Towards Explainable Educational Recommendation through Path Reasoning Methods

Current recommender systems in education lack explainability and interpretability, making it challenging for stakeholders to understand how the recommended content relates to them. Path reasoning methods are an emerging class of recommender systems that provides users with the reasoning behind a recommendation. While these methods have been shown to work well in several domains, there is no extensive research on their effectiveness in the context of education. In this ongoing project, we investigate the extent to which the existing path reasoning methods meet utility and beyond utility objectives in educational data. Experiments on two large-scale online course datasets show that this class of methods yields promising results and poses the ground for future advances.