Overcoming the Incentive Collapse Paradox

TL;DR

This paper introduces a sentinel-auditing payment mechanism and an incentive-aware inference framework to maintain human effort and optimize statistical accuracy under budget constraints in AI-assisted tasks.

Contribution

It proposes a novel incentive-robust payment scheme and an integrated active inference approach that jointly optimize effort enforcement and statistical performance.

Findings

Sentinel-auditing mechanism enforces positive human effort at finite cost.

The framework achieves better cost-error tradeoffs than standard methods.

Experiments validate improved efficiency in budget-constrained settings.

Abstract

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.