Dr Paul Mulholland

Our project is dedicated to advancing socio-technical approaches to fairness in Higher Education within the UK, with a specific focus on addressing bias and discrimination in AI-driven learning analytics systems. As technology increasingly shapes the student experience and educational outcomes, institutions worldwide are recalibrating their data strategies for positive impacts on student success. However, the integration of fairness, both technically and socially, remains an understudied aspect of these strategies. Leveraging the successful OUAnalyse system at the Open University, our project seeks to establish a framework of best practices, risks, and compliance with UK regulations. This initiative aims to guide Higher Education institutions in the UK as they develop their digital strategies, emphasising the core principles of fairness and responsible AI. The OU Analyse Dashboard (OUA) serves as our primary use case, a Predictive Learning Analytics System supporting over 200,000 students. Through weekly alerts to tutors, the system identifies students at risk of dropping out, contributing to a noteworthy increase in retention rates, especially for ethnic minority and socioeconomically disadvantaged students. Our project is organised around five fundamental pillars: Fairness, Explainability, Transparency, Accountability, and Social Justice. In the pursuit of fairness, we will conduct a comprehensive examination of PLA's performance across diverse student cohorts, paying special attention to intersectionality. Scrutinising prediction errors, and whether they affect differently to different student groups based on sensitive characteristics, such as gender, ethnicity, disability, Index of Multiple Deprivation, and previous qualifications, we aim to uncover biases and enhance overall system fairness. Transparency is addressed by providing students with predictions and explanations, piloting a student dashboard to make predictions visible and empowering students to take proactive measures in collaboration with tutors. In terms of accountability, we will investigate tutor reactions to predictions, ensuring equitable support for all students. The project will contribute to the broader discourse on ethical AI implementation in education, aligning with the UK government's innovation-focused approach to AI regulation. Collaboration with regulatory bodies, such as The Equality & Human Rights Commission and The Information Commissioner's Office, will ensure compliance of the provided framework with equality and human rights law, data protection, and sector-specific legislation. Beyond technical considerations, our commitment to Equality, Diversity, and Inclusion (EDI) is central. We actively address societal inequalities, striving to mitigate biases in PLA systems and reducing awarding gaps, ensuring fair and equal opportunities for all students. Our dissemination efforts encompass scientific conferences, collaboration with regulatory bodies, and engagement with higher education communities, fostering knowledge sharing and responsible AI practices. The project's potential economic impact lies in improving student investment returns, contributing to a socially inclusive educational landscape with diversified job opportunities. Success metrics include compliance with government priorities, regional impacts on other universities, and economic and social impacts on student quality of life and inclusion. Through this comprehensive approach, our project aims to reshape the AI in the education landscape, fostering fairness, transparency, and inclusivity.