Among the sessions at UNESCO Digital Learning Week 2026 was a fireside conversation between Khan Academy founder Sal Khan and UCL's Wayne Holmes — an AI-tutoring-industry founder in the same seat as an academic whose career has been spent asking hard questions about exactly that industry. Holmes's own research strand is worth knowing about on its own terms, because it's one of the more UK-relevant threads in this whole global conversation.

Who Wayne Holmes is

Wayne Holmes is based at UCL (University College London), one of the UK's most influential institutions in education research. His work sits in a specific and increasingly important corner of the AI-in-education conversation: human-centred AI design, and what's often called algorithmic accountability — the question of who is responsible when an AI system makes a decision about a student's learning, and whether that decision can be understood, questioned, or overturned.

That's a different emphasis from most of the AI-in-education commentary aimed at parents, which tends to focus on content ("is this appropriate for my child to see?") or outcomes ("does this improve grades?"). Holmes's strand asks a structural question underneath both: when an algorithm decides your child needs more practice on fractions, or flags a piece of writing as likely AI-generated, or chooses what hint to give — who is accountable for that decision, and can anyone actually inspect it?

Why "accountability" is the harder problem

It's relatively easy to audit an AI system for obviously harmful content. It's much harder to audit an AI system for a subtler failure: consistently under-challenging a child who could handle more, or over-flagging one child's writing style as "AI-generated" because of how they happen to write, or nudging every student toward the same learning path regardless of what actually suits them. These aren't safety failures in the dramatic sense — nobody gets hurt in an obvious way — but they're exactly the kind of decision an opaque algorithm can get quietly, persistently wrong, at scale, for years, with no parent or teacher ever knowing to ask the question.

This is why "can this be understood and challenged" matters as much as "is this safe" — it's the difference between a system that might occasionally get something wrong (true of any teacher, human or AI) and a system where nobody, including the people who built it, could tell you why it got something wrong or fix it once they knew.

Why this matters specifically for UK schools

UCL isn't a government body, and Holmes's research isn't policy in the way the DfE's own guidance documents are. But UCL's Institute of Education is one of the most influential voices feeding into UK education policy, which means work like this shapes the evidence base underneath decisions the DfE and individual schools eventually make. We look at how UNESCO's broader principles line up against the UK's own DfE guidance directly in how UNESCO's AI education principles compare to UK DfE guidance.

For a UK parent, the practical thread that connects all of this is simple, even if the research underneath it is not: can you actually get a real answer to "why did the AI do that?" A product that can explain its own reasoning — why it gave a hint instead of an answer, why it flagged a subject as needing more practice — is doing the accountability work Holmes's research strand is about. A product that can only say "the algorithm decided" is not, regardless of how safe its content filters are.

FAQ

Who is Wayne Holmes?

Wayne Holmes is an academic at UCL (University College London) whose research focuses on human-centred AI in education and algorithmic accountability in schools. He took part in a fireside conversation with Khan Academy founder Sal Khan at UNESCO Digital Learning Week 2026.

What does "algorithmic accountability" mean in a school context?

It means being able to answer, for any decision an AI system makes about a student — a hint it gives, a grade it suggests, a path it recommends — who is responsible if that decision is wrong, and how a parent, teacher or child could challenge or understand it. It's the opposite of a black-box system nobody can question.

Does UCL's AI-in-education research apply directly to UK schools?

UCL is a UK institution and its Institute of Education is one of the country's most influential voices on education policy, so its AI-ethics research feeds directly into UK debates — but it isn't government guidance itself. The DfE's own generative-AI guidance for schools is the more direct policy document; UCL's work is closer to the evidence base underneath it.

Duke Harewood built aitutors.me for his own KS3 daughter. "Can you explain why the AI did that" is a question we try to be able to answer honestly, every time — the hint ladder exists precisely so the reasoning is never a black box. Published 11 September 2026.