When I read this article by Yong Zhao and colleagues, I had to pause for breath. They recently published The Death and Rebirth of Research in Education in the Age of AI (2025). And honestly? I am still unsure whether I should be impressed or worried.
First, the numbers were quite nauseating for someone who thought he was keeping up with educational research. According to their calculations, around 140,000 academic articles on education are published annually. That figure alone is dizzying.
And yet the question remains: how many of these actually make a real difference in classrooms or in policy? Zhao and colleagues point out sharply that the impact is often minimal. Not because researchers are not doing their best, but because the whole system is clogged with noise: peer review that falters, an excessive fixation on numbers, the old “paradigm wars” between quantitative and qualitative, and the tendency to generalise results across contexts and learners who are, in reality, very different. At the same time, I do see a lot of good work being done, but it often takes considerable effort to separate the wheat from the chaff.
Where the article is strong is in its analysis, which suggests that AI does not solve these problems but actually makes them more pressing. The authors sketch how generative AI undermines research stability: a study on GPT-3 is already outdated by the time GPT-4 or GPT-5 appears. What is more, AI speeds up and broadens literature reviews, but at the same time risks eroding critical depth – something I wrote about earlier in this blogpost.
There are also new opportunities – more participatory research, more attention to complexity, perhaps even an “epistemological rebirth”. However, risks also creep in: AI spitting out patterns without meaning, researchers hiding behind efficiency, and a system producing even more publications without any guarantee of higher quality. Before you know it, we will have AI reviewing AI.
My reflections? First, the call for pluralism and imagination is justified, but hardly a new one. The real question is how to put this into practice in an academic ecosystem that still rewards publication volume. Second, the idea that AI causes a paradigm shift sounds appealing, but perhaps we should recognise that many of the old problems – bias, lack of impact, oversimplification – will not disappear through technology. They get a new jacket. One of the major issues with AI is that it is trained on human-made data and texts. Bias is not eliminated; it becomes amplified.
In brief, this article serves as an invitation to reinvent educational research. But anyone who thinks AI will be the saviour of science is mistaken. The rebirth Zhao and colleagues predict will require more than clever algorithms. It demands courageous choices, critical reflection, and above all, the will to publish less, but more meaningfully. And that plea, I fear, was already on the table long before ChatGPT, Gemini and other agreeable nodding machines came along.
P.S.: The image with this post was created using AI. Yes, I’m well aware of the irony.
Abstract of the article:
Purpose
This article critically examines the enduring problems and emerging possibilities of educational research in light of rapid advances in artificial intelligence (AI). It seeks to understand why educational research has struggled to influence practice and policy meaningfully and explores how AI necessitates a fundamental rethinking of research purposes, methods, and epistemologies.
Design/Approach/Methods
The article adopts a conceptual and critical review approach, drawing on historical, philosophical, and methodological literature. It identifies and analyzes seven major problems in traditional educational research, including flaws in peer review, quantification bias, methodological fragmentation, overgeneralization, neglect of individual learner diversity, limited educational imagination, and narrow outcome measures. It then explores how AI technologies challenge and reshape core assumptions about knowledge production and educational inquiry.
Findings
Traditional educational research is constrained by outdated paradigms that emphasize generalizability, stability, and typicality at the expense of contextual sensitivity, individual variability, and imaginative possibilities. The rapid evolution of AI further undermines assumptions of stable treatments, linear causality, and human-centered cognition. AI opens new opportunities for participatory, iterative, and systems-oriented research, while also raising ethical and epistemological concerns that demand critical reflection.
Originality/Value
This article offers a timely and provocative analysis of the limitations of traditional educational research and articulates a vision for its rebirth in the age of AI. It contributes to the growing discourse on paradigm shifts in education by integrating critiques of research orthodoxy with emerging insights into AI-enabled learning. The article calls for methodological pluralism, ethical vigilance, and epistemological innovation, positioning researchers to better respond to the complex and evolving landscape of education in a post-AI world.