Pathways linking expectations for AI chatbots to loyalty: A moderated mediation analysis

Abstract

Despite the prevalence of generative AI chatbots such as GPT and Bard, scholarly inquiry into how users' enduring expectations influence their engagement with AI chatbots remains scant. Drawing on expectation violation theory, the present study examines how user expectations, informed by belief in machine heuristics and concerns over human uniqueness, impact chatbot loyalty through a moderated mediation framework. A questionnaire survey of 900 participants in China revealed that belief in machine capabilities bolsters users' perceptions of machine intelligence, which in turn, enhances user loyalty. Interestingly, when users encounter service failures that challenge their expectations, their perceived intelligence of the chatbot intensifies rather than diminishes. In contrast, expectations shaped by human uniqueness concerns diminish users' perceptions of machine intelligence and, consequently, their loyalty, which remains consistent regardless of the encounter with AI failure. The study also delves into the theoretical contributions of these findings to the evolution of expectation violation theory within the sphere of human-robot interaction.