Proof of Theorem 4.2, p. 18 — E[S(ω_S)] − E[S(ω_T)] = b_S(v̄_S − v̄_T) + (b_S − b_T)(v̄_T − W) + W(a_T − a_S)
ProvedEngagementOpt.Salad.p18_gap_identitybehavioral-economicsengagement-optimizationp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1
Let and be points of , let , and write , . If the user participates at both points, and , then the utility gap is
This decomposition is the starting point of the bound in Theorem 4.2: the first term is controlled by the value spread , the second by (5), and the third by (7).
Formalization Note The statement uses as in (1); the participation hypotheses make at both points, which is how the paper's display reads.
Preamble
import Mathlib import Definitions.Def_EngagementOpt_Salad_Model
Formal statement
namespace EngagementOpt.Salad
/-- The identity on p. 18 (proof of Theorem 4.2, first three lines of the display): for two points
of `Ω` at which `g_S ≥ 0`,
`E[S(ω_S)] − E[S(ω_T)] = b_S(v̄_S − v̄_T) + (b_S − b_T)(v̄_T − W) + W(a_T − a_S)`. -/
theorem p18_gap_identity (W : ℝ) (ωS ωT : EngagementOpt.Misalign.Params) (hS : EngagementOpt.Misalign.InOmega ωS) (hT : EngagementOpt.Misalign.InOmega ωT)
(hgS : 0 ≤ EngagementOpt.Misalign.gS W ωS) (hgT : 0 ≤ EngagementOpt.Misalign.gS W ωT) :
ES W ωS - ES W ωT =
bCoef ωS * (ωS.v - ωT.v) + (bCoef ωS - bCoef ωT) * (ωT.v - W)
+ W * (aCoef ωT - aCoef ωS) := by sorry
end EngagementOpt.Salad
Source
Kleinberg, Mullainathan, Raghavan, The Challenge of Understanding What Users Want, arXiv:2202.11776v3, p. 18, proof of Theorem 4.2, first three lines of the display
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.