Mining Sequences of Gameplay for Embedded Assessment in Collaborative Learning

This poster presents a sequence mining analysis of collaborative game-based learning for middle school computer science. Using pre-post test results, dyads were categorized into three groups based on learning gains. We then built first-order Markov models for the gameplay sequences. The models perform well for embedded assessment, classifying gameplay sequences with 95% accuracy according to whether the group learned the target concepts or not. These results lay the groundwork for accurate embedded assessment of dyads in game-based learning.