Systematic review of research on artificial intelligence applications in higher education: where are the educators?

This systematic review is one of the most cited entry points for understanding AI in higher education because it asks a deceptively simple question: if artificial intelligence is becoming important to teaching and learning, where are the educators in the research conversation? The authors searched peer-reviewed literature from 2007 to 2018 and screened 2656 initial records down to 146 studies for synthesis. Their review shows that AI in higher education had already become a substantial research area before the recent generative AI wave, but that the area was shaped heavily by computer science, STEM authorship, quantitative methods, and technical performance concerns.

The paper organizes AI applications into four major areas: profiling and prediction, assessment and evaluation, adaptive systems and personalization, and intelligent tutoring systems. This taxonomy is useful for AIEDHK because it maps research outputs to concrete product and institutional functions. Profiling and prediction include systems that estimate dropout risk, performance likelihood, or learning needs. Assessment and evaluation include automated grading, feedback, and evaluation support. Adaptive systems and personalization include tools that change learning paths or resources based on student data. Intelligent tutoring systems simulate forms of one-to-one guidance, often using learner models and domain models to select tasks or feedback.

The most important contribution, however, is not just the taxonomy. The review argues that the AIED literature in higher education often lacks strong pedagogical framing and critical attention to risk. The authors observe limited engagement with educational theory, limited educator authorship, and insufficient discussion of ethics, privacy, transparency, and institutional consequences. This is still highly relevant because many readers are tempted to evaluate AI education tools by technical novelty alone. Zawacki-Richter and colleagues instead make the case that research-to-practice translation needs educators, learning theory, and ethical analysis at the center.

An editorial reading should also notice the paper's methodological discipline. The authors do not simply collect examples of AI tools; they define inclusion criteria, code application areas, and examine authorship patterns. That makes the review useful as a model for future Research News work: every paper summary should distinguish what the technology does, what evidence supports it, and what educational assumptions are visible or missing. The paper also helps identify enduring gaps. If educator participation, ethical reasoning, and student voice are thin in a literature base, then a product team should treat that as a design warning rather than a side note. AIEDHK can use this paper to keep the research feed from becoming a technology catalogue.

For AIEDHK, this paper can anchor a why-review-matters narrative. It supports the idea that a knowledge hub should not simply report AI tools or model capabilities. It should ask who the system is for, what educational purpose it serves, what assumptions it makes about teaching and learning, and what risks it introduces. The paper is especially useful for higher education partnerships, university teaching innovation, and policy-oriented research briefings.