AI-Based Screen Addiction Risk Prediction in Toddlers Using Behavioral Data

Abstract

Excessive screen exposure in early childhood has become a growing pediatric and public-health concern, especially because toddlers are in a rapid period of language, socioemotional, and executive-function development. The WHO recommends no screen time for children younger than 1 year and no more than 1 hour per day for children aged 1 to 4 years, within broader 24-hour movement guidance for children under 5. Recent studies report that screen exposure above this threshold is associated with poorer language development, behavioral difficulties, and weaker caregiver-child interaction quality in toddlers and preschoolers. This paper proposes a master’s-level research framework for predicting screen addiction risk in toddlers using behavioral data collected from parents, caregivers, and digital-use logs. The framework combines structured behavioral features, duration-pattern features, and context-aware indicators such as co-viewing, device type, bedtime use, and interruption resistance, then applies supervised machine learning for early risk stratification. A feature pipeline and evaluation design are presented for binary and ordinal risk classification using logistic regression, random forest, gradient boosting, and calibrated neural models, with AUROC, F1-score, and confusion matrices used for assessment. The paper further discusses ethical safeguards, clinical limitations, and translational implications for pediatric screening tools, early-intervention apps, and parental monitoring systems.pmc.ncbi.nlm.nih+8

Keywords