I spent a few days recently sitting in primary school classrooms, mostly beside Learning Support Officers. If you haven’t met one, they’re the person crouched next to the student who can’t get started, quietly rewriting questions or activities so it actually makes sense. Or helping drive focus and task completion, emotional regulation and generally supporting classroom engagement.
Here’s the thing that stuck with me. In Australia, adjusting learning for students with disability isn’t a nice-to-have. Under the Disability Standards for Education 2005, schools have to make reasonable adjustments so students with disability can take part on the same basis as everyone else.
That’s the right call, but a requirement doesn’t always come with a toolkit. If we expect educators to adjust, they need techniques, tools and realistic ways to do it in a class of 25, while juggling all the other responsibilities and potential classroom chaos.
Same goal, but different way in
The name for this is differentiated learning, and Carol Ann Tomlinson is the person most associated with it. Her model is pretty clear-cut. You can change the content (what students work with), the process (how they make sense of it) or the product (how they show what they know). And you base those changes on a student’s readiness, their interests, or how they learn best.
What you don’t change is the goal.
A student learning fractions still learns fractions. It isn’t about making the work easier. It’s about removing the bits that were never the point in the first place.
The barrier is usually in the design
When I’m designing learning at work, Universal Design for Learning is the framework I lean on most. It was developed by CAST, and I wrote about it in a previous post on there being no default learner. The short version: the barrier sits in the design, not in the learner. So you build in more than one way to engage, more than one way to take it in, and more than one way to show it.
Tomlinson and UDL come at it from slightly different angles. Differentiation adjusts for the student in front of you, whereas UDL designs the flexibility in from the start, so you need fewer adjustments later. In a real classroom, I think you need both.
Think about what actually stops a student getting through a worksheet:
- Dyslexia. The sheer amount of text.
- ADHD. Ten questions on one page and no obvious place to start.
- Autism. A topic that just doesn’t connect.
- Low vision. A graph they can’t see.
None of those are about whether the student can do fractions. They’re about how the task was put together, and the catch has always been time. Rewriting one activity five different ways isn’t realistic for one teacher on a busy day.
Starting with what they love
This was the bit that got me most excited.
One student is autistic and loves Ford cars. Not “likes”. Loves.
So we took a photo of the class fractions sheet, gave it to Copilot, and asked it to rewrite the activities around Ford cars with the same learning goal. Out came: “A Ford Ranger has 4 tyres and 1 needs replacing. What fraction is that?”
Same maths but very different reason to pick up the pencil.
There’s good research behind this. Gunn and Delafield-Butt (2016) reviewed the studies on teaching autistic children through their special interests. Across them, bringing a child’s interest into the learning generally improved engagement, behaviour and learning, with very little evidence of harm. For years the instinct has often been to manage special interests away, or use them as a reward once the “real” work is done. The research suggests they’re something to build on instead.
The same idea works for reading. Ask a student what they’re into, then ask AI for three minutes of text on that topic at their year level to read aloud. I suspect it beats the tired readers on the shelf every time.
What this looks like in a classroom
Not everything needs AI. Some of the most useful tools are already sitting on school devices. A few ideas the support staff came up with:
- Use what’s already there. Immersive Reader reads text aloud and lets a student choose their own colours and spacing. Dictation lets a student whose handwriting is slow get ideas down at the speed they think.
- Give handwriting a second life. Take a photo of a student’s handwritten work, bring it onto the computer and have it read back to them. Hearing their own words often helps them see what to add next.
- Practise without an audience. For a student who dreads presenting, PowerPoint’s Speaker Coach lets them rehearse alone and get feedback on pace and filler words. As many goes as they need, and no teacher sitting through every one.
- Shrink the task, not the goal. A long comprehension task becomes five short clues to crack the case, one at a time.
- Stretch it the other way too. For the student who finds it all too easy, the Ford fractions become fuel efficiency ratios.
A few guardrails
The tools are new, and the research on AI for this is still catching up. The principles they’re applying aren’t new at all. But judgement still matters:
- Describe the need, not the child. Keep names and personal details out of AI tools, and use the ones your school has approved.
- Check before it reaches a student. AI is very confident and occasionally very wrong, especially with maths.
- Keep the bar where it is. Change how the task is presented, not what’s expected, unless the student’s learning plan says otherwise.
- Don’t single anyone out. A Ford-themed worksheet often appeals to half the class. Share it widely and nobody’s the kid with the “different” sheet.
Requirement, meet toolkit
The Standards say adjust, Tomlinson tells us what we can adjust, and UDL tells us to design for it from the start.
And for the first time, the tools make it realistic for a busy educator to actually do it, in minutes rather than the hour nobody had.
The fractions didn’t change. The Ford Ranger just gave them a reason to start. And I think much of this will be a game changer for roles that need fast flexibility and options to suit learner needs.