PISA 2025 has landed, and the recent “Beyond the results: What PISA 2025 means for education in Australia” turned into a discussion about where technology belongs in classrooms. Its finding that where curiosity declined, performance declined moves the question onto design. What does a tool invite a student to do, and what does it tell a teacher they can act on tomorrow to keep curiosity alive?
A hammer can build a house. Yet, that same hammer can demolish a whole house. Nobody argues about the problematic nature of hammers, because everybody accepts that the outcome sits with the person holding the tool, and what they set out to do.
Recently, the OECD published PISA 2025: more than 760,000 students across 91 countries, science as the major domain, and for the first time a measure of how students learn in a digital world, alongside the first large-scale international evidence on how they are using tools, like AI, for schoolwork.
The coverage has started, and in the coming weeks there will be a lot of discussion about which tools, in which countries, lead to the highest performance. While many things change over the years, the evidence and the throughline of the narrative stay constant: at the heart of all good performance are teachers who harness the best tools to spark curiosity and sustain student interest along the learning journey.
What PISA can and cannot settle
Three things from the launch have stayed with me.
The first is about curiosity. Where curiosity declined, performance declined. That sentence is the ‘ah-ha’ moment every teacher is looking for: motivation is built out of success a student can take credit for, and curiosity is what a student brings to the next task once they believe they can do it.
The second is about what students are doing on devices, rather than for how long. The OECD described a shift from creating to consuming, and it was personal device use for leisure that correlated with decline. Those are different activities using the same tool. .
The third is about AI as a tool. The framing was to use technology to deepen learning, with cognitive activation as the marker: enjoyment, curiosity, persistence. AI can be a useful tool for feedback and for personalised practice. And the strongest results came from East Asian systems where AI sat in the hands of teachers rather than in the hands of students working alone.
PISA is a snapshot of fifteen-year-olds taken at one moment, and its associations are not causes. It can tell us what tends to sit alongside what. It cannot tell us which way the arrow points, and it wasn’t built to.
Cautions from a learning scientist
Rose Luckin, Professor Emerita of Learner-Centred Design at UCL’s Knowledge Lab and founder of Educate Ventures Research, predicted the day before publication that somewhere in the release there would be a chart plotting performance against how much technology students use at school, and that it would be screenshotted by the people who believe technology is damaging schools and by the people who believe schools use too little of it, each presenting it as support. She was right.
Her first caution is about who is left in the comparison group. In 2012, 75 per cent of students in Shanghai reported using no technology at school. By 2018 the equivalent figure in Denmark was 0.6 per cent. When the no-technology group shrinks to almost nothing, the shape of any curve drawn against it becomes partly a fact about who remains in that group rather than about technology use itself.
Rose is leading an independent study of Education Perfect, so this isn’t an entirely neutral citation: she’ll be marking EP’s output, and seeing whether the platform’s claims hold up under third-party scrutiny.
Fit for purpose is the whole argument
Name the purpose, and you can judge if the tool is effective. Skip that step and you end up defending or attacking a category, which gets nobody anywhere.
The purposes that matter in a classroom have been well evidenced for a long time. Explicit instruction. Purposeful practice at the right level of difficulty. Feedback that arrives while a student can still act on it. Rosenshine’s principles, Hattie’s visible learning, decades of work on feedback. None of it is in dispute, post-PISA.
So the question to put to any tool is whether it serves one of those. Does it make the student do the thinking, or does it do the thinking and hand back a result? That distinction is doing more work than the word ‘AI’ is. A student who writes an answer and gets questioned on it is practising. A student who prompts for an answer and tidies it is not. The technology is identical. The design isn’t.
The impact is not in the tool, it is in whether the tool is used by design or by default.
The screen at the front of the room
The most consequential screen in most classrooms is the one at the front of the room. How often does this screen feature in the debate? It’s used by a teacher to model, to work through an example, to check that thirty-something students have caught what was taught before anyone moves on. Nobody counts those minutes, and nobody is anxious about them.
The OECD made this point at the launch, and it’s the part I hope gets picked up. The highest-performing systems are the ones using technology this way, with Singapore named as an example. The technology at its best, is directed by the teacher, at the front, towards a piece of thinking the class is doing together.
That tells us the argument was never really about screens, but instead about who is directing the attention on them, and towards what.
What by design actually looks like
St Mary MacKillop College, on the edge of Canberra, is an Australian school modeling a similar approach to the Singapore teacher-led innovation example.
Trent Wilson, their Head of Digital Learning, has adapted Leon Furze’s AI Assessment Scale across both of the school’s campuses, so that every assessment task carries an explicit level, from no AI through to full AI. Students are told which level applies and why. On the junior campus, students start at the bottom of that scale and work with feedback that is teacher-set and task-specific rather than open-ended.
“We’re trying to change the thinking that AI is just something that does the work for you,” Trent says. “We want students to see it as a tool that assists, not replaces, their effort.”
St Mary’s is a school deciding what the tool is for before deciding whether to use it. Effort is where the learning happens, and their design protects it. The rest is implementation.
Which is why efficacy has to be the bar
Governments have paid real attention to AI in schools, and the frameworks are strong on safety, privacy and transparency, but say little about efficacy. We seem to be regulating on what is easy to count: access, minutes, cheating. While there seems to be very little requirement for a tool to demonstrate if it helps a student learn. There are a strong movement to change this and bring a higher bar for technology in the classroom. This is a positive move and is the reason why Education Perfect decided to join forces with Professor Rose Luckin and Educate Ventures Research to provide this kind of evidence across 4 key subject areas: Maths, Science, English & Languages.
Education matters too much to run on anything less than the best evidence available, and that has to include the evidence about individual EdTech companies.
Rose’s interim findings on EP, and the interim results have just been released at educationperfect.com/impact
Next
One thing worth knowing before the PISA commentary settles. Learning in the Digital World has two parts, and only the computational problem-solving results are out. The self-regulated learning analysis, which is the part that speaks most directly to everything above, doesn’t publish until 2027. So for now, we’re reading half a domain.
This post is for my initial, immediate thoughts on PISA. In my next newsletter, I’ll do a deeper dive into PISA 2025 and what it says so far about learning in a digital world.
Read the rest of the series
Well-MeasurED is an ongoing series from Dr Jill McGuire, EP’s Group Head of Learning Innovation, on the hardest part of teaching and what the evidence actually says about it. She publishes it on her own newsletter, most weeks.
Subscribe to The Well-MeasurED Newsletter
About the author
Dr Jill McGuire is the Group Head of Learning Innovation at Education Perfect. She has a doctorate in multimodality, leads EP’s research partnership with Professor Rose Luckin, and co-hosts Well-MeasurED, EP’s industry leader interview series.
