Human–Avatar Interaction: Perceptions of Voice and Appearance Among Older Adults in a Residential Care Setting

Abstract

This study explores the acceptance and perception of a conversational AI avatar among older adults in a residential care setting. Against the backdrop of increasing interest in human-AI collaboration in health and social care, we developed a photorealistic avatar named Ann-Sophie, designed to support well-being through everyday conversation. The study involved 14 participants across two iterative rounds and employed a mixed-methods approach, combining questionnaires with semi-structured interviews. Results show that perceptions of the avatar’s voice and appearance depend on individual experiences and needs. While some appreciated its lifelike features, others found it emotionally distant. Clarity and responsiveness of the voice, rather than realism, were key to engagement. The study emphasizes the importance of usability and context-sensitive design over mere aesthetics. These findings offer design implications for voice-based AI systems in care settings and call for further exploration of embodied AI in shared human contexts, ensuring that such systems move beyond novelty to sustainable adoption.