A research agenda for studying how AI couples educational functions in broad access higher education

Abstract Research on AI in higher education has expanded rapidly, but much of it remains organized around tools and adoption rather than around institutional reorganization. This article proposes a research agenda for studying what it calls institutional coupling : the degree to which AI-mediated systems draw together educational functions — authorship, tutoring, advising, assessment, analytics, and governance — that institutions have historically governed as separate domains. The article argues that adoption counts reveal little about how educational work is being reorganized, and that the more consequential research question concerns the redistribution of judgment, the coupling of institutional functions, the visibility of process, the design of contestability, and the uneven effects of optimization across student populations. Drawing on literature on AI in higher education, guided pathways, advising redesign, educational integrity, learning analytics ethics, and governance, the article specifies five research questions, a set of observable indicators grouped into five families, a comparative framework, candidate methods, and the principal threats to inference — including over-attribution, tool fetishism, policy–practice gaps, and sector flattening. It argues that community colleges and other broad-access institutions are especially valuable research sites because advising, scheduling, pathways, student support, and triage are already tightly coordinated there, which makes coupling easier to observe. The contribution is a design for inquiry rather than a finding: a way of specifying what would have to be observed for claims about AI-driven institutional convergence to be tested rather than asserted.