The Embodiment Trap: When AI Has a Body in Education

Artificial Intelligence (AI) is increasingly imagined, designed, and prototyped in embodied forms (e.g., avatars or virtual humans). Within HCI and educational technology, embodiment is commonly framed as a means to improve engagement, presence, or motivation. We argue that such framings are incomplete for educational settings. When AI is given a body in education, embodiment functions as a claim to legitimacy that redistributes pedagogical authority, often without explicit scrutiny. Drawing on prior work in embodied interaction, social presence, AI in education, and critical computing, we identify what we call the embodiment trap; a field-level tendency to treat bodily presence as a design benefit while sidelining its normative consequences. We articulate four tensions introduced by embodied AI in education and examine why these tensions persist. With this paper, we invite the HCI community to reconsider embodiment not as an interface enhancement, but as a normative intervention shaping educational roles, responsibilities, and futures.