AI-Enhanced Multimedia for Personalised Higher Education: A Critical Review
TL;DR
The review examines four interrelated design challenges: the pedagogical quality of AI-generated multimedia; the validity and fairness of automatic feedback; the interpretability of learning analytics; and the impact on established instructional design processes.
Abstract
In higher education, multimedia authoring, feedback, analytics and adaptive learning systems are increasingly embedded with artificial intelligence (AI). The impact of these changes on instructional design for personalised learning is examined. Instead of viewing personalisation as automatic optimisation, the review examines four interrelated design challenges: the pedagogical quality of AI-generated multimedia; the validity and fairness of automatic feedback; the interpretability of learning analytics; and the impact on established instructional design processes. Research suggests that AI can broaden the diversity and responsiveness of resources for learning, but evidence of lasting learning gains is more patchy and very much dependent on human oversight, curricula and student agency. The systems may then use non-transparent models or data which are unrepresentative, thus reproducing equity concerns. The review suggests a series of principles for responsible design: pedagogical purpose; cognitive coherence; human verification; contestable personalisation; and proportionate data use. AI should be viewed as a limited design resource instead of an independent system of instruction.