Student success prediction is one of the many applications of artificial intelligence (AI) which helps educators identify the students requiring tailored support. The intelligent algorithms used for this task consider various factors to make accurate decisions. However, the decisions produced by these models often become ineffective due to lack of explainability and trust. To fill this gap, this paper employs several machine learning models on a real-world dataset to predict students’ learning outcomes from their social media usage. By leveraging the SHapley Additive exPlanations (SHAP) to investigate the model outcomes, we conduct a critical analysis of the model outcomes. We found several sensitive features were considered important by these models which can lead to questions of trust and fairness regarding the use of such features. Our findings were further evaluated by a real-world user study.
CITATION STYLE
Afrin, F., Hamilton, M., & Thevathyan, C. (2022). On the Explanation of AI-Based Student Success Prediction. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 13351 LNCS, pp. 252–258). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-3-031-08754-7_34
Mendeley helps you to discover research relevant for your work.