Presentation
Usability of Artificial Intelligence Systems in Mental Health Settings
SessionWednesday Poster Session
DescriptionThe growing use of AI systems as mental health aids has gained media attention, often accompanied by concern over potential harms, including links to extreme outcomes. These risks highlight the need to improve AI safety while maintaining usability, especially for developers aiming to create effective tools. However, limited research has examined how usability is influenced by an AI system’s perceived helpfulness in mental health contexts. This study aims to address that gap by exploring the relationship between counseling quality and usability. Existing research has focused more on clinician-AI interaction or user satisfaction rather than overall usability. Prior findings suggest AI can provide general guidance but may lack depth, empathy, and contextual understanding, limiting its effectiveness. This study uses a custom GPT to generate transcripts of “helpful” and “unhelpful” AI counseling interactions across common student concerns. Helpful models emphasize psychoeducation, transparency, and cognitive reframing, while unhelpful models provide vague, overly agreeable responses. Participants evaluate these interactions based on counseling quality, usability, and personal reactions. The findings aim to demonstrate that interaction quality directly influences usability perceptions. This research can guide developers and researchers in designing more effective, user-centered AI mental health systems.
Event Type
Poster
TimeWednesday, October 21st5:30pm - 6:30pm PDT
Location
Similar Presentations