Depression Detection on Social Media: A Classification Framework and Research Challenges and Opportunities

Abstract

Abstract Objective: Social media has become a safe space for discussing sensitive topics such as mental disorders. Depression dominates mental disorders globally, and accordingly, depression detection on social media has witnessed significant research advances. This study aims to review the current state-of-the-art research methods and propose a multidimensional framework to describe the current body of literature relating to detecting depression on social media. Method: A study methodology involved selecting papers published between 2011 and 2022 that focused on detecting depression on social media. Three digital libraries were used to find relevant papers: Google Scholar, ACM digital library, and ResearchGate. In selecting literature, two fundamental elements were considered: identifying papers focusing on depression detection and including papers involving social media use. Results: In total, 46 papers were reviewed. Multiple dimensions were analyzed, including input features, social media platforms, disorder and symptomatology, ground truth, and machine learning. Various types of input features were employed for depression detection, including textual, visual, behavioral, temporal, demographic, and spatial features. Among them, visual and spatial features have not been systematically reviewed to support mental health researchers in depression detection. Despite depression's fine-grained disorders, most studies focus on general depression. Conclusion: Recent studies have shown that social media data can be leveraged to identify depressive symptoms. Nevertheless, further research is needed to address issues like depression validation, generalizability, causes identification, and privacy and ethical considerations. An interdisciplinary collaboration between mental health professionals and computer scientists may help detect depression on social media more effectively.