Robust Human-AI Complementarity under Uncertainty

2026-07-07

AI Summary

This study addresses the challenge that humans often fail to achieve complementary gains from AI predictions, even when AI signals are informative. The authors identify a negatively correlated error structure between human and AI predictions as a critical condition for realizing such complementarity. Building on this insight, they propose a robust decision-making mechanism that guarantees utility improvement under asymmetric information quality. Through theoretical analysis grounded in expected utility theory, examination of error correlation patterns, and empirical evaluation across multiple real-world prediction tasks, the work demonstrates the widespread presence of this negative error correlation and shows that the proposed strategy consistently enhances human-AI collaborative performance.

Abstract

Machine learning models are often intended to augment rather than replace human decision makers, by providing information that is complementary to human judgement. Yet, in practice, human decision makers routinely fail to realize such complementary gains, even when models provide useful signal. In this work, we study how asymmetric information about the quality of information available to a human decision maker vs. an AI impacts the ability of a decision maker to extract complementary value from AI predictions. We show that a key factor is the error correlation structure between human and AI predictions. In particular, when the AI's prediction errors are \textit{negatively correlated} with those of the human, the decision maker can construct robust strategies which guarantee improvements in expected utility. We empirically investigate whether these conditions for complementarity arise in practice, using real-world forecasting benchmarks.

Problem

• Human-AI complementarity
• asymmetric information
• error correlation
• decision making under uncertainty
• complementary gains

Innovation

• human-AI complementarity
• error correlation
• negative correlation
• robust decision-making
• asymmetric information