Human-centered Design and Evaluation of a Knowledge-based Conversational Agent for Student Well-being

Abstract

This study extends prior exploratory research on student acceptance and expectations of well-being conversational agents by developing and evaluating a high-fidelity conversational agent using a human-centered AI design approach. The prototype incorporates evidence-based mental health screening instruments, including the Patient Health Questionnaire-9 (PHQ-9) and assessments for anxiety, sleep, and panic symptoms, alongside a structured knowledge base from verified resources from the Anxiety and Depression Association of America (ADAA), the Mayo Clinic, and the American Academy of Sleep Medicine (AASM). An empirical usability study was conducted with twenty university students engaged in task-oriented usability testing and post-interaction interviews to assess system usability, user satisfaction, and perceived conversational intelligence. Findings indicate that participants perceived the conversational agent as credible, easy to use, and emotionally supportive, while expressing persistent concerns regarding privacy and data security. The results substantiate the user-centered design framework articulated in earlier work and demonstrate its evolution into a functional and deployable model for student well-being support. This research provides empirically grounded recommendations for the design of credible, trustworthy, and knowledge-based AI-driven wellness interventions in higher education.