Publication
Analyzing Student Success and Mistakes in Virtual Microscope Structure Search Tasks
Benjamin Paaßen; Andreas Bertsch; Katharina Langer-Fischer; Leo Sylvio Rüdian; Xia Wang; Rupali Sinha; Jakub Kuzilek; Stefan Britsch; Niels Pinkwart
In: François Bouchet; Jill-Jênn Vie; Sharon Hsiao; Sherry Sahebi (Hrsg.). Proceedings of the 15th International Conference on Educational Data Mining. International Conference on Educational Data Mining (EDM-2021), June 29 - July 2, Paris (virtual), France, Pages 1-7, International Educational Datamining Society, 2021.
Abstract
Many modern anatomy curricula teach histology using virtual microscopes, where students inspect tissue slices in a computer program (e.g., a web browser). However, the educational data mining (EDM) potential of these virtual microscopes remains under-utilized. In this paper, we use EDM techniques to investigate three research questions on a virtual microscope dataset of N = 1,460 students. First, which factors predict the success of students locating structures in a virtual microscope? We answer this question with a generalized item response theory model (with 77% test accuracy and 0.82 test AUC in 10-fold cross-validation) and find that task difficulty is the most predictive parameter, whereas student ability is less predictive, prior success on the same task and exposure to an explanatory slide are moderately predictive, and task duration as well as prior mistakes are not predictive. Second, what are typical locations of student mistakes? And third, what are possible misconceptions explaining these locations? A clustering analysis revealed that student mistakes for a difficult task are mostly located in plausible positions ('near misses') whereas mistakes in an easy task are more indicative of deeper misconceptions.