Adding eye-tracking AOI data to models of representation skills does not improve prediction accuracy

Visual representations are ubiquitous in STEM instruction. Representation skills allow students to use visual representations to learn about concepts. It seems reasonable to hypothesize that we can gather useful information about representation skills from eye-tracking AOI data that assesses how students pay attention to representations. We tested this hypothesis by comparing cognitive models with and without eye-tracking AOI data. Specifically, we used Bayesian Knowledge Tracing and Long Short Term Memory models. We evaluated these models based on their accuracy in predicting students learning of knowledge components that assess representation skills. Eye-tracking AOI data did not improve the prediction accuracy of our cognitive models. We compare our results to prior research to generate hypotheses for future research.