My encounter with research on AI in education has been mostly through teaching an MA course in Education and Technology. Over the last 4 years or so, a recurring pattern in classroom conversations between students is the treatment of AI as equivalent to ‘personalised learning’, with the latter characterised in terms of solving three main problems in education. First, the ‘one size fits all’ model, which monopolistically imposes State-defined curricula, standards and timetables on a diverse population, privileging the interests of large powerful institutions like schools and universities over the specific and varied needs of individuals. Second, personalised learning enables people to ‘learn at their own pace’, by contrast to the fixed temporal divisions and speeds of the academic calendar and assessment frameworks. And third, teachers’ time is saved and their workload lightened, with the burdensome tasks of whole class teaching lifted off their shoulders.

These claims characterise classroom conversations because they can be found, in varied forms, in both research literature and in the descriptions of products and solutions advertised to, and within, education institutions, including my own university. And if this is what ‘personalised learning’ is about, how could anyone ever be against it? Furthermore, if this is what AI makes possible, then surely AI in Education brings only good tidings to all diverse learners….

Nonetheless, the celebration of personalisation begs questions which, historically, have been central to education as a professional practice and a field of study. And my argument in this short piece is that these are questions for which AI research in education ought to have nuanced, thoughtful and empirically-grounded responses. These responses seem overlooked or are largely unarticulated, because AI, like much technology before it in education, has been treated as a technique or a method: a means of doing the same thing in an incrementally enhanced way, more efficiently or more effectively, a move which has the effect of de-politicising education and framing it as a location (AI in Education) where stuff is explained, and expertise is more or less well transferred from the more to the less knowledgeable. The questions can be phrased as followed:

(1) What is the purpose of education? Education has emerged as a field of research, with associated philosophies, because it has historically been seen as a matter of public and collective concern, which ought to be subject to continuous democratic contestation and deliberation, precisely because it deals with what is or should be valued in public life. How do arguments for ‘personalised learning’ respond to this question? And more generally, what kinds of resources does AI research offer to develop understanding of the scope of this question and the different ways it has been and could be answered?

(2) What vision of a better world does education sustain? So, more specifically here, what kind of vision of a better society, a better economy, a better way of living together, does ‘personalised learning’ imply or promote? What features and qualities does the better world have to which ‘personalised learning’ is contributing?

(3) How does education practise equality, if we understand equality as a political claim, which sustains education’s commitment to self-determination, and marks its difference from indoctrination and mindless control? More specifically here, what kind of person is produced by ‘personalised learning’? How does ‘personalised learning’ convince this person of their own political and intellectual power, one which is equal to anyone else’s, insofar as what anyone is capable of thinking and doing, of becoming, is emergent rather than determinate? Education has always wrestled with the problem of how to treat people equally, not in the sense of the same, but equal in their capacity to become someone different from who they are as a person, in a way which isn’t pre-determined in advance, for instance by the needs of the existing economy and social order. What new resources, material and intellectual, does ‘personalised learning’ bring to this problem?

Other questions could be formulated but their gist gets at what education is about and for, a question which is inherently political as well as philosophical, and which, very broadly speaking, is not attended to very fully in educational technology research. One could object that such questions are not within researchers’ and teachers’ remit, since they have a job to be getting on with whose purpose has been defined by others higher up, including in senior management or government. But teaching or researching invokes a sense of purpose. And if there is a case to be made to personalise learning, doesn’t it touch precisely on what makes it meaningful, by contrast to determined only by more powerful others?

In raising these questions, I don’t mean to suggest that everyone should have clear and categorical answers. But the questions might have use in troubling certain assumptions and opening up possibilities for considering what we want in and out of education. I’ll return to the three main claims made about the benefits of personalised learning to illustrate this.

In defence of ‘one size fits all’: the figuring of a ‘one size fits all’ model of education relies on two analogies, the first that education institutions are like industrial era factories, and the second that they are like State monopolies. However, faith in the virtues of the comprehensive school, in the UK at least, was not the result of industrial era values, but rather a commitment to democracy, and the creation of a progressive civil society in which distinctions inherited at birth, not least of wealth, should not determine people’s future. ‘One curriculum fits all’ could perhaps be a way of interpreting a call to use education to keep students’ social roles open, rather than matching them to a fixed calling (as in, academic education for the managing or ruling elites, vocational training for the managed and the poor). Pursuing an education is, or arguably should be, a way of discovering that I am like everyone else: capable of thinking and doing what anyone has thought and done. Furthermore, this starting presumption collectively shares moral responsibility for creating, rather than registering, individual choices. Which in practice means that both teachers and students adapt to each other’s interests and circumstances: I have yet to meet a working teacher or lecturer, or go to a school or university, which practises ‘one size fits all’ in the classroom.

