ReDepress: A Cognitive Framework for Detecting Depression Relapse from Social Media

2025-09-22

AI Summary

Depression exhibits high relapse rates (up to 80% after a second episode), yet social media–based relapse detection remains underexplored due to scarce annotated data and insufficient theoretical grounding. To address this, we propose ReDepress—the first cognition-driven framework for depression relapse detection—grounded in empirically validated cognitive theories: attentional bias, interpretive bias, memory bias, and rumination. We formalize these constructs into computable cognitive bias features, release ReDepress—the first publicly available, relapse-oriented, annotated social media dataset—and design a Transformer-based temporal model integrating cognitive features. Experiments demonstrate that cognitive markers significantly distinguish relapsing from non-relapsing users (p < 0.01), and our model achieves an F1-score of 0.86. This work pioneers the systematic integration of cognitive theory into computational relapse prediction, establishing a novel, low-cost, and clinically actionable paradigm for early mental health intervention.

Technology Category

• Cognitive Modeling & Cognitive Systems: Social Cognition And Interaction
• Application Domains: Humanities & Computational Social Science
• Machine Learning: Feature Construction/Reformulation

Application Category

• Social Networks and Social Media: Computational social science
• User Modeling, Personalization and Recommendation: Social recommender systems and personalization for the social Web
• Economics, Online Markets and Human Computation: Trust and reliance of crowd workers and data experts on GenAI

Abstract

Almost 50% depression patients face the risk of going into relapse. The risk increases to 80% after the second episode of depression. Although, depression detection from social media has attained considerable attention, depression relapse detection has remained largely unexplored due to the lack of curated datasets and the difficulty of distinguishing relapse and non-relapse users. In this work, we present ReDepress, the first clinically validated social media dataset focused on relapse, comprising 204 Reddit users annotated by mental health professionals. Unlike prior approaches, our framework draws on cognitive theories of depression, incorporating constructs such as attention bias, interpretation bias, memory bias and rumination into both annotation and modeling. Through statistical analyses and machine learning experiments, we demonstrate that cognitive markers significantly differentiate relapse and non-relapse groups, and that models enriched with these features achieve competitive performance, with transformer-based temporal models attaining an F1 of 0.86. Our findings validate psychological theories in real-world textual data and underscore the potential of cognitive-informed computational methods for early relapse detection, paving the way for scalable, low-cost interventions in mental healthcare.

Problem

• Detecting depression relapse from social media data
• Addressing the lack of curated datasets for relapse detection
• Incorporating cognitive theories to differentiate relapse users

Innovation

• Cognitive framework incorporating attention bias and rumination
• Transformer-based temporal models achieving 0.86 F1 score
• First clinically validated social media relapse dataset