Group Counterfactual Explanations: A Use Case to Support Students at Risk of Dropping Out in Online Education

This paper proposes the novel application of group counterfactual explanations to the problem of predicting student at risk of drop-out. Our objective is to provide explanations for trying to re-cover the largest number of students with less effort and cost. Using group counterfactuals, in-structors and institutions could recover large groups of students with minimal remedial actions. For testing it, we have used the well-known public educational OULAD dataset that contains student’s clicks made throughout interaction with online courses. We have modified and adapted the only existed algorithm for generating group counterfactual named GROUP-CF. We also used the DICE individual counterfactual algorithm with the K-means clustering method and new op-tions for discovering the most representative counterfactuals for a group of students. The results obtained are very promising and they show that our approach can be successfully applied to re-cover 99.3% of students at risk of failing in a shorter time in comparison to traditional individual counterfactuals. And although, a group counterfactual proposes to change a greater number of student’s features, the values are lighter and therefore seem easier to apply than the ones obtained with individual counterfactuals. This work opens a new line of research in education.