Theorem 4.10 — no truthful additive mechanism is a -approximation for task scheduling for any
ProvedAlgMechDesign.Additive.no_additive_mechanism_below_nConsider task scheduling with agents and tasks. Let be a truthful direct mechanism whose prices are additive: for every agent , positive type vector and set of tasks . Then for every real the allocation rule is not a -approximation: there are a positive type vector and an allocation with
By the revelation principle the same holds for every mechanism whose truthful implementation is additive. The bound is tight: the MinWork mechanism, which gives each task to the fastest agent and pays it the second-best time, is truthful, additive and an -approximation.
Formalization Note The threshold is the one the paper's proof uses ("Let "); the theorem's printed statement does not mention . The statement is kept for every ([NeZero n]); at it holds because every allocation is the same. Truthfulness and additivity quantify over positive types; running time is not modelled.
import Mathlib import Definitions.Def_AlgMechDesign_Additive_Model import Definitions.Def_AlgMechDesign_Additive_Price
namespace AlgMechDesign.Additive
/-- Theorem 4.10, p. 180: for `n ≥ 1` agents and `k ≥ n²` tasks, no truthful additive mechanism
for task scheduling is a `c`-approximation for any `c < n`. -/
theorem no_additive_mechanism_below_n {n k : ℕ} [NeZero n] (hk : n ^ 2 ≤ k)
(alloc : (Fin n → Fin k → ℝ) → (Fin k → Fin n))
(pay : (Fin n → Fin k → ℝ) → Fin n → ℝ) (htr : IsTruthful alloc pay)
(hadd : IsAdditive alloc pay) (c : ℝ) (hc : c < n) :
¬ IsApprox c alloc := by sorry
end AlgMechDesign.Additive
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.