NursingLLM: A Retrieval-Augmented Health Chatbot for Supporting Nursing Home Residents

Abstract

Elderly individuals in nursing homes often experience loneliness and health-related anxiety. To explore how LLM-based chatbots might support this population, we conducted in-depth interviews with 23 elderly residents to understand their emotional and informational needs. Based on these findings, we designed and implemented NursingLLM, a conversational agent tailored for aging-related health support. Guided by principles from HCI and gerontology, the system emphasizes ease of use, trust-building, and empathetic interaction. NursingLLM integrates a multimodal (voice and text) interface to improve accessibility, Retrieval-Augmented Generation (RAG) to provide verifiable health information, and prompting strategies to sustain natural and caring conversations. We then conducted a preliminary user study with 16 residents in two care centers to assess usability and perceived supportiveness. Results suggest that NursingLLM helped users better understand their symptoms, recall daily care routines, and feel more emotionally supported. This work offers design insights and a system prototype for future AI-driven tools to promote older adults' well-being in institutional settings.

Ähnliche Arbeiten