Abstract
Large language models (LLMs) are increasingly used as private, always-available conversational systems, yet little is known about how individuals with depressive symptoms utilize them. Building on CSCW research regarding disclosure and peer support, we examine ChatGPT as an emerging informal support infrastructure: private, persistent, responsive, and available outside regular hours.
We analyzed 187,093 ChatGPT conversations from 766 participants who completed the PHQ-8, comparing those below the moderate symptom threshold (score of 10) with those at or above it. Higher-PHQ participants engaged ChatGPT more for mental health, interpersonal, loneliness, self-focused, and support-seeking conversations, with notable late-night and recurring monthly patterns. Their language featured more first-person singular pronouns and absolutist terms. They also engaged in high-disclosure contexts more frequently, although professional redirection was not higher. Language-based prediction was modest and insufficient for screening (AUROC 0.591). We argue that these histories should not be treated as clinical screening data but as evidence that LLMs are increasingly utilized as informal support infrastructure.
Blogger's Review: This study reveals the potential of large language models in providing mental health support, particularly for individuals with depression, highlighting how LLMs can serve as a new form of informal support tool. However, the research also underscores the limitations of language pattern analysis, reminding us to exercise caution when interpreting such data and not to treat it as a substitute for professional healthcare.