V₀ M; reason: the hypothesis is the proof's last line; without it the observation is false
- generalization: M is the dominant auxiliary mechanism (a, p, 0) of M̃ → any TFM M with the same allocation rule as M̃ that is U-BNIC; reason: the proof uses only θ = a(p̃ − p) and U-BNIC of M; the dominant auxiliary is an instance
- generalization: V₀ has a pdf ρ on [0,1] → any probability measure V₀ on ℝ with V₀([0,1]) = 1; reason: the proof uses no density; every law with a pdf on [0,1] is an instance
- encoding: E[θ_i(b_i, b_{−i})] = 0 → θ_i(b_i,·) integrable against V₀^{⊗n} and its integral 0, for every i and b_i ∈ [0,1]; reason: integrability only excludes the case where the expectation is undefined
- encoding: users 1,…,n; b_{−i}; bids and values in [0,1] → Fin n; a vector b ∈ [0,1]ⁿ whose i-th entry is overwritten (Function.update); bids, values and deviations in Set.Icc 0 1; reason: Lean convention; same instances
Preamble
import Mathlib import Definitions.Def_BlockchainTFM_Model import Definitions.Def_BlockchainTFM_Properties open MeasureTheory
Formal statement
namespace BlockchainTFM
/-- **Observation 1**, Chen–Simchi-Levi–Zhao–Zhou, arXiv:2209.13099v7, p.ec10. Let
`M̃ = (a, p̃, r̃)` and its auxiliary mechanism `M = (a, p, 0)` share the allocation rule, let
`θ_i = a_i (p̃_i − p_i)` be their payment difference (8), and let `M` be U-BNIC. Then `M̃` is
U-BNIC if `E_{b_{-i}∼V_{-i}}[θ_i(b_i, b_{-i})] = 0` (EC.34) for every user `i` and bid `b_i`.
**Formalization Note.** The hypothesis "`M` is U-BNIC" is the last line of the paper's proof
("As mechanism `M` is U-BNIC, it holds that `M̃` is also U-BNIC"). `θ` is pinned to (8) by `hθ`.
(EC.34) is required for every bid `b_i ∈ [0,1]`, with `θ_i(b_i, ·)` integrable so that the
expectation exists. Only the allocation and payment rules of `M` and `M̃` matter (U-BNIC ignores
the miner revenue). `V₀` is a probability law carried by `[0,1]`. -/
theorem observation_1 {n : ℕ} (V₀ : Measure ℝ) [IsProbabilityMeasure V₀]
(hV₀ : V₀ (Set.Icc 0 1) = 1) (M Mt : TFM n) (θ : (Fin n → ℝ) → Fin n → ℝ)
(halloc : Mt.alloc = M.alloc)
(hθ : ∀ b i, θ b i = M.alloc b i * (Mt.pay b i - M.pay b i))
(hM : IsUBNIC V₀ M)
(hθint : ∀ i : Fin n, ∀ x ∈ Set.Icc (0 : ℝ) 1,
Integrable (fun b => θ (Function.update b i x) i) (Measure.pi fun _ : Fin n => V₀))
(hθzero : ∀ i : Fin n, ∀ x ∈ Set.Icc (0 : ℝ) 1,
∫ b, θ (Function.update b i x) i ∂(Measure.pi fun _ : Fin n => V₀) = 0) :
IsUBNIC V₀ Mt := by sorry
end BlockchainTFM
Source
Chen, Simchi-Levi, Zhao, Zhou, Bayesian Mechanism Design for Blockchain Transaction Fee Allocation, arXiv:2209.13099v7, Appendix EC.4.1, Observation 1 (EC.34)–(EC.35), p.ec10
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.