Researcher(s)
- Farzana Mohseni, Nursing, University of Delaware
Faculty Mentor(s)
- Xiaopeng Ji, School of Nursing/College of Health Sciences, University of Delaware
Abstract
Introduction: Young Black/African American (BAA) adults (vs. other racial/ethnic groups) have worse sleep. Artificial intelligence (AI) chatbots, mimicking human conversations, can automatically deliver sleep interventions. This study explored the usability and accessibility of an AI sleep chatbot intervention for young BAA adults.
Methods: Using a mixed-method approach, 15 BAA adults (18-25 y.o., 11 females) who completed a 4-week sleep chatbot intervention completed an interview and questionnaires. Guided by the AI Chatbot Behavior Change Model, interview questions included chatbot characteristics, relational/persuasive capacity, and perceived outcomes. We performed content analyses of interview data. Participants also completed the System Usability Scale (SUS) and Acceptability Scale.
Results: The average SUS score was 74.16, exceeding the favorable usability cutoff (68). The total acceptability score was 28.13, while all participants rated the overall acceptability question as acceptable/completely acceptable. In interviews, participants found coaching dialogues and sleep logs most useful, while free-input Q&A and resource tabs were less essential. They found the chatbot user-friendly with good design and coloring but noted some texts were lengthy, intervals between texts too fast, and occasional technical issues. The chatbot’s relational capacity was evident in its encouraging, engaging, and personalized conversations and its perceived emotional support. Features enhancing its persuasive capacity included reminders at preferred times, easy-to-understand language, end-of-session knowledge checks, daily sleep logs, graphic sleep dashboard, and progress reviews. Participants consistently reported improved knowledge, sleep self-efficacy, and adoption of sleep hygiene practices, strategies for addressing racing mind and relaxation skills. However, they found changing their sleep schedule and stimulus control strategies difficult to implement.
Conclusion: Sleep chatbot intervention has good acceptability and usability, and can facilitate positive sleep behavioral changes among young BAA adults. An updated chatbot program is needed to troubleshoot technical issues and add features to facilitate sleep schedule changes.