When AI Service Fails: The Role of Procedural Fairness in Shaping Citizens’ Usage Intentions

With the increasing deployment of artificial intelligence (AI) across organizations, this study examines citizens’ interactions with AI chatbots under service failure conditions. Unlike existing studies, it focuses on the public sector and investigates the impact of service failure on citizens’ intentions to use digital public services. A survey experiment is conducted by embedding AI chatbot interactions into controlled scenarios. Drawing on expectation disconfirmation theory, the results show that, compared to AI service failure, human service failure leads to a more substantial decline in citizens’ intentions to continue using human-provided services. Mechanism analysis reveals an expectation shift effect, whereby human service failure lowers citizens’ expectations toward human services and shifts their intentions toward AI. Procedural fairness negatively moderates this process, attenuating the shift in citizen intentions. At the same time, increased intention to use AI services generates an intention backflow effect, partially restoring citizens’ intentions toward human services. By uncovering the dynamic and reciprocal relationship between AI and human service usage under failure conditions, this study advances our understanding of how citizens make choices between AI and human services and provides practical implications for designing more reliable and user-centered AI systems in services.