Fostering Social Connection and Emotional Engagement to Mitigate Social Isolation in Care Homes Using Empathetic Human-Robot Conversation

Social isolation is a critical health concern for elderly individuals, particularly in institutional care settings, where limited social interaction can lead to depression, cognitive decline and physical ill health. Empathetic human-robot interaction (HRI) presents a promising solution to address these emotional and mental health challenges. This paper reports how social robots powered by prompted Large Language Models (LLMs) can provide emotional support through empathetic conversations. We validated these capabilities in a controlled experiment with 40 participants, followed by a 10-week study in a nursing home, where five elderly residents engaged in free-form conversations with the robot lasting up to 45 minutes. The results demonstrate that LLM-powered robots can adapt to users’ emotions, build trust, and provide emotional support over time and that, for the first time, autonomous empathetic spoken language support is possible.