Passive behavioral sensing: Using within-person variability features from mobile sensing to assess self-regulated learning

Abstract

Self-regulated learning (SRL) significantly influences students' learning behaviors and academic performance. However, research has focused on "between-person" differences, neglecting "within-person" variability. Traditional SRL assessments rely on self-reports, which fail to capture fine-grained behavioral changes (such as hourly variations). We propose a novel approach using mobile sensing to assess SRL through within-person variability. We use passive sensing data from the phones of 211 university students to explore this relationship. To assess behavioral variability, we focus on five sensing behaviors—physical activity, social interactions, sleep, location, and app usage data—and calculate four within-person variability features: standard deviation, circadian rhythm, regularity index, and flexible regularity index. Our findings reveal significant associations between these variability features and self-reported SRL skills, particularly in dimensions such as environment structuring, time management, and help seeking. This research provides new insights into assessing SRL and offers a theoretical foundation for future personalized interventions in educational settings.