Beyond the dichotomy of use and Not-Use: Forms and motivations of user non-use behaviors toward AI customer service

Abstract

In the era of artificial intelligence, the dichotomy of acceptance versus abandonment falls short in capturing the complexities of user behavior. This study focuses on AI customer service (AICS)—a pervasive but difficult-to-abandon AI application—to explore how users’ perceptions shape diverse non-use behaviors. Through a sequential mixed-methods design, we develop and test an extended Expectation Confirmation Model (ECM). Qualitative findings first reveal that users evaluate AICS through functional and emotional dimensions, and engage in two forms of non-use behavior: active (e.g., self-education, seeking help) and passive (e.g., teasing, verbal abuse). Building on these insights, quantitative results show that both functional and emotional performance perceptions significantly predict satisfaction. Crucially, cognitive satisfaction negatively predicts passive non-use, whereas emotional satisfaction positively predicts active non-use—a finding that challenges conventional models where satisfaction predicts continuance. Moreover, expectation confirmation significantly moderates the relationship between performance perception and satisfaction. Our research extends ECM by introducing a dual-pathway model of non-use and offers managers and designers nuanced insights for improving human-AI coexistence.