Demographic-Aware Multi-Modal Deep Learning Framework for Digital Behavior and Mental Health Analysis

Digital device usage has become ubiquitous, raising questions about its psychological impact. While existing research has established correlations between screen time and mental health, studies often overlook demographic factors, usage modalities, and temporal patterns. We introduce a demographic-aware multi-modal deep learning framework that integrates a recurrent-convolutional hybrid model with demographic embeddings and specialized neural pathways processing different aspects of digital behavior. Using data from 2,000 participants, our framework achieved a 17.3% reduction in mean absolute error for mental health prediction and 12.8% improvement in anxiety classification compared to traditional models. Results revealed significant demographic variations: tablet usage showed stronger associations with anxiety in younger demographics (r = 0.112 in 18-24 vs. r = 0.041 in 55+ group), while social media use correlated differently with depression across genders (r = 0.086 for females vs. r = 0.027 for males). Geographic location also emerged as significant, underscoring the necessity of demographic-aware approaches to digital wellness research.