Mental Health on Social Media: AI-Driven Detection and Response

Abstract

Mental health issues are increasingly prevalent, with stress playing a critical role in the development of severe mental and physical health conditions. Early detection and effective intervention are essential for mitigating these challenges. In an increasingly digital world, social media serves as a valuable repository of large-scale data on how individuals vent and express stress. This data source captures two critical dimensions or perspectives: the individual and the social. The individual dimension is revealed through direct expressions of stress in users’ posts, where emotional states and linguistic patterns provide important indicators. In a synergistic manner, the social dimension is discerned from the reactions of others, offering contextual cues that reflect the broader environment’s influence on the user’s mental state.

My dissertation builds on this dual perspective by integrating social science and psychological theories to inform a methodologies,that strengthens AI’s capacity to recognize stress-related cues and also to engage with mental health discourse in a refined and contextaware manner. To achieve this, I propose three innovative detection strategies that capture the individual and social dimensions. The first strategy focuses on analyzing the finegrained linguistic and emotional features to identify stress within individual posts, directly addressing the individual perspective. The second strategy extends this analysis by examining the broader contextual nuances embedded in these posts, thereby deepening the understanding of individual stress expressions. The third strategy shifts attention to the social perspective by incorporating emotional cues from community responses as auxiliary signals to enhance the stress classification. Finally, drawing on the insights from these works, I established a data-supported refinement process that improves AI’s ability to produce more supportive responses that are both contextually aware and socially attuned. This research exemplifies how interdisciplinary innovation can redefine AI’s role in addressing complex challenges in mental health.