Section 5, p. 9 — no optimal policy in the three-value general-cost instance is outcome monotonic
ProvedStrategicPolicy.NonMonotone.no_outcome_monotonic_optimumgeneral-costsoptimal-policyoutcome-monotonicityp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1strategic-classification
For the fixed three-value instance of Section 5, at least one admissible optimal policy exists, but every admissible policy maximizing the strategic utility fails outcome monotonicity:
This gives the paper's counterexample to the existence of an outcome-monotonic optimum under general costs. The existence clause makes the counterexample substantive.
Formalization Note Outcome monotonicity uses the one-way reading of equation (1), . Consequently the negative conclusion is stronger than failure of the literal biconditional printed there. The maximum is taken over all real-valued policies in , not a finite grid.
Preamble
import Mathlib import Definitions.Def_StrategicPolicy_NonMonotone_Setting import Definitions.Def_StrategicPolicy_NonMonotone_Example
Formal statement
namespace StrategicPolicy.NonMonotone
/-- Section 5, page 9: an optimum exists, and none obeys outcome monotonicity. -/
theorem no_outcome_monotonic_optimum :
(∃ π : Fin 3 → ℝ, π ∈ StrategicPolicy.MonotoneBinary.Policies 3 ∧ IsMaxOn (StrategicPolicy.MonotoneBinary.u p q γ c) (StrategicPolicy.MonotoneBinary.Policies 3) π) ∧
(∀ π : Fin 3 → ℝ, π ∈ StrategicPolicy.MonotoneBinary.Policies 3 → IsMaxOn (StrategicPolicy.MonotoneBinary.u p q γ c) (StrategicPolicy.MonotoneBinary.Policies 3) π →
¬ StrategicPolicy.MonotoneBinary.OMPolicy q π) := by sorry
end StrategicPolicy.NonMonotone
Source
Tsirtsis et al., Optimal Decision Making Under Strategic Behavior, arXiv:1905.09239v5, p. 9, §5 bold heading and general-cost example
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.