Examining students' cognitive load in the context of self-regulated learning with an intelligent tutoring system

Students process qualitatively and quantitatively different information during the dynamic self-regulated learning (SRL) process, and thus they may experience varying cognitive load in different SRL behaviors. However, there is limited research on the role of cognitive load in SRL. This study examined students' cognitive load in micro-level SRL behaviors and explored how cognitive load affected metacognitive judgments and SRL performance. In this study, thirty-four medical students solved two diagnostic tasks of varying complexity (i.e., simple and complex) in BioWorld, an intelligent tutoring system designed for promoting clinical reasoning skills. Think-aloud protocols (TAPs) were utilized to code students' SRL activities, i.e., Orientation, Planning, Monitoring, Evaluation, and Self-reflection. Using the text mining techniques, we also extracted students' linguistic features from TAPs to represent their cognitive load in each SRL behavior. The results demonstrated that students experienced significantly different cognitive load during different SRL behaviors as they solved the complex task. Moreover, students' cognitive load during the Orientation and Evaluation behavior significantly increased as task complexity increased. Furthermore, the linear mixed-effects model (LMMs) indicated that students' cognitive load in the Orientation and Monitoring behaviors negatively predicted confidence ratings but did not affect diagnostic performance. This study not only provides theoretical and methodological insights about cognitive load during SRL but inspires the design of intelligent tutoring systems regarding the effective self-regulation of cognitive load.