Assessing Student's Mental Representations of Complex Problem Spaces with EEG Technologies

Abstract

We have developed a neurophysiologic-based assessment of student's understanding of complex problem spaces that blends the population-based advantages of probabilistic performance modeling with the detection of neurophysiologic signals. It is designed to be rapid and effective in complex environments where assessment is often imprecise. Cohorts of novices, and experts encoded chemistry problem spaces by performing a series of online problem solving simulations. The stable memory encoding was verified by comparing their strategies with established probabilistic models of strategic performance. Then, we probed the neural correlates of the encoded problem space by measuring differential EEG signatures that were recorded in response to rapidly presented sequences of chemical reactions that represented different valid or invalid approaches for solving the chemistry problems. We found that experts completed performances in stacks more rapidly than did novices and they also correctly identified a higher percentage of reactions. Event related potentials revealed showed increased positivities in the 100–400 ms following presentation of the image preceding the decision when compared with the other stack images. This neural activity was used to explore reasons why students missed performances in the stack. One situation occurred when students appeared to have a lapse of attention. This was characterized by increased power in the 12–15 Hz range, a decrease in the ERP positivities at 100–400 ms after the final image presentation, and a slower reaction time. A second situation occurred when the students' decisions were almost entirely the reverse of what were expected. These responses were characterized by ERP morphologies similar to those of correct decisions suggesting the student had mistaken one set of chemical reactions for another.