A LLM-Based Chatbot for Mindfulness Practice with Older Adults: A Development and Usability Study.
Lucy McCarren, Sanna Kuoppamäki
Studies in health technology and informatics May 15, 2025 DOI: 10.3233/SHTI250493 via PubMed
Summary
AI-generated from the abstractA chatbot powered by a large language model (LLM) can support mindfulness practice among older adults. In a pre-study, 15 older adults used a commercial mindfulness app with human guidance for three weeks, and their feedback highlighted the importance of voice, personality, interactivity, and ease of use. Based on these insights, a prototype AI mindfulness coach was built by fine-tuning OpenAI's ChatGPT on a custom dataset of mindfulness scripts. The prototype was then assessed through participatory design workshops, suggesting that LLM-based chatbots may offer a promising avenue for delivering mindfulness support tailored to older adults.
Study at a glance
| Characteristics | Qualitative study Peer reviewed |
|---|---|
| Sample size | 15 |
| Population | Older adults |
| Interventions | Mindfulness app with recorded human guidance AI mindfulness coach |
| Duration | Three weeks |
| Keywords | Ai in healthcare Chatgpt Chatbot Large language models Digital health |
| Key finding | Qualitative feedback from older adults using a mindfulness app identified key design themes—voice, personality, interactivity, and ease of use—that informed the development of a prototype AI mindfulness coach. |
Abstract
This study explores the development and usability of a Large Language Model (LLM)-based chatbot designed to support mindfulness practice for older adults. In a pre-study, 15 older adults used a commercially available mindfulness app with recorded human guidance for three weeks. Qualitative feedback revealed key design themes including voice, personality, interactivity, and ease of use. Building on the findings of the pre-study, we developed a prototype AI mindfulness coach by fine-tuning OpenAI's ChatGPT on a custom dataset of mindfulness scripts. The AI coach was evaluated through participatory design workshops.