목차

Associate Professor, Department of Global Business Administration, Anyang University With the growing interest in Physical AI, attention toward humanoid AI has also been increasing. Humanoid AI has now evolved beyond technical functionalities to a level where it can engage in emotional exchanges and foster social connections. However, academic research has largely approached humanoid AI from a functional perspective, paying relatively little attention to emotionally driven modes of interaction. In particular, there is still a lack of systematic research on verbal embodiment, which lies at the core of user-AI interaction. This study empirically examines the effects of the humanoid AI’s role type (task-oriented vs. social-oriented) and the level of verbal embodiment (turn-taking vs. grounding) on users’ evaluations and behavioral intentions. The focus is on situations where users receive product recommendations from a humanoid AI. The results show that, in such recommendation contexts, users not only feel more satisfied with the interaction during the grounding phase—where conversations evolve across diverse topics—compared to turn-taking exchanges, but also exhibit higher purchase intentions for the recommended products. However, the positive effect of grounding was found to be reduced when the AI adopts a task-oriented role. Based on these findings, the study presents both theoretical and practical implications.