Article

Examining the Association Between Social Media Use and Self-Reported Social Energy Depletion: A Machine Learning Approach 

---

Abstract

The pervasive integration of social media and digital technologies into daily life has been linked to a rise in digital fatigue and burnout. While the quantity of screen time is often implicated, the more nuanced behavioral patterns contributing to this phenomenon, termed "social energy depletion," are less understood. This study sought to move beyond simple correlational analyses by employing a machine learning framework to predict social energy depletion and identify its most significant behavioral predictors. This quantitative, cross-sectional study utilized a dataset of self-reported digital behavior and psychological metrics from 500 participants. A binary target variable for social energy depletion was created using a median split on a self-reported mood score. A key aspect of the methodology was feature engineering, creating interaction and polynomial features (e.g., screen_focus_ratio, sm_time_squared) to capture the context and non-linear effects of digital use. Three supervised machine learning models (Logistic Regression, Random Forest, Gradient Boosting) were trained and optimized using GridSearchCV to classify the presence of social energy depletion. The Gradient Boosting Classifier achieved the highest predictive accuracy at 58.7%. While the overall predictive power was modest, the feature importance analysis yielded the study's central finding. Engineered features representing the quality of attention (screen_focus_ratio) and the non-linear impact of social media duration (sm_time_squared) were among the most influential predictors, ranking third and fourth, respectively. This demonstrates that the context of digital engagement is a critical factor alongside the total volume of screen time. Predicting a complex psychological state like social energy depletion from behavioral data is challenging. However, this study successfully demonstrates that a machine learning approach can uncover complex, non-linear patterns that traditional analyses may miss. The findings strongly suggest that the quality and nature of digital engagement, not just the quantity, are key drivers of digital fatigue.

Keywords: Digital Well-being, Feature Engineering, Machine Learning, Screen Time, Social Media

How to Cite:

Kumar P, V., S, P. & Batumalay, M., (2025) “Examining the Association Between Social Media Use and Self-Reported Social Energy Depletion: A Machine Learning Approach ”, Journal of Digital Society 1(3), 216-229. doi: https://doi.org/10.63913/jds.v1i3.50