Overcoming the Incentive Collapse Paradox

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Yin, Qichuan, Su, Ziwei, Li, Shuangning
Format: Preprint
Veröffentlicht: 2026
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866915896665047040
author Yin, Qichuan
Su, Ziwei
Li, Shuangning
author_facet Yin, Qichuan
Su, Ziwei
Li, Shuangning
contents AI-assisted task delegation is increasingly common, yet human effort in such systems is costly and typically unobserved. Recent work by Bastani and Cachon (2025); Sambasivan et al. (2021) shows that accuracy-based payment schemes suffer from incentive collapse: as AI accuracy improves, sustaining positive human effort requires unbounded payments. We study this problem in a budget-constrained principal-agent framework with strategic human agents whose output accuracy depends on unobserved effort. We propose a sentinel-auditing payment mechanism that enforces a strictly positive and controllable level of human effort at finite cost, independent of AI accuracy. Building on this incentive-robust foundation, we develop an incentive-aware active statistical inference framework that jointly optimizes (i) the auditing rate and (ii) active sampling and budget allocation across tasks of varying difficulty to minimize the final statistical loss under a single budget. Experiments demonstrate improved cost-error tradeoffs relative to standard active learning and auditing-only baselines.
format Preprint
id arxiv_https___arxiv_org_abs_2603_27049
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Overcoming the Incentive Collapse Paradox
Yin, Qichuan
Su, Ziwei
Li, Shuangning
Machine Learning
AI-assisted task delegation is increasingly common, yet human effort in such systems is costly and typically unobserved. Recent work by Bastani and Cachon (2025); Sambasivan et al. (2021) shows that accuracy-based payment schemes suffer from incentive collapse: as AI accuracy improves, sustaining positive human effort requires unbounded payments. We study this problem in a budget-constrained principal-agent framework with strategic human agents whose output accuracy depends on unobserved effort. We propose a sentinel-auditing payment mechanism that enforces a strictly positive and controllable level of human effort at finite cost, independent of AI accuracy. Building on this incentive-robust foundation, we develop an incentive-aware active statistical inference framework that jointly optimizes (i) the auditing rate and (ii) active sampling and budget allocation across tasks of varying difficulty to minimize the final statistical loss under a single budget. Experiments demonstrate improved cost-error tradeoffs relative to standard active learning and auditing-only baselines.
title Overcoming the Incentive Collapse Paradox
topic Machine Learning
url https://arxiv.org/abs/2603.27049