Stressed or engaged? Addressing the mixed significance of physiological activity during constructivist learning

The current paper qualifies previous assumptions on the link between physiology and learner states during constructivist learning, employing a study design where electrodermal activity data (EDA) is collected across two distinctive pedagogies - direct instruction and hands-on learning. Our results are surprising in that higher arousal during hands-on learning is not indicative of higher learning gains, nor is the strength of the EDA response clearly different when transitioning across pedagogies. However, qualitative investigations show that varying sources of physiological arousal likely to be present in hands-on learning, such as stress, may have led to this mixed connection between EDA and learning during hands-on learning. Moreover, identical tests for the direct instruction section show that higher EDA during this more passive form of learning is indeed indicative of higher learning gains. Our machine learning prediction models achieved fair performance for both learning gains and engagement, providing support for the use of physiological data in gauging and supporting learner states despite its polysemy. We conclude with informed suggestions for building theory-based, nuanced connections between EDA and constructivist learning.

Automated summary

This study reevaluates the link between physiology and learner states during constructivist learning by collecting electrodermal activity data (EDA) in direct instruction and hands-on learning. Surprisingly, higher arousal during hands-on learning does not correspond to higher learning gains, and the strength of EDA response does not differ significantly across pedagogies. Qualitative investigations suggest that varying sources of physiological arousal, such as stress, in hands-on learning may contribute to the mixed connection between EDA and learning. However, higher EDA during direct instruction is indeed indicative of higher learning gains. Machine learning prediction models perform well in predicting learning gains and engagement using physiological data, supporting its use in assessing and supporting learner states despite its multiple meanings. In conclusion, we provide informed suggestions for establishing theory-based, nuanced connections between EDA and constructivist learning.