Critical educational technology asks us to examine how digital tools become part of education: who shapes their use, whose interests they serve, and what happens in everyday classrooms.
Neil Selwyn’s 2010 essay, Looking beyond learning: notes towards the critical study of educational technology, develops this argument. Reading it amid current discussions of generative AI, I find its questions particularly useful.
A tool might help a student complete an assignment while also introducing new demands on teachers, changing expectations about assessment, or making access more dependent on institutional resources. Evaluating its learning potential gives us one part of that picture. Selwyn asks researchers to investigate the surrounding arrangements as well.
Why critical educational technology matters
Educational technology discussions often begin with possibilities. We ask what a tool could improve, how teachers might integrate it, and which obstacles prevent wider adoption. These questions can quietly assume that increased use is the desirable outcome.
Selwyn challenges that assumption. Technology use develops through decisions involving teachers, students, institutions, businesses, and governments. A tool’s educational role depends on those relationships and the circumstances in which people encounter it.
This also changes how we understand non-use. A teacher who declines a platform may have concerns about its suitability, the work it requires, or the conditions attached to it. We need to investigate those reasons before interpreting the decision as a lack of competence or enthusiasm. The same applies when students use a tool differently from its designers’ expectations.
What Selwyn’s essay argues
Published in the Journal of Computer Assisted Learning, the paper is a conceptual discussion drawing on social theory and educational technology scholarship. Selwyn argues for accounts of technology as socially shaped and negotiated.
He proposes three broad questions: what technology use actually looks like, why it takes that form, and what consequences follow. His emphasis falls on ordinary, sometimes constrained practices that optimistic accounts can overlook.
The analysis extends across three levels. Individual teachers and learners matter, but their experiences sit within institutional routines and wider political, economic, and cultural relationships.
A classroom account becomes more informative when it considers the policies, resources, expectations, and commercial arrangements surrounding it. Selwyn also calls for greater attention to democracy, social justice, and empowerment.
The essay does not report an intervention or demonstrate that a particular technology improves or harms learning. It predates today’s generative AI systems. Its contribution is a way of framing research questions, alongside research into learning and effective technology design. Any application to contemporary AI requires further evidence.
Using these questions in AI teaching and research
One implication I take from the essay is that we should examine what an AI tool becomes in a particular setting. Consider an AI feedback system. A demonstration may show clear comments produced quickly.
In a classroom, we would also want to know whether students understand those comments, how teachers check them, and who resolves disagreements about their accuracy.
These are suggested applications of Selwyn’s argument. His paper did not test AI feedback systems. They help identify what we might observe before deciding whether a tool supports a workable educational practice.
For teachers, a manageable starting point would be to examine one assignment. What does the tool contribute? What must students and teachers do before and after using it? Does it create additional checking work? Can students question a response or choose an alternative route? These questions make the evaluation more specific than a general judgment about whether AI is useful.
I would also include students who use a tool infrequently or avoid it. Their accounts may reveal circumstances missing from an evaluation based only on enthusiastic users. For example, access arrangements, unclear expectations, or uncertainty about how to disclose assistance could warrant investigation. We should establish which explanations apply through evidence.
Researchers can follow the same approach by connecting classroom observations with institutional decisions. Who selected the platform? What expectations accompanied its introduction? What training and support were available? How do assessment policies shape the ways students use it? Interviews, observations, and relevant documents could help develop that account.
Selwyn’s argument also asks us to examine whose priorities guide an evaluation. An institution might emphasize adoption, a provider might emphasize efficiency, and a teacher might emphasize whether students can explain their reasoning. Investigating those priorities can reveal why people disagree even when they are discussing the same tool.
This connects with my discussion of generative AI in education and neutrality: decisions about educational technology involve assumptions about what education should achieve. My wider AI teaching practice discussions consider how those decisions enter everyday teaching.
A critical approach should also support practical change. If an evaluation identifies unequal access, excessive checking demands, or limited opportunities for students to challenge a system, the next question concerns what can be changed. That might involve revising an assignment, improving support, adjusting institutional expectations, or reconsidering the tool’s role.
Selwyn’s essay gives me a reason to slow down before equating adoption with improvement. A useful evaluation asks how people experience a technology, how those experiences are shaped, and whether the resulting arrangements serve the educational purposes we value.
Reference
Selwyn, N. (2010). Looking beyond learning: Notes towards the critical study of educational technology. Journal of Computer Assisted Learning, 26(1), 65–73. https://doi.org/10.1111/j.1365-2729.2009.00338.x
