When artificial intelligence apologizes: Dynamic trust updating and the role of personality.

Artificial intelligence (AI) systems are susceptible to error, and many are designed to apologize.Yet, it remains unclear which apologies restore user trust, a question requiring repeated interactions with an AI agent and models capturing how trust evolves across encounters.Drawing on models of apology from human-human interaction research, we examined whether apology components that repair interpersonal trust function similarly in AI.Participants (N = 245) solved five business problems with a chatbot preprogrammed to err in two tasks at fixed intervals.They were randomly assigned to one of seven conditions: error with no apology or error followed by one of six apology strategies (Expression of Regret, Explanation, Acknowledgment of Responsibility, Declaration of Repentance, Offer of Repair, Request for Forgiveness).Using a dynamic trust-updating model, we estimated how trust shifted after each trial, following errors versus correct performance.Trust trajectories differed significantly by apology condition.Relative to no apology, only Expression of Regret buffered the typical immediate decline in trust following the errorapology-correction sequence.None of the apology conditions further enhanced trust recovery once the system resumed correct performance.Larger apology-related reductions in trust loss occurred among participants with more relational, relative to functional, postinteraction evaluation profiles.In separate bivariate analyses, all five Big Five traits were associated with baseline trust.When considered jointly, only Agreeableness and Openness remained associated with baseline trust, and Openness was also linked to dynamic trust updating.These findings suggest that apology-based trust repair in AI is selective and context-dependent.