Understanding Human–AI Interaction in Aging Populations: Usability Differences between Robots and Avatars

As AI systems become more prevalent in elderly care, their form of embodiment, robot or screen-based, may influence user acceptance and engagement. This exploratory study examines how older adults interact with two distinct embodiments of an AI coaching system: the 2D avatar AnnSophie and the humanoid robot, Ameca. Through qualitative interviews with four participants, we examined perceptions of appearance, voice, usability, and integration potential in care settings. Participants generally accepted both systems but noted different strengths. Ann-Sophie was seen as visually pleasant and approachable, though some criticized its limited expressiveness and conversational depth. Ameca elicited stronger emotional responses: while appreciated for its realism and social cues, it occasionally provoked unease and prompted reflections on the role of robotics in caregiving. Across both systems, participants emphasized the importance of understandable language, personalization, and practical benefits. Our findings suggest that embodiment influences user experience, but does not solely determine acceptance. Context, expectation, and perceived usefulness were equally decisive. This points to the need for adaptive, context-aware AI agents and user-centered development strategies. Despite the limited number of participants, the study contributes to the ongoing debate on human-centered AI and provides insights for the design of socially acceptable digital coaching systems for the aging population.