NB. this is the draft introduction to the final chapter of my still-being-written book on AI and education (‘AI and education: the challenge of thinking differently’ Polity, due late 2026).

As we have seen throughout this book, there are many reasons to push back against the creep of AI into education. These are technologies that shrink the complexities of classrooms and schools down to what is quantifiable and can be captured into data. These are technologies that reduce the art of teaching into patterns of symbols that can be taught to machines. These are technologies with inherent computational frailties in the form of statistical errors and biases, technical glitches and data hallucinations. These are technologies that enforce a constant state of monitoring and surveillance. These are technologies that exacerbate the most discriminatory and unfair aspects of schooling and society. These are technologies with disproportionate environmental impacts, and that depend on the exploitation of ‘ghost’ labour in some of the world’s poorest regions. Oftentimes, these are technologies that simply do not do what they are supposed to. By any decent reckoning, these are technologies that anyone claiming to care about education should not be welcoming with open arms.

Yet perhaps the most disheartening aspect of the recent push for AI is how it devalues and debases the status of schools, teachers, students and everyone else involved in education. Any insistence that current forms of AI can improve education is an insistence that we accept a hollowed-out version of education that is routine, formulaic and utterly mindless. Seen in these terms, then, good teachers and students are those who always maximise their time, diligently respond to feedback and can be nudged to make correct decisions. Good education is simply a matter of repeatedly and efficiently doing the same things ‘at scale’. This is a robotic version of education that appeals to managerialists, consultants and tech-bros with little experience and/or interest in the unavoidable messiness of schools and classrooms. Much of the enthusiasm for ‘transforming education through AI’ is rooted in thinly-veiled contempt for the forms of education that we currently have. As Anthony Moser (2025) put it, “the makers of AI aren’t damned by their failures, they’re damned by their goals”.

Of course, efficiency, precision and standardisation are not wholly undesirable qualities in and of themselves. Nobody wants to see teachers who are completely disorganised, students who are blatantly wasting their time and repeatedly failing to take heed of their mistakes. Yet, the ‘AI-first’ approach to education is a dead-eyed, soulless version of what should be a lively, joyful and unpredictable endeavour. Pushing for AI in education is to give up on many of the things that ideally make education more than just a procedural matter of filling up empty vessels with knowledge – the joys of improvising, going off-script and thinking outside the box … the benefits of repeatedly struggling to get things right … the unexpected things that occur when learning together with others.  AI technologies are utterly tone deaf to all of these subtleties, nuances and frictions. The best forms of education are full of elements that simply do not compute (and therefore do not count) in the AI classroom. The underpinning premise of this book is that there is much more to education than can be automated; that the unpredictable and chaotic nature of education is something that we should value and defend.

Of course, this is not to say that there is no room at all for any form of AI ever in education. Please do not conflate the arguments in this book with reactionary knee-jerk responses that there is no place for any digital device or screen in classrooms and schools. This is not a book arguing that a traditional ‘back to basics’ approach will somehow correct all that is currently wrong with the world. In contrast, if AI technologies can be designed in ways that are genuinely focused on improving education experiences for all, or empowering students in respectful and non-harmful ways, then we should not dismiss them out of hand. The challenge that we face, however, is whether it is possible to build these other forms of educational AI in ways that are not harmful. Is it feasible to have forms of AI in our classrooms, schools and other education settings that do not result in all of the problems previously identified? In short, the prospect of AI in education is something that we need to think differently – not dismissively – about.