Reproduce “SAFE-Teen: Evaluating Chatbot Safety Across Adolescent Mental Health Conversations” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Andrea Mock, Valerie Chen, Eric Lin, Rayan Krishnan, Diyi Yang, Caroline Figueroa
SAFE-Teen tests chatbot safety across simulated adolescent conversations by comparing matched branches in which teens disclose, respond to help offers, or react to chatbot limits. Across 18 configurations, its bundled teen-aware prompt lowered applicable criterion failures from 43.3% to 24.9%, while indirect suicide-risk failures remained at 52%. These results concern clinician-authored scenarios and simulated replies, not actual adolescents.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.