When AI–chatbots disappoint – the role of freedom of choice and user expectations in attribution of responsibility for failure

Purpose The implementation of AI-powered chatbots in the frontline may enhance efficiency, yet failures are still common. This paper aims to explore users' attribution of responsibility for service failures when using AI–chatbots and to examine how contextual factors influence perceptions of blame. Design/methodology/approach This work utilizes a mixed-methods approach, leveraging the findings from 39 exploratory interviews to develop the research framework and hypotheses. Subsequently, two experimental studies evaluated the type of interaction, failure type and failure severity. Findings The qualitative study identified voluntary and forced interaction types perceived by users based on contextual factors and demonstrated how these types impact expectations and responsibility attribution post-failure. The experimental studies showed that forced interactions intensify responsibility attributions toward the company and that disconfirmation of expectations mediates the relationship between forced interactions and responsibility attribution. Furthermore, failure type and severity level have a moderating influence on responsibility attribution. Originality/value This paper contributes to the theoretical understanding of user interactions with AI-powered frontline technology, by revealing the nuanced ways in which users perceive and react to failures.