Restoring Trust in Human–Centered AI: Differential Impacts of Trust Repair Strategies in Human–AI Versus Human–Human Teams

The rapid advancement of artificial intelligence (AI) is reshaping human collaboration, making trust repair critical in human–AI teams. However, limited guidance exists on how to restore trust after AI-related violations. Drawing on social exchange theory and trust repair theory, this study examined the effects of trust repair strategies in team decision-making. We conducted a 3 × 2 between-subjects experiment with 134 valid participants, manipulating trust repair strategy (commitment, apology, no repair) and team type (human–AI vs. human–human) in an urban risk management task. Results showed that team type significantly affected team performance and job satisfaction, and that repair strategies moderated these effects. Perceived support and collective efficacy mediated, and jointly chain-mediated, the relationships among team type, repair strategy, and outcomes. Commitment-based repair produced the most favorable results, particularly in human–AI teams. These findings advance trust repair theory in human–AI collaboration and inform human-centered AI design.