What does it mean to widen participation? a machine learning analysis of historical and contemporary access discourses

Diversifying the demographic makeup of higher education institutions remains at the forefront of both government and university policy agendas. This paper examines the evolving discursive landscape of Widening Participation (WP) in the English higher education system through a machine learning analysis of institutional texts. Drawing on a novel dataset produced for this research entitled the Access Corpus, which is comprised of twenty years of Access Agreements and Access and Participation Plans, this paper applies machine learning tools to explore discursive shifts that have taken place in the domain of WP. Topic Modelling shows that both conceptual understandings of key terms and the focus of widening participation discourses have been transformed over the past twenty years as discourses have moved towards marketisation, and regulatory compliance. Importantly these shifts are not felt equally across the sector, statistical analysis indicates that high-status institutions display greater autonomy in their responses to regulatory policies and thus a more stable discourse over time. This paper’s empirical findings are situated within a broader theoretical understanding of discourse which emphasises its inherently untethered nature, where meanings are constituted and reconstituted through use over time; aligning with Ball’s ‘policy as discourse’ framework. Thus, ‘widening participation’ is neither fixed nor stable, but instead continually re-understood through institutional responses to shifting political, economic, and regulatory circumstances. This paper provides both methodological innovations through its use of topic modelling and a critical examination of how widening participation is constructed by higher education institutions over time.