AI2030 Responsible AI Tools Marketplace

Key Takeaways

Human errors damage trust in experts more than AI corrections build trust in AI: When advisors made mistakes, participants significantly reduced their trust in the human expert. However, when AI visibly corrected those same errors, trust in the AI system did not increase proportionally, revealing an asymmetry in how people evaluate human versus AI performance.
People remain anchored to trusting the human expert regardless of AI involvement: Across all experimental conditions—whether AI support was absent, advisor-controlled, or automatically monitoring—participants maintained their primary trust in the human advisor rather than redistributing trust toward the AI system or the human-AI team as a whole.
Making AI oversight visible doesn't shift trust toward AI: The study tested whether making AI corrections visible to users would increase confidence in the AI system, but changing the visible structure of AI involvement "does not substantially redistribute trust toward the AI assistant." This suggests that transparency alone may not address trust imbalances in human-AI teams.
Epistemic dependence creates vulnerability in AI-assisted expert settings: In domains like medicine, law, and finance where users cannot independently verify expert recommendations, the asymmetric trust effects become particularly concerning—people may over-rely on human judgment even when AI has demonstrated superior error-catching ability.
Corrective AI has limits as a governance tool: While visible AI oversight can help recover from errors and improve workflow outcomes, it should not be relied upon as the primary mechanism for building appropriate trust in AI systems or ensuring balanced human-AI team performance.

Abstract

The increasing integration of AI-powered tools into expert workflows, such as medicine, law, and finance, raises a critical question: how does AI involvement influence a user`s trust in the human expert, the AI system, and the human-AI team? This question is especially important in expert advisory settings where users are epistemically dependent on human-AI systems: they are recipients of guidance produced by an expert using AI support, but often lack the domain knowledge needed to independently verify the recommendation. We investigated these dynamics through a user study (N=157) using a simulated course-planning task. Our design varied advisor performance and the structure of AI involvement, including whether AI support was present and, when present, whether it was invoked by the advisor or automatically monitored the interaction. Across all conditions, workflows ultimately produced correct schedules. Results show an asymmetric trust effect: advisor errors significantly reduce trust in the human advisor, but visible AI correction does not produce a corresponding increase in trust toward the AI assistant. Trust judgments remain anchored to the advisor across multiple trust measures, and changing the visible structure of AI involvement does not substantially redistribute trust toward the AI assistant or the human-AI team. These findings suggest a limit of corrective AI as a governance mechanism: making AI oversight visible may improve workflow recoverability, but it does not necessarily redistribute trust or responsibility away from the human-facing expert. In epistemically dependent settings, correctness alone may be insufficient for trustworthy AI integration, as users may continue to assign responsibility to the human expert even when AI assistance visibly shapes the final outcome.

Revised version. Accepted to AIES 2026. Includes technical appendices and additional analyses

Key Takeaways

Human errors damage trust in experts more than AI corrections build trust in AI: When advisors made mistakes, participants significantly reduced their trust in the human expert. However, when AI visibly corrected those same errors, trust in the AI system did not increase proportionally, revealing an asymmetry in how people evaluate human versus AI performance.
People remain anchored to trusting the human expert regardless of AI involvement: Across all experimental conditions—whether AI support was absent, advisor-controlled, or automatically monitoring—participants maintained their primary trust in the human advisor rather than redistributing trust toward the AI system or the human-AI team as a whole.
Making AI oversight visible doesn't shift trust toward AI: The study tested whether making AI corrections visible to users would increase confidence in the AI system, but changing the visible structure of AI involvement "does not substantially redistribute trust toward the AI assistant." This suggests that transparency alone may not address trust imbalances in human-AI teams.
Epistemic dependence creates vulnerability in AI-assisted expert settings: In domains like medicine, law, and finance where users cannot independently verify expert recommendations, the asymmetric trust effects become particularly concerning—people may over-rely on human judgment even when AI has demonstrated superior error-catching ability.
Corrective AI has limits as a governance tool: While visible AI oversight can help recover from errors and improve workflow outcomes, it should not be relied upon as the primary mechanism for building appropriate trust in AI systems or ensuring balanced human-AI team performance.

Abstract

Revised version. Accepted to AIES 2026. Includes technical appendices and additional analyses

Asymmetric Trust Effects of Corrective AI in Expert Advisory Workflows under Epistemic Dependence

Key Takeaways

Abstract

Paper Info

Responsible AI Topics

Links

Asymmetric Trust Effects of Corrective AI in Expert Advisory Workflows under Epistemic Dependence

Key Takeaways

Abstract

Paper Info

Responsible AI Topics

Links