Securing Equitable Success in the First Year: A Whole-Institution Case Study of Data-Driven Early Warning and Critical AI Literacy

Widening access has transformed who enters higher education but not who thrives once inside: inequality resurfaces as first-year failure and dropout, while the rapid spread of generative artificial intelligence (AI) threatens to open new divides. This article presents a whole-institution case study of how one Portuguese public university, Iscte—University Institute of Lisbon, has pursued equitable success in the first year through a single strategy with two reinforcing arms: a reactive arm that detects at-risk students and responds with personalized support, and a proactive arm that builds a baseline of critical AI literacy that is mandatory on one campus and optional on the other. Drawing on institutional records from 2016/17 to 2024/25, it combines a rule-based early warning alarm system, a comparison of machine learning models, and descriptive evidence on the literacy initiatives. The findings show that academic risk is concentrated in identifiable groups, that flagged students who engaged with support re-enrolled at higher rates than those who did not (though not distinguishably so once students who had already formalized a withdrawal are set aside, a group that took up no support and left without exception), and that a common literacy baseline is operationally achievable. The findings also expose the limits of each, as support reaches only a minority of those flagged and exploratory predictive models transfer poorly to the following cohort. Together, these observational results frame reactive support and proactive capacity-building as facets of one inclusion strategy.