Towards an Explainable Learning Analytics: Advancing Transparency, Bias-free and Trust

Enfoque

With the widespread availability of online education, there has been a surge in the amount of data that is available and accessible. This abundance of educational data has provided ample opportunities for the field of learning analytics and educational data mining to expand and yield numerous advantages. Machine learning and deep learning techniques are commonly employed in these fields to analyse the data and provide actionable insights. However, it is not enough to apply such techniques without enabling humans to understand it. In this paper, we will refer to enabling the process of understanding machine learning techniques in the context of learning analytics as “Explainable Learning Analytics: XLA”. With particular focus on a case study on explainable machine learning techniques for predicting student performance in online courses, the contributions of this paper are twofold. First, we demonstrate how explainable machine learning tailored by the needs of stakeholders can make a big difference in learning analytics by effectively improving the transparency, trust, and model bias understanding and fairness of complex models. Second, this paper will enhance our understanding of machine learning techniques in learning analytics scenarios and provide practitioners with a better understanding of complex techniques. In addition, this paper will also facilitate further applications of machine learning techniques in learning analytics.