Generative Artificial Intelligence in Higher Education: A Decision-Support Framework for Practical Use, Governance, and Teaching Decision-Making

Abstract

Generative artificial intelligence has entered higher education faster than institutions have governed it, leaving lecturers to decide case by case what may responsibly be handed over. This paper reframes the lecturer not as a user of AI tools but as a decision-maker who delegates parts of the teaching workflow to an analytical assistant under explicit human oversight. Drawing on an integrative review of literature from 2019 to 2025 in Scopus, Web of Science, IEEE Xplore and ERIC, the paper synthesises the documented uses, benefits and risks of AI in university teaching. A consistent finding is that the benefits are conditional, arising from the human-machine pairing rather than the tool alone, while the most structural risks cluster around academic integrity, honest disclosure and equity. One question organises them: what may be delegated, and how far? Adapting the classical model of types and levels of automation from human-factors engineering, the paper maps four teaching decision functions, namely acquiring, analysing, deciding and implementing, against a graded scale of AI assistance. The resulting pedagogical decision-support framework couples each function with governance guardrails covering transparency, academic integrity, data protection, equity, and accountability, and proposes a decision-rights rule that keeps evaluative and relational judgements with the teacher while allowing higher assistance for preparatory and generative work. The contribution is a per-task rule: policy frameworks organise institutions and AI-literacy frameworks build competence, while this one tells a lecturer how far to delegate a given task. The paper closes with limitations and an agenda for validation