Human-AI Complementarity: A Goal for Amplified Oversight

TL;DR

This paper investigates how AI can assist humans in fact-verification tasks to improve oversight of AI systems, emphasizing the importance of appropriate trust and effective AI-human collaboration methods.

Contribution

It introduces methods for combining AI and human ratings based on AI confidence and evaluates different AI assistance displays to optimize human oversight.

Findings

Combining AI and human ratings improves fact-verification accuracy.

AI assistance with search results and evidence fosters better human trust.

Over-reliance occurs when AI explanations and confidence are displayed.

Abstract

Human feedback is critical for aligning AI systems to human values. As AI capabilities improve and AI is used to tackle more challenging tasks, verifying quality and safety becomes increasingly challenging. This paper explores how we can leverage AI to improve the quality of human oversight. We focus on an important safety problem that is already challenging for humans: fact-verification of AI outputs. We find that combining AI ratings and human ratings based on AI rater confidence is better than relying on either alone. Giving humans an AI fact-verification assistant further improves their accuracy, but the type of assistance matters. Displaying AI explanation, confidence, and labels leads to over-reliance, but just showing search results and evidence fosters more appropriate trust. These results have implications for Amplified Oversight -- the challenge of combining humans and AI to supervise AI systems even as they surpass human expert performance.