Quantifying digital habits

Abstract Digital habits remain poorly quantified despite ever-present concerns regarding problematic social media use. Through a mixed-methods study of 6,816 participants and 12,899 app-specific responses across six major platforms, we compare self-reported habit measures (via the Self-Reported Habit Index) with longitudinal behavioural tracking data collected over a 6 week period. Our analysis reveals three key insights: First, app usage shows strong habitual patterns (Facebook: mean SRHI = 4.88), exceeding established benchmarks for health-related habits like smoking. Second, we find weak correlations (r=0.28-0.32) between SRHI scores and actual usage metrics, exposing significant discrepancies in self-assessment accuracy. Third, machine learning demonstrates that simple behavioural history (sessions/hours) predicts future usage at least 64% more effectively than SRHI measures. These results challenge the predictive validity of self-report instruments in digital habit research and suggest observational data offers superior predictive power. Our findings have immediate implications for intervention design, platform accountability, and the methodological evolution of habit research in computational social science.