Early methods for studying affective use and emotional well-being on ChatGPT

Source: OpenAI Date: 2025-03-21 URL: https://openai.com/index/affective-use-study

Summary

OpenAI research on studying how users interact with ChatGPT for emotional support and what effects this has on well-being — developing early methodologies for measuring “affective use” (using AI for emotional regulation, companionship, or therapeutic conversations). The paper likely covers survey methods, conversation analysis techniques, and early findings on whether AI emotional support is net positive or negative.

Implications

Safety/well-being thread. Measuring emotional well-being effects of AI use is methodologically hard and commercially sensitive — findings that show negative effects (increased loneliness, reduced human connection) create regulatory and reputational pressure. OpenAI publishing this methodological work signals they’re taking the measurement problem seriously before publishing substantive findings. The study design here likely informed the Expert Council on Well-being and AI (October 2025) and the mental health update (February 2026). The research is genuine but also positions OpenAI as the responsible actor studying its own effects.