Using multi-channel data with multi-level modeling to assess in-game performance during gameplay with Crystal Island

Game-based learning environments (GBLEs) have been touted as the solution for failing educational outcomes. In this study, we address some of these major issues by using multi-level modeling with data from eye movements and log files to examine the cognitive and metacognitive self-regulatory processes used by 50 college students as they read books and completed the associated in-game assessments (concept matrices) while playing the Crystal Island game-based learning environment. Results revealed that participants who read fewer books in total, but read each of them more frequently, and who had low proportions of fixations on books and concept matrices exhibited the strongest performance. Results stress the importance of assessing quality vs. quantity during gameplay, such that it is important to read books in-depth (i.e., quality), compared to reading books once (i.e., quantity). Implications for these findings involve designing adaptive GBLEs that scaffold participants based on their trace data, such that we can model efficient behaviors that lead to successful performance.