🤖 AI Summary
This study investigates the nonlinear causal effect of posting behavior on psychological state evolution among users in Reddit’s depression-related subreddits. To overcome limitations of conventional linear assumptions, we propose a state-transition-based causal pathway identification framework that integrates sequential modeling, regression discontinuity design (RDD), counterfactual reasoning, and survival analysis to capture multi-stage emotional dynamics from long-term user behavioral logs. We uncover, for the first time, a threshold-dependent, nonmonotonic relationship between posting frequency and depression remission: only after users’ affective states cross a critical threshold does high-frequency posting significantly accelerate recovery—empirically validating the “expressive window hypothesis.” Our model achieves an AUC of 0.82 in predicting depression remission, providing actionable, temporally grounded criteria for timing online psychological interventions.
