Date of Original Version
Abstract or Description
The rich interaction space of many educational games presents a challenge for designers and researchers who strive to help players achieve specific learning outcomes. Giving players a large amount of freedom over how they perform a complex game task makes it difficult to anticipate what they will do. In order to address this issue designers must ask: what are students doing in my game? And does it embody what I intended them to learn? To answer these questions, designers need methods to expose the details of student play. We describe our approach for automatic extraction of conceptual features from logs of student play sessions within an open educational game utilizing a two-dimensional context-free grammar. We demonstrate how these features can be used to cluster student solutions in the educational game RumbleBlocks. Using these clusters, we explore the range of solutions and measure how many students use the designers’ envisioned solution. Equipped with this information, designers and researchers can focus redesign efforts to areas in the game where discrepancies exist between the designers’ intentions and player experiences.
Proceedings of The 6th International Conference on Educational Data Mining.