While the current literature on algorithmic fairness has rapidly expanded over the past years, it has yet to fully arrive in educational contexts, namely, learning analytics. In the present paper, we examine possible forms of discrimination, as well as ways to measure and establish fairness in virtual learning environments. The prediction of students’ course outcome is conducted on a VLE dataset and analyzed with respect to fairness. Two measures are recommended for the prior investigation of learning data, to ensure their balance and fitness for further data analysis.
Keyword(s): Learning Analytics; At-risk Prediction; Moocs; Fairness.