Modeling and Analyzing Inquiry Strategies in Open-Ended Learning Environments

Modeling and predicting student learning in computer-based environments often relies solely on sequences of accuracy data. Previous research suggests that it does not only matter what we learn, but also how we learn. The detection and analysis of learning behavior becomes especially important, when dealing with open-ended exploration environments, which do not dictate prescribed learning sequences and skills. In this paper, we work with data collected from an inquiry-based environment. We demonstrate that 1) students' inquiry strategies indeed influence the learning outcome, and 2) students' inquiry strategies also seem to be predictive for their academic achievement. Furthermore, we identified a new positive inquiry strategy, which has not yet been described in the literature. We propose the use of a probabilistic model jointly representing student knowledge and strategies and show that the inclusion of learning behavior into the model significantly improves prediction of external posttest results compared to only using accuracy data, a result that we validated on a second data set. Furthermore, we cluster the children into different groups with similar learning strategies to get a complete picture of students' inquiry behavior. The obtained clusters can be semantically interpreted and are not only correlated to learning success in the game, but also to students' science grades and standardized math assessments. We also validated the cluster solution on a second data set. The inquiry-based environment together with the clustering solution has the potential to serve as an assessment tool for teachers and tutors.