Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Depression Detection on Social Media with Large Language Models

Xiaochong Lan, Zhiguang Han, Yiming Cheng, Li Sheng, Jie Feng, Chen Gao, Yong Li

TL;DR

This paper addresses the challenge of detecting depression from social media while requiring medical grounding and model explainability. It proposes DORIS, a hybrid framework that uses LLMs to annotate posts against DSM-5 criteria and to summarize mood trajectories, which are then fused with a Gradient Boosting Trees classifier for accurate predictions. The approach demonstrates a substantial AUPRC gain over strong baselines on the SWDD dataset and generalizes to the Twitter Mental Disorder dataset, with ablations confirming the critical role of diagnostic-criteria features and mood-course representations. By delivering evidence-based explanations grounded in clinical knowledge, DORIS offers a practical, clinician-supportive tool that can enhance large-scale depression screening in low-resource settings.

Abstract

Limited access to mental healthcare resources hinders timely depression diagnosis, leading to detrimental outcomes. Social media platforms present a valuable data source for early detection, yet this task faces two significant challenges: 1) the need for medical knowledge to distinguish clinical depression from transient mood changes, and 2) the dual requirement for high accuracy and model explainability. To address this, we propose DORIS, a framework that leverages Large Language Models (LLMs). To integrate medical knowledge, DORIS utilizes LLMs to annotate user texts against established medical diagnostic criteria and to summarize historical posts into temporal mood courses. These medically-informed features are then used to train an accurate Gradient Boosting Tree (GBT) classifier. Explainability is achieved by generating justifications for predictions based on the LLM-derived symptom annotations and mood course analyses. Extensive experimental results validate the effectiveness as well as interpretability of our method, highlighting its potential as a supportive clinical tool.

Depression Detection on Social Media with Large Language Models

TL;DR

This paper addresses the challenge of detecting depression from social media while requiring medical grounding and model explainability. It proposes DORIS, a hybrid framework that uses LLMs to annotate posts against DSM-5 criteria and to summarize mood trajectories, which are then fused with a Gradient Boosting Trees classifier for accurate predictions. The approach demonstrates a substantial AUPRC gain over strong baselines on the SWDD dataset and generalizes to the Twitter Mental Disorder dataset, with ablations confirming the critical role of diagnostic-criteria features and mood-course representations. By delivering evidence-based explanations grounded in clinical knowledge, DORIS offers a practical, clinician-supportive tool that can enhance large-scale depression screening in low-resource settings.

Abstract

Limited access to mental healthcare resources hinders timely depression diagnosis, leading to detrimental outcomes. Social media platforms present a valuable data source for early detection, yet this task faces two significant challenges: 1) the need for medical knowledge to distinguish clinical depression from transient mood changes, and 2) the dual requirement for high accuracy and model explainability. To address this, we propose DORIS, a framework that leverages Large Language Models (LLMs). To integrate medical knowledge, DORIS utilizes LLMs to annotate user texts against established medical diagnostic criteria and to summarize historical posts into temporal mood courses. These medically-informed features are then used to train an accurate Gradient Boosting Tree (GBT) classifier. Explainability is achieved by generating justifications for predictions based on the LLM-derived symptom annotations and mood course analyses. Extensive experimental results validate the effectiveness as well as interpretability of our method, highlighting its potential as a supportive clinical tool.
Paper Structure (57 sections, 13 equations, 4 figures, 5 tables)

This paper contains 57 sections, 13 equations, 4 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Methodology
  3. Problem Formulation
  4. Diagnostic Criteria Feature Extraction
  5. Annotation with LLM
  6. Annotation with LLM
  7. Efficient Annotation.
  8. Mood Course Representation Construction
  9. Posts Filtering
  10. Representation Construction
  11. Post History Representation Construction
  12. Training and Predicting
  13. Design for Explainability
  14. Experimental Setup
  15. Implementation Details
  16. ...and 42 more sections

Figures (4)

  • Figure 1: Depression detection on social media analyzes users' social media post history to identify potential depression, offering insights to support mental health professionals in depression assessment.
  • Figure 2: Illustration of DORIS. Through the collaboration of the LLM and the text embedding model, we obtain three key features: depression symptom feature, post history representation, and mood course representation. The classifier uses these three features to make its judgment; the LLM uses annotations of depression symptoms and descriptions of the mood course to generate explanation for the system's decision.
  • Figure 3: A case study of DORIS's output.
  • Figure 4: Results of hyperparameter study.