Enhancing Depression Detection in Online Social Networks with Explainable AI (XAI)

This study enhances depression detection through social media text analysis using a BERT-based approach. Traditional ML/DL models struggle to interpret nuanced language and emotions in online interactions. BERT’s contextual embeddings improve sentiment understanding, while Explainable AI (XAI) techniques like LIME and SHAP ensure interpretability. Among tested models, MentalBERT achieved the highest accuracy (96.08%), effectively capturing mental health-related language patterns. This combination of accuracy and transparency makes it a promising tool for real-world mental health applications, enabling better detection and interpretation of psychological cues in social media text for early intervention and support.