Stephanie Holt teaches English at Riverside International School in Prague, works on school transformation, and writes The Educator's Notebook. Recently she put into words something we had struggled to say ourselves. Schools measure how much a student has accumulated. They rarely measure whether that student's thinking is developing, or whether any of it carries over to something new. She calls those three dimensions Up, Along and Across. We asked her to talk us through them.
Up is accumulation. How much a student knows, how much of it is correct, how that compares to the rest of the class. It is countable, and it is what nearly every school reports on. Along is development over time: practice, feedback, mistakes, another attempt, and whether responsibility for the learning is gradually shifting from the teacher to the student. Across is transfer. Whether a student recognises that something learned in one place is useful in another, and can defend the choice they made.
One line from the article "We Measure Everything Except the Learning" stayed with us most. "Much of what schools call assessment is really an autopsy," she writes. "We inspect the finished product after the intellectual life has gone out of it, identify what succeeded or failed, attach a number and file the evidence." That is more or less the problem we built Examplary to work on, written down more sharply than we had managed ourselves.
We spoke to Stephanie about the framework and where assessment is heading. Here's our conversation, and what we learned from it.
Where did Up, Along and Across come from?
"My daughter. She's an English language student at Lancaster, and we had one of those Sunday conversations you get when your mum is an English teacher. She told me about this theory she was working on and I'd never heard of it. I thought: that fits assessment exactly. So I took it to Alexander Harris, who I co-research with, and said I think this is the framework for assessment as AI becomes normal in classrooms."
Where are schools now?
"On Up. You start at five years old and the only thing that matters at the end is the grade, whether that's the International Baccalaureate (IB), A-level or SAT. Everything in between is an attempt to move a child from here to there as fast as is humanly possible. That's Up logic. It's getting on the motorway at Leeds and driving to London. You miss everything worth seeing and you stop at one terrible service station for bad food. That's pretty much the diet our children are on. As teachers, we've always known it isn't a good system."
England scrapped the national exams at the end of Key Stage 3. Didn't that give schools more room to teach beyond the exam?
"No. Key Stage 3, so eleven to fourteen, used to end with exams in Year 9. Now it doesn't, and because there's nothing high-stakes left there, it's become a bit of a wasteland. It's just pre-GCSE. It was supposed to be freeing. In practice it just took out a motorway service station. That's all it did. And it isn't surprising. If a school is judged on whether it delivers GCSEs well, it will make sure it delivers GCSEs well."
"In practice it just took out a motorway service station. That's all it did."
In the Netherlands we see something similar. Teachers find formative assessment important, but very few change how they actually work, because the standardised exam at the end is what the school gets judged on.
"I don't agree with that, actually. If the learning process is done well, the exam results follow. Any teacher worth their salt can teach an exam in about two hours, because all you're teaching is its format. That isn't the important bit. The real question is what the exam asks for. If it's regurgitation-heavy, it's memorisation, and the process doesn't matter at all. Look at India's medical entrance exam, where students need close to full marks to get a place. That tells me the assessment is wrong, not the students. Compare that with the IB. It's far from perfect, but it includes unseen texts, so you can't get through on memory alone. If you can't analyse a text you've never seen before, you can't do it. So the process has to be there."
Say a teacher reads this and wants to work differently. Where do they start?
"They have to accept it first, and that's quite difficult. When a child hasn't done well on an assessment, I always believe I need to teach them in a different way. It's on me, not on the child for not studying hard enough. Once you accept that, you rewrite your processes. Look at how industry is handling AI. The companies that rewrote how they work are doing better than the ones that bolted AI onto what they already had. It's exactly the same in teaching. If you try to bolt formative assessment onto an Up system, it just doesn't work."
"If you try to bolt formative assessment onto an Up system, it just doesn't work."
And what does it take from school leaders?
"It has to come from the top. Most primary schools that follow the UK curriculum don't give grades at all any more. They use descriptors: working towards, at standard, above standard, exceptional. That stops children and parents getting obsessed with what is basically a made-up number, and you can still see clearly whether a child needs extra help or is flying. In my last school we moved Key Stage 3 onto the same system, so the conversation is about progress instead of attainment. That matters, because the moment you put a grade on a piece of work, the child will not read the feedback. There's no point putting it on. They just get obsessed with the grade."
What about the students who are already top of the class?
"If you focus on attainment, the A-star child usually makes no progress at all, because they're already doing great. They're top of the class at eleven, they do very little for three years, and then GCSEs arrive and suddenly they have to work. Focusing on progress says something different: you're great, but you can be greater."
"The moment you put a grade on a piece of work, the child will not read the feedback."
So they never learn how to study, because they've never had to?
"Very much so, and with AI, metacognition is becoming even more important. I've got an academic article coming out on this. When I was working with metacognitive AI tutors, the high-achieving children wouldn't engage with the metacognition, because their mindset is fixed and they already think they're amazing. They rushed it and produced an Up product. The middle and lower groups listened to the tutor, and the kids who went Along did far better in the end. I had an eleven-year-old who had never vibe coded before teach herself and build an app, because the metacognition helped her do it. She's literally my perfect specimen for the study."
Does any of this make the teacher less important?
"More important. Much more. In an Up system the teacher is a deliverer. You deliver knowledge, you deliver skills, and the job is quite dull. Anyone can stand in front of a class and read from a textbook, and that's not what any of us went into teaching for. The less time I spend marking by hand and working out what to teach next, which used to be my skill, the more I can use my skill to support a child one-to-one through their social and emotional problems. That's how my job is changing."
Do you use AI for feedback yourself?
"Since the moment ChatGPT came along. I was like, huzzah! My students didn't like it at first, because they wanted me. So I explained what they were getting. The feedback AI can give is far more detailed and specific than anything I can physically write for every child, and it's personalised rather than one broad brushstroke across the class. I don't have the time to mark like that. But I put a personal comment on every piece, a smiley face or a 'brilliant effort'. Children love their teachers and want to do well for them, so it matters to them that I've read it and not just fed it to a machine."
"The feedback AI can give is far more detailed and specific than anything I can physically write for every child, and it's personalised."
How do you get from Along to Across? What does that look like in a classroom?
"Across doesn't really exist yet. But go and watch an early years class. Reception teachers, who teach the four- and five-year-olds, don't plan lessons the way you'd think. They plan a theme, autumn leaves say, and then they capture the children's learning in the moment: a photograph, a short video, a note. To sign off an early learning goal, the child has to have done it three times, like tying their shoelaces. I think that's where we're going. If a child has pulled a leaf apart and got interested in its skeleton, the teacher moves their science on right there. If another child is doing sink or float with a leaf, they plan for that child, in the moment. Our reception teachers are probably the most skilled practitioners in the world right now."
That sounds hard with thirty children at once, when every subject is its own silo.
"Not if you've got AI helping, because we do it already. I've even taught classes of forty-two, and the only way is to teach in slices: these children need extra support, these are the middle, these are my top end. You teach the content, and then in the moment you move it up and you move it down. To see where each child is, use descriptors instead of percentages. Whatever the task, you can say this shows a transferable skill in English, that shows a skill in history. And yes, subjects are silos. That's why it's going to be AI. In primary school one teacher does it all, which is a big part of why children progress so well there. In secondary it splits, and each subject teacher is effectively an individual AI agent for their subject. What I'm talking about is bringing it all together, so the English AI feeds the humanities AI, which feeds the science AI and the maths AI."
Stephanie Holt writes The Educator's Notebook. "We Measure Everything Except the Learning" is here.