The Relative Importance of Self-Regulated Learning, Emotions, and Cognitive Load in Clinical Reasoning

• In Event: Data Mining and Self-Regulated Learning: Aligning Constructs With Measurement

Abstract

Purpose
This paper addresses the role that cognitive load and emotions play in self-regulated learning in medicine. In particular, we examine medical students as they establish a clinical diagnosis in the context of solving clinical cases in BioWorld (Lajoie, 2009), a computer supported learning environment designed for medical students to deliberately practice (Ericsson, 2004) their diagnostic reasoning strategies while solving virtual cases. We are interested in how SRL (self-regulated learning) strategies, cognitive load, and emotions jointly affected diagnostic performance.

Method
Twenty-six medical students volunteered for the study and solved 3 patient cases of varying difficulty, resulting in 78 cases that were analyzed. Think alouds were collected as students solved patient cases resulting in 78 transcripts. These transcripts were segmented into meaning units and coded for cognitive load and SRL strategies. For cognitive load, two raters coded 30% of the transcripts. We then used five types of supervised machine learning algorithms (i.e., Naïve Bayes, Decision Tree, Deep Learning, Support Vector Machines, and Rule Induction) to predict the variables of interest for the rest of the think-aloud protocols based on their linguistic features extracted by the Linguistic Inquiry and Word Count (LIWC) program (see Table 2-3) (Pennebaker, Boyd, Jordan, & Blackburn, 2015). We calculated the relative importance of SRL strategies, cognitive load, and five types of emotions (pride, anger, shame, relief, enjoyment) in predicting diagnostic performance (i.e., efficiency and confidence) using the method developed by Lideman, Merenda, and Gold (1980). Specifically, we built two models named efficiency model and confidence model, whereby we examined the contributions of the seven predictors to diagnostic efficiency and confidence, respectively. To test whether one predictor is significantly more important than the other, we conducted 1000 bootstrap replicates at 95% confidence level.

Results
The results in Table 4 showed that the SRL strategies, the emotions of pride and enjoyment positively predicted diagnostic efficiency, while the emotions of anger, shame, relief, and cognitive load negatively predicted diagnostic efficiency. In terms of their relative importance in predicting efficiency, the results revealed that the emotions of anger, pride, and shame were the three most influential factors, followed by SRL strategies. However, there were no statistically significant differences between these predictors as shown in Table 5.
In addition, the emotion of pride and SRL strategies positively predicted diagnostic confidence while the other variables negatively predicted confidence (see Table 6). Specifically, the emotions of shame, pride, and anger influenced students’ confidence the most. As shown in Table 7, the emotion of pride was statistically significantly and more important than relief, enjoyment, cognitive load, and SRL strategies in terms of predicting confidence. Moreover, the emotion of shame was significantly more important than relief, cognitive load, and SRL strategies.

Significance
In general, SRL strategies and positive emotions contributed positively to diagnostic performance and negative emotions contributed negatively to performance. Cognitive load contributed negatively to both diagnostic efficiency and confidence. In terms of the relative importance of the variables, SRL strategies, emotions, and cognitive load functioned differently in predicting different aspects of diagnostic performance.

Authors

• Susanne P. Lajoie, McGill University

• Shan Li, McGill University

• Juan Zheng, McGill University

• Tianshu Li, McGill University

• Alejandra Ruiz-Segura, McGill University

• Kirsten Nynych, McGill University