The Physical, Emotional, and Autonomous Anthropomorphism of Service Chatbots
Prior research on chatbot anthropomorphism has often treated it as a unidimensional construct, producing inconsistent findings regarding how human‐like design cues shape user responses. This study conceptualizes anthropomorphism as a multidimensional higher‐order perception comprising three distinct forms of anthropomorphic attribution (physical, emotional, and autonomous) and examines how each dimension independently and interactively influences customer satisfaction with service chatbots. Drawing on the stereotype content model and expectancy violation theory, the study proposes that different anthropomorphism dimensions activate distinct evaluative pathways: physical and emotional cues primarily signal warmth, whereas autonomous cues primarily signal competence. A mixed‐method approach was employed across four studies, combining LLM‐based text analysis with experimental designs. Study 1 identifies the three dimensions from consumers' own descriptions of chatbot human‐likeness. Study 2 demonstrates that autonomous anthropomorphism uniquely drives satisfaction through perceived competence, whereas emotional anthropomorphism enhances perceived warmth, which does not translate into higher satisfaction in utilitarian service contexts, and physical anthropomorphism shows no significant effect on satisfaction. Study 3 reveals a counterintuitive interaction: the positive effect of autonomous anthropomorphism on satisfaction is attenuated when emotional anthropomorphism is high, suggesting that warmth‐signaling cues can undermine competence‐driven benefits in utilitarian service contexts. Study 4 replicates the interaction found in Study 3, and confirms that the attenuation happens in the translation of autonomous cues into competence perceptions and it is this diluted competence perception, not any change in warmth, that carried the interaction into satisfaction. By disentangling anthropomorphism into discrete dimensions and empirically testing their interplay, this research provides a structured lens through which previously mixed findings can be better understood. The findings challenge managers to prioritize autonomous capabilities over superficial human‐like traits when designing AI‐driven services, particularly in efficiency‐focused industries such as banking and e‐commerce.