Key points are not available for this paper at this time.
Recent advances in large language model (LLM)-based chatbots have increased interest in anthropomorphizing AI systems. However, current research emphasizes human-like appearance and behavior rather than examining how specific anthropomorphic features functionally shape user outcomes. Addressing this issue, this study integrates the Humanness-Value-Loyalty (HVL) model with dehumanization theory to develop a differentiated account of how four key anthropomorphic traits, namely warmth, competence, civility, and morality, are associated with user loyalty via perceived service value. Based on a quota-sampled survey of 1,010 users of Doubao, China’s leading LLM-based chatbot, structural equation modeling reveals that perceived warmth, competence, and civility foster loyalty through enhanced value perception, whereas perceived morality paradoxically diminishes loyalty by reducing value. Besides, competence exerts the strongest and most direct influence, followed by warmth, civility, and finally morality. Our findings challenge the prevalent ‘more human-like is better’ assumption and advocate a context-sensitive approach to AI design that foregrounds value alignment over superficial human resemblance. Furthermore, this study proposes a layered architecture of humanness effects, enabling scholars and designers to tailor anthropomorphic features to optimize specific outcomes.
Xi et al. (Tue,) studied this question.