Proceedings of the Annual Meeting of the Cognitive Science Society

Early detection of self-regulation challenges in university students is critical for academic success, yet conventional assessment methods struggle to capture these behavioral processes' multidimensional and dynamic nature. To address this, we develop iSense - a smartphone-based system that enables privacyaware, passive monitoring of self-regulated learning (SRL) and academic emotional regulation (AER) through multi-modal behavioral sensing (social interactions, mobility, app usage, and sleep). In our 12-month longitudinal study involving 211 college students, the system successfully identified significant behavioral correlates of SRL/AER states. Our hybrid populationpersonalized prediction model achieved mean absolute errors of 7.9% (SRL) and 9.3% (AER), representing a 22% improvement over conventional baselines. Importantly, the results demonstrate how knowledge transfer from student populations can facilitate the development of accurate, personalized models that require minimal individual-specific data for rapid adaptation to new students. This scalable solution overcomes limitations of traditional self-report methods for diverse educational settings.