Validating game-based measures of implicit science learning

Rowe, E., Asbell-Clarke, J., Eagle, M., Hicks, A., Barnes, T., Brown, R., & Edwards, T.
Paper presented at the Ninth international conference on Educational Data Mining, Raleigh, NC. June 29 - July 2, 2016


Building on prior work visualizing player behavior using interaction networks, we examined whether measures of implicit science learning collected during gameplay were significantly related to changes in external pre-post assessments of the same constructs. As part of a national implementation study, we collected data from 329 high school students playing an optics puzzle game, Quantum Spectre, and modeled their gameplay as an interaction network, examining errors hypothesized to be related to a lack of implicit understanding of the science concepts embedded in the game. Hierarchical linear modeling (HLM) showed a negative relationship between the science errors identified during gameplay and implicit science learning. These results suggest Quantum Spectre gameplay behaviors are valid assessments of implicit science learning. Implications for how gameplay data might inform classroom teaching in-game scaffolding is discussed.