Explainable AI-driven depression detection from social media using natural language processing and black box machine...

Area of Science:

• Computational psychiatry
• Artificial intelligence in mental health

Background:

• Mental disorders, particularly depression, impose significant burdens on individuals and society.
• Social media offers a rich source of user-generated data for computational mental health research.
• Early detection of depression is crucial for timely intervention and improved outcomes.

Purpose of the Study:

• To explore the early detection of depression using machine learning (ML) models on social media data.
• To integrate Explainable AI (XAI) methods to enhance the interpretability of black-box ML models.
• To evaluate the combined predictive performance and interpretability of ML and XAI for depression detection.

Main Methods:

• Utilized black-box ML models: Support Vector Machines (SVM), Random Forests (RF), Extreme Gradient Boosting (XGB), and Artificial Neural Networks (ANN).
• Employed Natural Language Processing (NLP) techniques including TF-IDF, LDA, N-grams, BoW, and GloVe embeddings for feature extraction.
• Integrated Local Interpretable Model-Agnostic Explanations (LIME) to provide insights into model predictions.

Main Results:

• Support Vector Machines (SVM) demonstrated the highest accuracy in detecting depression from social media content.
• LIME successfully provided granular explanations for model predictions, identifying key linguistic markers.
• The identified linguistic markers aligned with established psychological research on depression.

Conclusions:

• This study highlights the efficacy of SVM for depression detection in social media data.
• The integration of LIME significantly improved the interpretability and clinical trustworthiness of ML models.
• Combining predictive accuracy with explainability is crucial for advancing computational approaches in mental health.