Université de Montréal
Predicting User Learning Performance From Eye Movements During Interaction With a Serious Game
Abstract: This paper explored the relationship between eye movements’ measures and learners’ performance during interaction with Crystal Island, a narrative-centered learning game environment. We gathered gaze data from 20 participants using Tobii Tx300 eye tracker while they were reading books and answering multiple-choices quizzes. Statistical analysis as well as classifications were performed. Random forest classifier reached 70% accuracy and was able to discriminate between the learners who successfully completed the quizzes and the learners who do not, providing thus insight for using eye tracking technique to assess learner’s outcomes.