Artificial intelligence could soon help build education systems that improve every time they are used, according to Professor Tom Mitchell (Carnegie Mellon University), who delivered the 2025 Peter Kirstein Lecture, Where Can AI Take Education by 2030?
Drawing on examples from the US and China, he described how AI-powered platforms are already being used by millions of learners – and argued that the next five years could see a dramatic shift in how we teach and learn.
At the heart of his talk was the idea that AI can learn how to teach more effectively by observing how students learn. By analysing student performance and tailoring support in real time, Mitchell argued, AI could one day offer the kind of personalised help that only a human tutor currently provides – but at scale. “If AI is ever going to make a difference in education,” he said, “this is the decade it will happen.”
Professor Mitchell shared research showing how AI systems can analyse student performance, select the most helpful follow-up action, and even generate new content. His team’s work has focused on creating ‘teaching rubrics’ – human-readable frameworks that guide AI in generating educational hints, which can be reviewed and edited by teachers.
One key takeaway was that AI isn’t just a tool for automation – it can be a tool for learning how to teach better.
Panel reflections on AI in the classroom
A panel discussion followed the lecture, chaired by Professor Kaska Porayska-Pomsta, Director of the UCL Knowledge Lab at the Institute of Education. The panel featured experts in AI, education, philosophy and industry.
They welcomed the potential of AI to personalise learning and reduce teacher workload, but cautioned against letting the technology dictate what or how we teach.
Professor Ivana Drobnjak (UCL Computer Science) reflected on the broader purpose of education, suggesting that AI should support – not define – what we want learners to become. She stressed the importance of designing tools with teachers and students, so that technology supports meaningful goals rather than efficiency alone.
Matthew Pryor, founder of the education platform Graffinity, spoke about the difficulty of knowing which AI tools actually work in classrooms. He said many projects involve collaboration between schools, researchers and developers, but the evidence about what makes a real difference is still developing. As a result, schools often lack the guidance they need to make confident, informed decisions about which tools to use.
Professor Daniel Rothschild (UCL Philosophy) raised concerns about over-reliance on large language models. He warned that while these systems may sound convincing, they don’t truly understand the content they produce – and shouldn’t be trusted to replace human teachers. Instead, he called for clear and regularly updated guidance on how such tools should be used in education.
Teachers still take the lead
Ben Bartlett, CEO of the Hinchley Wood Learning Partnership, attended the event and later shared his reflections, welcoming the opportunity to explore how AI might shape the future of education – while emphasising the continued importance of teachers.
“Both the lecture and follow-up discussion offered interesting insights about where we might find ourselves with regard to the use of AI in schools and education more generally, in the near future,” he said.
“As a Trust, we are actively seeking to embrace AI and all that it has to offer… However, the role of the teacher in building relationships and offering a broad and balanced range of learning experiences is key to a successful school and must not be forgotten.”
A legacy of disruption
The Peter Kirstein Lecture series was created to honour Professor Peter Kirstein, the founding Head of UCL Computer Science and a major figure in the development of the internet in Europe. A visionary who challenged orthodoxy and championed boundary-pushing ideas, Peter believed in the power of debate to drive progress.
“The Kirstein Lecture is about bold ideas and open debate – just as Peter intended,” said Professor Daniel Alexander, Head of UCL Computer Science. “This year’s event captured that spirit, challenging us to think critically about where computer science is heading and how it’s reshaping society. By focusing on AI in education, we explored not just the technology, but the values and choices that will shape its impact.”
Professor Tom Mitchell said: "It was an honour and a delight to give the 2025 Peter Kirstein Lecture, to meet Peter's family, and to hear the related ideas of top UCL researchers in this field."
Each year, the topic is chosen in collaboration with newer members of staff, reflecting Peter’s belief in fresh perspectives. This year’s lecture was organised by Wei Chen, Tigi Bhatnagar, Ramona Behravan, Victoria Cooper-Cahill and Sam Stockdale.
Watch the lecture
If you were unable to attend, the full lecture and panel discussion are available to watch online.
In pictures: The Peter Kirstein Lecture on AI in Education
