Decoding individual differences in STEM learning from functional MRI data

Abstract

Traditional tests of concept knowledge generate scores to assess how well a learner understands a concept. Here, we investigated whether patterns of brain activity collected during a concept knowledge task could be used to compute a neural 'score' to complement traditional scores of an individual’s conceptual understanding. Using a novel data-driven multivariate neuroimaging approach—informational network analysis—we successfully derived a neural score from patterns of activity across the brain that predicted individual differences in multiple concept knowledge tasks in the physics and engineering domain. These tasks include an fMRI paradigm, as well as two other previously validated concept inventories. The informational network score outperformed alternative neural scores computed using data-driven neuroimaging methods, including multivariate representational similarity analysis. This technique could be applied to quantify concept knowledge in a wide range of domains, including classroom-based education research, machine learning, and other areas of cognitive science.

Teacher-student neural coupling during teaching and learning

“…Overall, these findings suggest that neural coupling between teachers and students can be used as an index of learning. While the precise biological processes that give rise to neural alignment has not yet been elucidated, we suggest that neural alignment reflects shared representation of semantic knowledge (Vodrahalli et al, 2018;Cetron et al, 2019;Nastase et al, 2019b).…”

Think Like an Expert: Neural Alignment Predicts Understanding in Students Taking an Introduction to Computer Science Course

“…For example, parietal and frontal regions have been shown to play a key role in mathematical cognition (Anderson et al, 2011;Dehaene et al, 2004). In physics, concepts such as gravity and frequency have each been associated with a distinctive set of cortical regions, mostly on the lateral cortical surface (Mason and Just, 2016), and recent work using multivariate methods has localized representations of physics concepts to dorsal fronto-parietal regions and ventral visual areas (Cetron et al, 2019). One way to account for these apparent discrepancies and for the prominence of DMN regions in our outcome-based results is to consider the likely role of different cortical regions in learning.…”

“…Recent imaging work has begun addressing this gap, examining the process of learning new concepts and extending a large body of work that has studied changes in neuronal circuits during and after learning (Karuza et al, 2014;McCandliss, 2010). Cetron et al (2019) successfully used a multivariate neuroimaging approach to show that brain activity patterns recorded while students learned new categories in a Newtonian physics task can predict performance in a subsequent behavioral test. In an earlier study, Mason and Just (2015) reported a progression of activation throughout the cortex during learning, providing "snapshots" of the various cortical networks activated as participants progressed through explanations about different mechanical systems.…”

Using the force: STEM knowledge and experience construct shared neural representations of engineering concepts

“…To date, only a few neuroimaging studies have investigated physics and engineering concept knowledge [7][8][9][10] . Results of this research implicate dorsal stream regions-including motor cortex -in the explicit retrieval of task-specific physics knowledge.…”

“…Using multivariate pattern analysis (MVPA) of neuroimaging data, we first identified convergent patterns of neural activity among engineering students and among novices. Then, for each group, we performed an informational network analysis 7 , a variant of representational similarity analysis (RSA 12 ), to query those activity patterns for the presence of concept knowledge information about the mechanical categories of structures. As a control, we also evaluated the contribution of bottom-up perceptual information to these neural patterns using a forward-encoding model of primary visual cortex (HMAX 13 ).…”