How Does Chatbot Anthropomorphism Shape Psychological Reactance? An Experiment on Trust and Perceived Persuasive Intent in Vaping Cessation Conversations

AI chatbots are increasingly used to deliver health information and persuasive interventions, yet the consequences of making these systems appear humanlike remain unclear. Drawing on the Computers Are Social Actors and Media Are Social Actors paradigms, mind perception theory, and psychological reactance theory, this research examined whether anthropomorphism in AI chatbots shapes reactance through two competing pathways: trust in the chatbot and perceived persuasive intent. An online between-subjects experiment was conducted among U.S. adults who currently used e-cigarettes. Participants in the main study (N = 586) were randomly assigned to interact with either an anthropomorphic or a non-anthropomorphic AI chatbot, or to view a static vaping cessation message. The results showed that interacting with an anthropomorphized chatbot reduced reactance compared to reading a static message, although this advantage did not extend to attitudes toward vaping or intentions to quit. Additionally, while anthropomorphism can generally reduce psychological reactance, its two dimensions, mindful and mindless anthropomorphism, operated in different paths. Mindful anthropomorphism was associated with greater trust and, indirectly, lower reactance. Mindless anthropomorphism was directly associated with lower reactance. These findings support the potential use of anthropomorphic AI chatbots as a communication approach for vaping cessation. Furthermore, by distinguishing mindful from mindless anthropomorphism and clarifying their distinct mechanisms, this research offers a more nuanced explanation of how humanlike AI cues shape psychological reactance and provides practical guidance for designing AI-mediated health interventions.