The excessive use of social networks among university students has emerged as a critical factor affecting academic performance and emotional well-being. This research proposes a multimodal framework that integrates artificial intelligence and Human-Computer Interaction technologies to support diagnosis and intervene in Problematic Social Network use. The study adopts a longitudinal experimental design including students from University of Cuenca, divided into control and experimental groups. The framework combines self-reported psychometric data with physiological and behavioral indicators to construct individualized user profiles. It aims to dynamically detect patterns associated with social media addiction and provide context-aware, personalized interventions. The theoretical foundation is based on the Design Science Research, guided by the methodologies of Hevner and Wieringa, ensuring both conceptual consistency and engineering cycle. This research addresses gaps in the current literature, particularly the lack of comprehensive, culturally adaptive, and AI-enabled solutions for digital addiction. The expected contributions highlight the societal impact of the study by promoting mental well-being through the novel integration of affective computing, machine learning, and human-computer interaction into a human-centered, ethical intervention model.