Exploring the Role of Cognitive Load and Self-Regulated Learning in the Context of Diagnostic Reasoning

• In Event: Using Multimodal Trace Data to Support Learners Across Tasks and Contexts

Abstract

Objectives. Self-regulated learning (SRL) and cognitive load (CL) have been investigated separately, but few studies explore their interaction, especially in medical education. We bridge this gap by conducting two studies that examine diagnostic reasoning in a computer-supported learning environment, BioWorld (Lajoie, 2009). Study 1 examines the relationship between CL and engagement in SRL phases (forethought, performance, and self-reflection), diagnostic performance, and confidence. CL in this context is induced by case difficulty, and inferred from linguistic features. Diagnostic performance is a measure of expert efficiency. Study 2 examines mental effort (ME) during diagnosis (an aspect of CL, Paas et al., 2003), metacognitive judgements, and diagnostic performance. The findings can inform scaffolding decisions on how to optimize CL during SRL.

Methods & Results. Study 1. 27 medical students solved three cases of varying difficulty while doing think aloud protocols (TAPs), resulting in data for 81 cases. Two raters coded CL and SRL segments for 30% of TAPs transcripts. Five supervised machine learning algorithms predicted the variables of interest for the remaining TAPs by extracting linguistic features using a text analysis program (Pennebaker et al., 2015). Diagnostic efficiency indicated overlaps between medical students’ and experts’ diagnostic processes. Self-reported confidence rating, a form of metacognitive judgment (Boekaerts, 2017), was collected for each diagnostic hypothesis. RM-ANOVAs revealed that students had significantly higher CL in the performance phase compared to forethought, despite case difficulty. For the easy case, students demonstrated significantly lower CL in self-reflection compared to forethought and performance. Overall, CL did not different significantly across case difficulty but did contribute to specific phases of SRL, diagnostic efficiency and confidence ratings.

Study 2. 88 medical students solved a diagnostic case. Students’ SRL behaviors were extracted from logfiles. Total mental effort (ME_TOTAL) refers to overall time students spent in the entire SRL process; additionally, ME was calculated by SRL phase, including forethought (ME_FO), performance (ME_PE), and self-reflection (ME_SR).

Results of linear regressions demonstrated that ME_TOTAL positively predicted diagnostic efficiency, while the predictive role of ME_TOTAL in confidence ratings was marginally significant (p=.053). According to ME_FO, ME_PE, and ME_SR, latent profile analysis revealed three mental effort profiles, i.e., self-regulator, task enactor, and passive learner. One-way ANOVAs showed significant differences in diagnostic efficiency and confidence ratings across the clusters. Post-hoc analysis revealed that self-regulators had better diagnostic efficiency than enactors and passive learners, and self-regulators reported higher confidence than passive learners.

Significance. SRL is a mediating factor in diagnostic reasoning, leading to higher levels of accuracy in clinical diagnosis. The interaction of SRL with CL can enhance or hinder medical students’ diagnostic performance. Our results show that CL had a specific relationship to phases of SRL, case difficulty, and mental effort. The multimodal data demonstrate the varying effect of CL and ME on metacognitive judgements and diagnostic performance. Understanding the relationship between CL, SRL, confidence, and diagnostic performance can lead to design decisions for adaptation based on distinct student profiles.

Authors

• Susanne P. Lajoie, McGill University
  Presenting Author

• Tingting Wang, McGill University
  Non-Presenting Author

• Alejandra Ruiz-Segura, McGill University
  Non-Presenting Author

• Shan Li, Lehigh University
  Non-Presenting Author