Reproduce “Depression Symptoms and Relational Patterns in 187k ChatGPT Histories” 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

Neil K. R. Sehgal, Dunigan Folk, Lyle Ungar, Sharath Chandra Guntuku

Researchers at the University of Pennsylvania analyzed 187,093 ChatGPT conversation histories, finding that users with higher depressive symptoms engage more in mental health-related disclosure and support-seeking, particularly during late-night hours, but do not receive significantly more professional redirection from the AI. The study also determined that linguistic cues in these conversations are insufficient for clinical screening.

Key insights

  • ChatGPT functions as an informal, continuously available support infrastructure, particularly for individuals experiencing depressive symptoms.
  • Users with higher depressive symptoms exhibit distinct interaction patterns, including increased disclosure, support-seeking, and late-night engagement.
  • The AI's professional redirection rate does not robustly scale with the apparent need or disclosure level of users with higher depressive symptoms, indicating a boundary-setting gap.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.