WHEN AI FAILS IN SERVICES: HOW HUMAN–AI RELATIONSHIPS SHAPE CUSTOMER RESPONSES
(A2026-131828)

AUTHORS

Shuqin Liu, College of Management and Economics, Tianjin University/School of Fashion and Textiles, The Hong Kong Polytechnic University; Chundong Zheng, Tianjin University

ABSTRACT

Drawing on the attachment–aversion (AA) theory of human–AI relationships, this research investigates how human–AI relationships affect customer responses to AI service failures and the underlying mechanisms. Through one survey and three experiments, we reveal that using AI agents in a relational (vs. functional) mode intensifies customers’ negative responses when failures occur. Furthermore, a close and connected self–AI relationship mitigates such negative responses more effectively than a far and disconnected one. This buffering effect is driven by perceived betrayal and forgiveness. However, this mitigating influence only holds when the prominence of AI is low. By unpacking these nuanced effects, our findings advance the understanding of human–AI relationships. The insights offered by this research carry practical relevance for designing AI agents that are more appropriately relationship-oriented, particularly in service contexts.