AI-expressed empathy in misaligned recovery: consumer forgiveness following recommendation failures

AI-mediated recommendation recovery represents a misaligned recovery context in which a failure of algorithmic preference understanding is followed by AI-mediated emotional recovery. After consumers receive an unsuitable recommendation, an AI customer-service agent attempts to express understanding and concern, raising a theoretical question: when can AI-expressed empathy become psychologically supportive after AI has failed to understand consumers’ preferences? Drawing on forgiveness theory and the stress-buffering hypothesis, we argue that AI-expressed empathy fosters consumers’ willingness to forgive when it is appraised as socially present and emotionally supportive, thereby relieving post-failure stress. Across three scenario-based experiments in online shopping, travel planning, and online learning (N = 705), AI-expressed empathy increased consumers’ forgiveness intentions through perceived social presence and perceived stress relief. This process was stronger for consumers with a communal relationship norm orientation and under high failure severity. The study explains how AI-expressed empathy can support affective-relational recovery in AI-mediated recommendation failures.