Symptoms and Their Temporal Distributions: An Interpretable AI Approach for Depression Detection in Social Media

Depression is a common mental disorder involving a depressed mood or loss of pleasure for long periods, which induces grave financial and societal ramifications. Social media-based depression detection is an effective method for early intervention to mitigate those consequences. Such a high-stake decision inherently necessitates interpretability. Although a few studies explain this decision based on the importance of linguistic or demographic features, these explanations do not directly relate to depression diagnosis criteria that are based on symptoms. To fill this gap, we develop a Focused Temporal Prototype Network (FTPNet) to detect depression and provide interpretations based on depressive symptoms as well as their temporal distributions. Extensive evaluations using large-scale datasets show that FTPNet outperforms comprehensive benchmark methods with an F1-score of 0.864. Our result also reveals fine-grained and emerging manifestations of depressive symptoms, such as sharing admiration for a different life, that are unnoted in traditional depression surveys like the Patient Health Questionnaire-9 (PHQ-9). We further conduct a user study to demonstrate improved interpretability over the benchmark. This study contributes to the Information Systems (IS) literature by introducing an interpretable depression detection approach that models the temporal distribution of depressive symptoms. In practice, multiple stakeholders, such as social media platforms and volunteers, can apply our approach to identify depressed users and deliver targeted assistance.