Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

They Look Like Each Other: Case-based Reasoning for Explainable Depression Detection on Twitter using Large Language Models

Mohammad Saeid Mahdavinejad, Peyman Adibi, Amirhassan Monadjemi, Pascal Hitzler

TL;DR

ProtoDep tackles the lack of interpretability in social media depression detection by integrating prototype learning with LLM-generated symptom explanations, enabling symptom-level, case-based, and weight-based explanations while delivering competitive detection performance. The framework constructs PHQ-9–grounded symptom prototypes, learns user prototypes, and uses a multi-view similarity-based classifier, achieving a 94.4% average F1 across five datasets and yielding meaningful prototypes validated by lexical alignment and PRIDE scores. It also reveals that attention alone may not explain final predictions, underscoring the value of explicit prototypes for transparent reasoning. The approach promises to enhance trust and clinical utility in depression screening on social media and can be extended to other platforms and mental health domains.

Abstract

Depression is a common mental health issue that requires prompt diagnosis and treatment. Despite the promise of social media data for depression detection, the opacity of employed deep learning models hinders interpretability and raises bias concerns. We address this challenge by introducing ProtoDep, a novel, explainable framework for Twitter-based depression detection. ProtoDep leverages prototype learning and the generative power of Large Language Models to provide transparent explanations at three levels: (i) symptom-level explanations for each tweet and user, (ii) case-based explanations comparing the user to similar individuals, and (iii) transparent decision-making through classification weights. Evaluated on five benchmark datasets, ProtoDep achieves near state-of-the-art performance while learning meaningful prototypes. This multi-faceted approach offers significant potential to enhance the reliability and transparency of depression detection on social media, ultimately aiding mental health professionals in delivering more informed care.

They Look Like Each Other: Case-based Reasoning for Explainable Depression Detection on Twitter using Large Language Models

TL;DR

ProtoDep tackles the lack of interpretability in social media depression detection by integrating prototype learning with LLM-generated symptom explanations, enabling symptom-level, case-based, and weight-based explanations while delivering competitive detection performance. The framework constructs PHQ-9–grounded symptom prototypes, learns user prototypes, and uses a multi-view similarity-based classifier, achieving a 94.4% average F1 across five datasets and yielding meaningful prototypes validated by lexical alignment and PRIDE scores. It also reveals that attention alone may not explain final predictions, underscoring the value of explicit prototypes for transparent reasoning. The approach promises to enhance trust and clinical utility in depression screening on social media and can be extended to other platforms and mental health domains.

Abstract

Depression is a common mental health issue that requires prompt diagnosis and treatment. Despite the promise of social media data for depression detection, the opacity of employed deep learning models hinders interpretability and raises bias concerns. We address this challenge by introducing ProtoDep, a novel, explainable framework for Twitter-based depression detection. ProtoDep leverages prototype learning and the generative power of Large Language Models to provide transparent explanations at three levels: (i) symptom-level explanations for each tweet and user, (ii) case-based explanations comparing the user to similar individuals, and (iii) transparent decision-making through classification weights. Evaluated on five benchmark datasets, ProtoDep achieves near state-of-the-art performance while learning meaningful prototypes. This multi-faceted approach offers significant potential to enhance the reliability and transparency of depression detection on social media, ultimately aiding mental health professionals in delivering more informed care.
Paper Structure (14 sections, 13 equations, 6 figures, 3 tables)

This paper contains 14 sections, 13 equations, 6 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Method
  4. Preliminaries
  5. Step 1: Embedding User's Tweets
  6. Step 2: Learning Symptom Prototypes
  7. Step 3: Encoding the User
  8. Step 4: Learning User Prototypes
  9. Step 5: Classification
  10. Experiment Results
  11. Ablation Study
  12. Conclusion
  13. Ethical Consideration
  14. Limitations

Figures (6)

  • Figure 1: Overview of the Proto-Dep
  • Figure 2: (a) Symptom Prototype Layer (b) User Prototype Layer
  • Figure 3: Similarity between learned symptom prototypes and ground truth lexicon.
  • Figure 4: Visualization of the PRIDE score for learned prototypes for ProtoDep, ProtoDep-Acc, and ProtoDep (Sinkhorn) Models.
  • Figure 5: Classification weights (absolute values) for different symptoms over all datasets.
  • ...and 1 more figures