Leveraging Learning Analytics to Support Institutional Efforts in Addressing Awarding Gaps

Awarding gaps, the persistent disparities in academic outcomes across student groups, remain a major challenge in education. These gaps, often observed along dimensions such as ethnicity, socio-economic status, disability, and gender, reflect broader structural inequalities and can limit the life opportunities of marginalised learners. Addressing them is a priority for institutions aiming to deliver fair and inclusive education. This thesis investigates how Learning Analytics (LA) can contribute to addressing awarding gaps. While LA is increasingly used in institutional decision-making, its equity impacts remain underexplored. Using The Open University as a case study, the research addresses three questions: (1) To what extent do existing LA systems mitigate awarding gaps? (2) How fairly do predictive LA models perform across student demographics? (3) How can equity-focused LA tools be co-designed with institutional stakeholders? A mixed-methods approach is employed across three empirical studies. The first evaluates the existing LA system, OU Analyse, and finds that, when paired with active teaching practices, it can improve outcomes for students from ethnic minority and lower socio-economic backgrounds. The second examines fairness in LA predictive models, revealing disparities in false positive and false negative rates across ethnicity, gender, and disability. It also evaluates bias mitigation strategies, showing that subgroup-specific models do not consistently improve fairness. The third study focuses on the co-design of a novel Equality, Diversity and Inclusion LA dashboard, developed with institutional stakeholders to address awarding gaps. The dashboard provides intersectional, longitudinal views of awarding gaps and demonstrates how user-centred design can support equity-driven analytics. It also raises ethical questions, highlighting the need for fairness, explainability, and privacy in LA systems. Together, these studies offer empirical evidence, methodological guidance, and design principles for developing more equitable LA systems. The thesis concludes with directions for future work, including broader demographic modelling, responsible AI integration, and human-in-the-loop evaluations.