Uncovering Patterns in Process Data to Analyze Interactions and Learning Outcomes within a Computer-Based Learning Environment

Computer-based learning environments (CBLEs) are powerful tools to support student learning. Increasingly of interest is the data that is recorded as learners interact with a CBLE. This "process data" yields opportunities for researchers to examine learners' engagement with a CBLE and explore whether specific interactions are associated with learner variables, with direct implications for improving learning outcomes and CBLE design. As CBLEs increase in number and complexity, researchers continue to seek more effective strategies to analyze process data. While a variety of strategies are in use, from visualizations to predictive modeling, none yet offer the capabilities to both uncover hidden, meaningful interactions and descriptively analyze those interactions rapidly across the complete data set. This paper details a method that addresses current challenges, and then applies the method to existing data from a prior study which investigated the effects of adding a visual scaffold to a chemistry-based CBLE. Using a biochemical coding approach through a cultural-historical activity theory (CHAT) framework, the method successfully identified 257 unique, meaningful patterns of interaction that were strategically grouped into nine categories of mediated actions. Though no differences in mediated actions were observed between learners in the experimental (visual scaffolds) and control conditions, three mediated actions were significantly and positively associated with higher learning outcomes in the visual scaffold condition. The results not only provide insight into why the addition of visual scaffolds led to higher learning outcomes among participants but have broader implications for filling a gap in the field of process data analytics for CBLEs in science education.