Copy the page URI to the clipboard
Ekuban, Audrey and Domingue, John
(2023).
DOI: https://doi.org/10.1145/3543873.3587644
Abstract
When students interact with an online course, the routes they take when navigating through the course can be captured. Learning Analytics is the process of measuring, collecting, recording, and analysing this Student Activity Data. Predictive Learning Analytics, a sub-field of Learning Analytics, can help to identify students who are at risk of dropping out or failing, as well as students who are close to a grade boundary. Course tutors can use the insights provided by the analyses to offer timely assistance to these students.
Despite its usefulness, there are privacy and ethical issues with the typically centralised approach to Predictive Learning Analytics. In this positioning paper, it is proposed that the issues associated with Predictive Learning Analytics can be alleviated, in a framework called EMPRESS, by combining 1) self-sovereign data, where data owners control who legitimately has access to data pertaining to them, 2) Federated Learning, where the data remains on the data owner’s device and/or the data is processed by the data owners themselves, and 3) Graph Convolutional Networks for Heterogeneous graphs, which are examples of knowledge graphs.