Network dynamics between digital media use and depressive symptomatology: A computational phenotyping approach using ecological momentary assessment.

BACKGROUND

Despite extensive investigation into relationships between digital media use and psychological well-being, understanding of specific temporal dynamics and underlying mechanisms remains inadequate.

METHODS

This study employed intensive longitudinal methods to examine the complex network relationships between digital behavior patterns and depressive symptoms. We collected ecological momentary assessment (EMA) data from 1642 young adults (ages 18-29, 58.2 % female, 42.3 % currently enrolled in undergraduate programs) across 60 days, yielding 358,962 observations capturing smartphone usage metrics, digital engagement characteristics, and momentary psychological states. Using multilevel vector autoregression modeling with L1-regularization, we constructed temporal networks representing dynamic associations between digital media variables and depressive symptoms.

RESULTS

Computational phenotyping analyses revealed four distinct network typologies: (1) "Social Comparison" (28.4 % of sample, 68.7 % female) characterized by strong bidirectional connections between social media browsing and negative self-evaluation (network density = 0.31); (2) "Temporal Displacement" (23.7 %) marked by robust connections between nocturnal usage and subsequent psychomotor-cognitive symptoms (density = 0.26); (3) "Passive Consumption" (31.2 %) defined by relationships between non-interactive content viewing and anhedonic symptoms (density = 0.22); and (4) "Active Communication" (16.7 %) showing protective pathways between communication-focused usage and reduced loneliness (density = 0.17). Network centrality parameters significantly correlated with longitudinal symptom trajectories (R² = 0.39), with betweenness centrality of passive social media nodes demonstrating particular prognostic value (β = 0.47, p < 0.001). Gender differences emerged in network structures, with females showing stronger social media-rumination associations (β = 0.21 vs. 0.13, p = 0.003). Time-varying effect analyses indicated network structures were moderated by environmental context, with significantly stronger digital media-depression associations during periods of social isolation (coefficient multiplier = 1.86, p < 0.001).

CONCLUSIONS

These findings advance understanding of heterogeneous pathways linking digital media use to psychological functioning in community samples, identifying computational phenotypes that may inform personalized digital well-being interventions for subclinical depressive symptoms.