AI in education is often presented as a break with everything that came before. Williamson and Eynon’s (2020) editorial, Historical threads, missing links, and future directions in AI in education, offers a useful challenge to that framing.
They argue that educational AI emerged through connections among academic research, commercial edtech, corporate infrastructure, and data-driven governance. Understanding those histories helps us examine whose priorities shape educational change.
Written before ChatGPT’s public release, this paper deserves attention for the questions it asks about power, evidence, and educational purpose. I find those questions especially useful when evaluating promises about what schools should become.
What Historical Argument Do the Authors Make?
The article introduces a special issue on critical perspectives on AI in education. It draws on historical literature, policy and industry materials, and the issue’s contributions to trace selected developments.
The authors describe their approach as genealogical: they examine how particular relationships, interests, and circumstances helped produce educational AI. They explicitly acknowledge that the editorial cannot provide a full history.
They identify several intertwined threads. Academic researchers developed intelligent tutoring systems and methods for studying learning. Edtech businesses pursued markets for personalised learning.
Technology corporations supplied infrastructure on which educational services could depend. Policy and governance increasingly organised education through data, measurement, and prediction. These activities intersected, but the authors caution against collapsing academic research and commercial activity into a single enterprise.
This is an interpretive editorial, with no newly reported classroom intervention or participant study. Its examples reflect the period in which it was written. It provides a historical and critical lens for examining AI in education; it does not establish the effectiveness or harms of every application.

Why Should Teachers Care About Forgotten Histories?
What interests me is how easily forgetting the past can narrow the choices available in the present. If a technology is described as an inevitable revolution, educators may feel that their main task is to adjust. Historical examination opens up questions about who promoted a particular direction and what alternatives were available.
The authors show that ambitious promises about transforming teaching and learning accompanied earlier AI systems. That does not make every new claim false. It gives us a reason to examine the educational problem being defined, the evidence offered, and the interests behind the proposed solution.
For teachers and school leaders, my practical reading is to ask what a product assumes about learning. Which activities count as progress? What can its measurements capture, and what falls outside them?
These questions matter because adopting a system can also mean accepting its categories and priorities. A claim about efficiency leaves open whether the activity being accelerated serves the educational purposes we value.
What Happens Behind the Classroom Interface?
A particularly useful part of the editorial examines infrastructure. AI can enter education through the underlying services that connect applications, process data, and support institutional decisions. Its influence may be difficult to see from the interface teachers and students use.
The authors explain how these arrangements can create dependence on corporate systems and the forms of analysis they enable. Their concern extends beyond the merits of an individual classroom tool to the relationships surrounding it.
I think this supports a broader approach to evaluating educational technology. Alongside usability and teaching value, institutions need to understand the services involved, how data moves, and who controls decisions about its use. The editorial supplies grounds for asking those questions; it does not provide an audit of present-day products.
This connects with my earlier discussion of critical AI literacy, where power, representation, and the interests surrounding AI are central. Historical analysis adds another question: how did those arrangements become established?
Whose Experience Should Shape Educational AI?
Williamson and Eynon identify an important gap in the critical research available at the time: limited knowledge of how teachers and learners actually encountered AI in everyday educational settings. They call for ethnographic case studies that examine what happens in practice.
That qualification matters. A persuasive analysis of a system’s risks cannot, by itself, tell us how those risks unfold in a particular classroom.
The authors also question meaningful consent when students and teachers have little power to refuse institutional technologies. They propose participatory approaches that involve affected communities, particularly marginalised groups, in shaping development and use.
For educators working on AI policy and ethics, I read this as a reason to involve teachers and students while decisions remain open. Their experiences should inform what gets built and adopted.
The paper’s strongest contribution is to make educational futures available for debate: technology develops through choices, and those choices deserve scrutiny in relation to education’s purposes and values.
References
Williamson, B., & Eynon, R. (2020). Historical threads, missing links, and future directions in AI in education. Learning, Media and Technology, 45(3), 223–235. https://doi.org/10.1080/17439884.2020.1798995




