What a pastry cook’s classroom can teach us about AI, effort and learner agency
The second keynote at the FLANZ conference in Dunedin was given by Selena Chan who began with her personal story to set the scene for what she wanted to share.
At the age of fourteen, Selena was doing well at school and quietly disengaging from it. Growing up in Singapore, she found an education system that was exam-oriented, academic and heavily reliant on rote learning. She left with four O-levels and no appetite for the English literature and history that A-levels would have required. What she really wanted was to be a chef, but at the time girls couldn’t be, so she took the nearest open door and became an apprentice pastry cook.
It turned out to be the making of her. Pastry work demands precision, and precision demands persistence, perspiration and practice. It gave a wandering spirit a discipline she says she has drawn on ever since. She came to New Zealand in the late 1970s, started teaching baking in what was then the Christchurch Technical Institute, and never really left. Today she is an award-winning educator, a researcher and author on vocational pedagogy, and part of Ara Polytechnic‘s large AI Ara research project.
As I’ve reflected on her presentation I’ve come to think her story wasn’t simply a charming preamble to the argument she was about to make. It was the argument. Here was someone who disengaged from a system that rewarded getting the answer, and who found her way back to learning through work where you simply couldn’t fake it.
Effort you can’t skip
Selena’s best illustration comes from the bakery. She explained how a commercial dough isn’t one kilo, it’s twenty. When an experienced baker lifts it from the mixer, they can read its temperature, consistency and viscosity, and that tells them what to do next, so the scone comes out the same size and texture every single day. That isn’t recipe-following. It’s a relationship with materials, built over thousands of repetitions, and it ends up so embodied that the baker often can’t explain how they know.
The route to that kind of knowing is deliberate practice, and it follows a familiar pattern – try something, notice it didn’t work, ask what went wrong, try again. Around and around. Skip the loop and, in Selena’s words, nothing happens in your brain – no connections are made. She calls the result of doing that cognitive debt, the learning you thought you’d done but hadn’t, because you never did the work.
We’ve all met this debt. The calculator generation who hesitate over seven times eight. The driver who can’t find their way without GPS and has lost their feel for the place they’re in. Selena is careful to say these skills matter far more in some fields than others, but the trades, where numbers and spatial sense are the job, are exactly where she sees it bite.
What AI is built to do
Selena then spoke about the future role of AI in this learning process. She was adamant in her argument that generative AI is pattern recognition, that it still makes things up, and that it is designed to please us. It agrees with us and it hands over answers. That, she pointed out, creates the cognitive debt she talked about and is precisely the opposite of what the bakery teaches.
Selena used Ara’s own experiments to illustrate the approach she believes we should all be taking, because they’re honest about where this goes wrong. She shared about a chatbot that had been used to help foundation students who had been having major struggles with their writing. This is now in its third iteration and probably its last, because pattern matching turns out to be poor at giving feedback on writing. She described a numeracy tutor that has worked far better, partly because numeracy is more black and white, but mostly because of how the fact that it doesn’t give the answer. It prompts the learner to recall the method, lets them have a go, and walks them through a tutorial when they get it wrong, still without handing over the answer.
Their trades students find this frustrating. All they want is the answer.
That frustration is the point. It is productive struggle, designed in on purpose, and it’s what Selena means when she talks about using AI as a coach and not an answer supplier.
Whose struggle is it?
In my last couple of posts I’ve been circling the question of learner agency with Adrian Woodhouse’s challenge of who owns the learning, and Mark Nicholls’ Blendagogy and what it looks like when the learner is in the driver’s seat . Selena’s talk adds something of an additional twist. If agency means owning your record, your pathway and your choices about when, where and how you learn, then I’d suggest it has to include owning the effort too.
Think about this example. A numeracy tutor that withholds answers only works if the learner understands why. Otherwise it’s one more thing done to them, and they’ll go and find a tool that gives them what they want. Selena’s own classroom practice shows the alternative. When her students’ bakes came out, the products went on the bench with no names attached, and the whole class discussed what hadn’t worked. Too hot, too cold, something weighed wrongly? Her reasoning was that if you can’t look at something imperfect and work out how to correct it, you will never get to something perfect. Mistakes were information, not shame.
A learner who has absorbed that mindset, who sees a mistake as information and not a verdict, can use AI to strengthen the struggle – asking it to probe, to push back, to check their reasoning. A learner who hasn’t will use the same tool to get around it. The technology is identical. The difference is agency.
Becoming
Think again about Selena’s story. The bakery was no easier than school. By her own account it was a major struggle at times, with early starts and long hours. What changed was what the struggle was for. Every cycle of trying, failing and trying again was taking her somewhere: it was making her a baker. In her research she calls this occupational identity formation, learning to think, do, feel and be what the vocation expects of you.
I suspect that’s the piece we keep missing in our work as educators. Learners rarely choose the hard path because we’ve explained cognitive debt to them, or because a tutor won’t hand over the answer. They choose it when the effort connects to who they’re becoming, when they can see themselves in the work and can see their own progress. The learner who sees a future carpenter, nurse or teacher in the mirror has a reason to do the work, and a reason to use AI as a coach and not a shortcut. The learner who sees only an assignment has every reason to find a way around it.
Selena’s talk was situated in the context of vocational training at a polytechnic, but the more I reflect the more I’m feeling that there are important lessons here for those working in the compulsory sector als0 – particularly at the senior end of secondary school. So here’s a question I’ve been left pondering: “how do we help learners become the kind of people who choose the harder path, on purpose?”
