Tracing Affect in Digital Learning: Combining Log Data and Introspective Momentary Assessments

Self-regulated learning is a critical 21st-century skill, particularly in e-learning environments, where learners must manage cognitive, motivational, and affective processes with limited external support. Recent research has put emphasis on better understanding and fostering SRL by using self-reports and log data. However, studies that temporally align introspective and behavioral measures within intelligent tutoring systems to examine affective dynamics are missing, limiting our understanding of how affective states unfold during self-regulated learning activities. Therefore, we conducted two complementary empirical studies, combining momentary self-reports with log data at different temporal scales to investigate affective processes during exam preparation in an intelligent tutoring system. In Study 1 ( N learning days = 1046, N learners = 95), we analyzed within-person variability in situational appraisals, affective states, and planned versus actual study behavior at a daily level revealing that appraisals predicted affect and planned study time but not actual behavior. Study 2 ( N momentary assessments = 2136, N learners = 60) provided a complementary perspective by reversing the measurement modalities and using log-based learning success to predict self-reported affect at a finer-grained item level. This study captured moment-to-moment variability in valence and arousal and showed that valence increased and arousal slightly decreased with greater study success. Together, both studies demonstrate that integrating introspective and behavioral data provides a richer, more dynamic understanding of affective processes in self-regulated learning, but also highlight the importance of fine-grained granularity and temporal alignment in SRL research. This approach offers pathways for designing adaptive, learner-centered technologies that respond to students’ affective states in real-time. • Self-regulated learning (SRL) is a key process in e-learning environments • An integrated approach utilizing self-reports and logfile data is needed • Two field studies show benefits and caveats of combining both data sources • Research must address granularity and suitable data sources for distinct SRL phases