APS
2026 APS Annual Convention · 2026
Negative Affect As an Early Linguistic Signal of Health Distress In AI-Mediated Journaling
Session: Poster Session XIII
Authors
- Hongyu MaPresenting Author
Stanford University - Xianglu Tang
Stanford University - Xingyu Li
Stanford University
Abstract
We examined conversations from 5,000 people using a mental-health AI assistant to understand how they express emotion, distress, and personal values. Negative emotion often coincided with fatigue or illness, while a positive tone aligned with wellbeing and relationships. Early language patterns showed small but reliable links to later engagement.