I can feel AI failure: the impact of service failure type and failure assessment on customer recovery expectation

Purpose Based on appraisal theory and social response theory, this study aims to explore the mechanism of AI failure types on consumer recovery expectation from the perspective of service failure assessment and validate the moderate role of anthropomorphism level. Design/methodology/approach Three scenario-based experiments were conducted to validate the research model. First, to test the effect of robot service failure types on customer recovery expectation; second, to further test the mediating role of perceived controllability, perceived stability and perceived severity; finally, to verify the moderating effect of anthropomorphic level. Findings Non-functional failures reduce consumer recovery expectation compared to functional failures; perceived controllability and perceived severity play a mediating role in the impact of service failure types on recovery expectation; the influence of service failure types on perceived controllability and perceived severity is moderated by the anthropomorphism level. Originality/value The findings enrich the influence mechanism and boundary conditions of service failure types, and have implications for online enterprise follow-up service recovery and improvement of anthropomorphic design.