Does AI Outperform Humans? An Empirical Study of Social Media Addiction Notifications
(A2025-130319)

AUTHORS

Iman Goodarzi, Concordia University; Hamid Shirdastian, Bishop's University; Michel Laroche, John Molson School of Business, Concordia University; Michele Paulin, Concordia University

ABSTRACT

Social media addiction (SMA) is increasingly recognized as a public health concern, yet it remains unclassified as a formal disorder. Existing solutions emphasize self-control and mindfulness, overlooking technology-based interventions. Grounded in the Conservation of Resources (COR) and Feedback Intervention Theory (FIT), this research investigates whether AI-driven or human-generated notifications, via text or voice, can reduce excessive social media usage. Across two studies (N=856), we examine how individual SMA tendencies moderate these effects. Contrary to linear predictions, initial results show no direct main effects of source or modality. However, non-linear analyses reveal a conditional influence, with moderate withdrawal symptoms responding more favorably to AI prompts. These findings highlight the value of personalized, context-aware interventions in promoting digital well-being. By combining COR and FIT, our research offers practical insights for ethical, user-centred platform design and broader efforts to address compulsive social media use.