The Relational Roots of Algorithmic Forgiveness: How Reflective Skepticism and Self–AI Connection Shape Responses to AI Bias

As generative AI systems increasingly shape information consumption, understanding how users respond to the biased outputs generated by those AI systems is critical. While prior research has focused on fairness and trust evaluations, less attention has been given to users’ willingness to forgive algorithmic bias. Drawing on dual-process perspectives and identity-based theories, this study examines how dispositional cognitive reflection influences willingness to forgive AI-generated gender stereotypes through self-AI connection. Using a scenario-based survey, participants evaluated a generative AI system that produced gender-biased career recommendations. Results show that cognitive reflection is negatively associated with self-AI connection, which in turn increases willingness to forgive biased outputs. Self-AI connection fully mediates this relationship. These findings highlight the relational roots of algorithmic forgiveness and advance understanding of heterogeneous user responses to biased AI systems