This is not to deny the strong correlation between socio-economic background and educational outcomes, but rather to suggest that ‘one size fits all’ is neither an empirical reality nor a professional ideal. So where does it exist, apart from in the discourse of ‘personalised learning’? And if it is this discourse which has created it, to what ends? Researchers in the critical tradition of Ed Tech research (e.g. MacGilchrist, 2021, Williamson, 2018, Selwyn, 2021, Ball and Grimaldi, 2022, Watters, 2021, to name but a few) have argued that ‘personalised learning’ is justification for a neoliberal agenda set on taking education’s financial resources out of the hands of the State but only to put them into those of monopolistic IT corporations – by contrast to a return to the ‘learner-centred’ progressivism of the 1960s. I won’t repeat these arguments here. But if this is what ‘personalised learning’ is doing, at least in part, then what are the arguments for it within education, when this is understood, as it has been historically, as a practice concerned with equality and a democratic civil society? How can it be reconciled, in theory and in practice, with some of the ideals of education as a discipline and an institution? What vision does it have of the relationship between knowledge and the social order? What is the collective social, moral and political problem it is tackling and how does it do so? How is its approach better than a comprehensive, or ‘one size fits all’, model of education? This is what I would like to see research on ‘personalised learning’ address more fully, to appreciate less cynically and suspiciously what it is actually about.

Against ‘learning at your own pace’: the idea that there is a pace to learning presupposes it has a beginning and an end point. It moves in a certain direction, at a certain speed, as judged from a point of view external to its emergence. The pace of learning can be detected only in relation to the creation of a gap which it then crosses. In other words, the phrase elicits a particular view of knowledge – that it has an order to it, such as from less to more complex – and of education – that its purpose is to close the gap between two knowable and determined points, say between the student’s and the teacher’s knowledge, or their speech on a topic. Learning is then framed as a student becoming increasingly like the teacher, reducing the deficit of the former in relation to the latter. But of course, a pace cannot be maintained unless that deficit is continuously rediscovered, a move which some might understand in terms of progressing up the achievement ladder, and others as precisely what makes the pedagogic relationship so intellectually stultifying (e.g. Rancière, 1987, Freire 1970/2017). More pragmatically, I would argue that the concept of ‘learning at your own pace’ is a contradiction in terms, and also simply a polite way of claiming that some people are slow so that others can be quick, and therefore a revification of the old argument for education as legitimation for meritocracy (Young, 1994).

In teaching practice, there are arguments for streaming and setting. Not necessarily very strong arguments on the basis of a commitment to equity and equality, but arguments nonetheless. These however aren’t prominent in the literature on ‘personalised learning’. More commonly, ‘learning at your own pace’ is justified simply using the seductive appeal of the discourse of uniqueness. But if researchers on ‘personalised learning’ are concerned with valuing individuality, they might explore how a knowledge of one’s own, a voice of one’s own, might be recognised in education, and how students might speak as individuals, rather than only as objects to be measured against a limited number of set scales. Rather than pushing everyone along, at various speeds, along the same path, perhaps attention could shift to multiplying paths and minimising their ordering by those who consider themselves to have already arrived. This is one of educational sociology’s constructive agendas (Rancière, 1987, Bourdieu and Passeron, 1990), and it is perhaps one which research in ‘personalised learning’ could take forward.

In defence of increasing teachers’ workload, and not saving their time: This might seem a hard one to argue…Perhaps the first way to do so is to point out that ‘technology saves time and reduces teachers’ workload’ is a claim which goes back to the very beginning of the history of ‘personalised learning’ (according to Watters, 2021), and it is a claim which remains always on the point of being realised. There is little evidence that ‘personalised learning’ technologies have ever done anything of the kind. So what faith can anyone still have in it?

A second point is that imagining education in terms of ‘personalised learning’ leaves little room for the teacher to play much of a role. It doesn’t so much save teachers’ time as circumvent its necessity, relocating responsibility for education to the individual learner who goes at their own pace. The obviousness with which the benefits of ‘personalised learning’ are treated in much research literature on the topic suggests that the main obstacle to its realisation is its implementation, a move which conveniently relocates the problem from its conceptual justification to the backwardness of working digital immigrants. Teachers are once again blamed for the failure of technology to fully realise its supposed potential - a potential which continues to exist precisely in the abyss of empirical substantiation generated ‘in the wild’, or in actual classrooms and sites of education (Hutchins, 1996; Selwyn, 2021).

Perhaps rather than trying to save teachers’ time, research on ‘personalised learning’ could aim for something more concrete and credible, which is how to foster more personal attachment to teaching as a profession, how to makes its work more meaningful, including by sustaining more professional autonomy as well as better relations with the other people it works with, not least students.

Final Thoughts

The brief for this opinion piece was to explore possible new avenues of research on AI in education. I haven’t researched AI, and I wasn’t sure I had a good basis on which to express an opinion. But AI is reconfiguring what we understand by technology and its significance for education, which is something I do research and teach about. I have focused here on ‘personalised learning’ because it is a term which has been raised repeatedly in the Education MA classes I teach. And what I have argued revisits terrain I have explored with students. What does it actually mean for and in education? What is the problem to which it is a solution? How does the problem arise; when, where and for whom, and what is the evidence that it is a credible or desirable solution and for whom? What other solutions or concepts does it endeavour to colonise or marginalise? And how might it be used to open up new ways of tackling perennial questions in education? To begin to think about this, research on ‘personalised learning’ and AI in education would benefit from treating education less as a location in which incapacity is perpetually re-discovered, and more as a political practice in which definitions of the person are perpetually under negotiation.