Predicting Performance Using Gaze Alignment in an Augmented Reality Learning Environment

Abstract

Monitoring students’ learning and providing appropriate feedback in augmented reality (AR) environments is crucial, though it presents several challenges. This study examines quantifiable measures of how well students engage with a virtual instructor within an AR learning setting. Thirty-three undergraduate engineering students participated in our study. During the experiment, participants learned about biomechanics and answered a question at the end of each AR learning module. In this study, a baseline of gaze alignment, which reflects ideal alignment with the virtual instructor, was used to predict student performance. We applied the average difference method, which quantifies the distance between a student’s gaze position and predefined baseline coordinates, to predict performance levels. These baseline coordinates represent ideal gaze alignment corresponding to full attention to all critical instructional materials during learning. Logistic regression models were then evaluated to examine the effectiveness of this metric in predicting answer correctness.