Bounded AI Assistance in Higher Education: Re-Aligning Pedagogy, Assessment and Institutional Governance
The rise of large language models (LLMs) has led universities and other higher education institutions to rethink the ways in which knowledge is conveyed, learning is facilitated, and academic performance is evaluated. While such technologies present possibilities for tailored explanations, feedback that supports learning as it unfolds, and joint inquiry, they also give rise to concerns regarding inaccurate information, reliance on machines for thinking, academic integrity, and inequities in access. Drawing on a narrative synthesis of selected literature and institutional guidance, this conceptual paper explores the pedagogical ramifications of LLMs. The paper contends that integrating them effectively demands learning and research environments that bring together subject-specific knowledge, critical questioning, openness about bounded AI use, and demonstrable competence. Constructive alignment, self-regulated learning, cognitive load theory, and cognitive engagement supply the conceptual bases for linking intended learning outcomes, teaching and learning activities, and assessment. A cyclical process is put forward, comprising independent engagement, AI-assisted interaction under guidance, verification, application, and reflection. The paper goes on to elaborate dimensions of institutional policy, encompassing governance, curriculum, assessment, research integrity, privacy, intellectual property, equity, procurement, and quality assurance. Specific consideration is devoted to contexts with constrained ICT resources. The paper's conclusion is that the educational value of LLMs rests on purposeful design of specific institutional AI policy that is clear, transparent and progressive.
Authors
- Marien Matafwali
- Richard Mulenga (ORCID: https://orcid.org/0009-0003-0065-7432)
Publication Details
- Journal
- Journal of Innovative Research
- Published
- 2026-11-21
- DOI
- https://doi.org/10.54536/jir.v4i3.9008
- Primary Topic
- Artificial Intelligence in Education
- Type
- article
- Field-Weighted Citation Impact
- 0.00