Most of the argument about AI in education is an argument about permission. Should we restrict it or allow it? Should we prohibit it or give it a different purpose? Neither of these will transform education, not so long as we hold the same values, and those values turn outcomes into ends.
At Green School we go analog first. That does not mean analog only, and it does not mean we are restricted. It means that at the moment we come into an encounter, we ask what AI might help with, and especially how we might not use it. I will come back to what that looks like in practice. First it is worth being clear about why the permission argument cannot get us anywhere.
The outcomes do not move
AI will not transform education because the grammar of education stays the same: young people sorted based on demonstrations of knowledge and skills that can be easily commodified. The same outcomes are measured as mastery of content or of capabilities, the same outcomes individualise young people, and the same outcomes produce legibility and documentation of what a young person demonstrated at a specific moment in time. Legibility isn’t the problem, it’s who decides what gets written down and how decisions are made by those who read it.
Assessment is not stupid. A test or essay need not be regurgitation. It can ask a student to work content in response to a prompt and show real sophistication of thinking, and skills are taught as the means to do exactly that. Elsewhere a student reflects on experience or process, which tells the assessor something about decision-making and about what was learned. Or a student does something and earns a badge, because they met, or supposedly met, certain competencies at a certain point.
This is where the current grammar of school is strengthened. These are tweaks at best, tricks at worst. Nowhere is any of this forward looking. Nowhere does it ask what life-giving possibilities the mastery of that content and those capacities might open up.
Content does not really matter unless it is applied to something. I am not suggesting that content isn’t important, I am saying that it has to be applied in order to matter, in the different senses of the word. Competencies are only important as and why they’re applied. What does it mean to show you can collaborate, communicate or be creative? Some of the most brutal authoritarians of the twentieth century were tremendously good at collaboration, they were exhilarating communicators, and they were awfully creative. It is how these are applied that matters, and what they open up.
Ends and means
Regenerative education is learning through participation in place, with human and other-than-human life, in ways that deepen reciprocity. Deepening reciprocity means engaging with the world so that each encounter creates spaces for new life-giving possibilities to emerge.
Conventional education systems see content and capacities as an end. Regenerative education, which rests on relational ontology, sees them as the means to get to ends that are life-giving possibilities. What do they open up? How do they open up possibilities that are life-giving? This matters far more than treating them as outcomes in themselves, which might be used for all kinds of purposes.
The proof is in the pudding: Conventional education has left us with extractive systems. It has led to fossil fuel extraction, to the destruction of forests and biodiversity, to consumerism. Just because you know how to get oil out of the ground does not mean that you should.
It also produced everyone now naming the problem, and that is not a contradiction. Systems produce what they cannot contain. Gandhi was trained as a barrister at the Inner Temple, Fanon in the French medical faculty, and both turned that formation on the order that gave it to them.
The wrong conversation
Seen this way, AI is a different mode of getting to the same place. The mainstream conversation reinforces the current system rather than transforming it, because AI changes the vehicle through which the same outcomes are valued, and we still value the same things. Which is why the conversation is not about what we value, but about how or how not to achieve it. And through AI, greater efficiency doesn’t lead to transformation. Jevons paradox tells us why: When a system becomes more efficient at producing something, it does not consume less, it produces more. The saving is absorbed as expanded output. A system that gets better at producing the same outcomes will produce more of them, faster, cheaper, and at a bigger scale, and that is what AI offers education, which is the opposite of transformation.
So the questions we in education are asking are the wrong ones: Whether we should use it or not. How we stop cheating. How we expand learning through tutoring. Redesigning assessment so that it is cheat-proof is useful (is it?), but it still gets us to the same question of outcome. Using AI to accelerate learning through tutoring is still outcome. It is about efficiency rather than robustness. It is about efficiency and performance rather than the ability to deepen reciprocity, to consider relationships first, and to ask how we might apply what we have for life-giving possibilities.
Discernment
At Green School we appreciate that we are still inside the system. Our position on AI is not about prohibiting it and not about enhancing with it. It is about our relationship with AI, and specifically about not using AI, in order to open up life-giving possibilities in face-to-face encounters, with other humans and with the other-than-human world, the living world.
This happens through discernment. AI is a digital technology that can help us do things we do not necessarily want to do and that do not enrich us. We use it for low-level administrative tasks, things that are fairly mechanical, plainly boring, and that divert our energies from encounters and from work that is far more enriching. When we discern that AI corrupts or short circuits our capacity for deeper relationships, or for life-giving possibilities, or for learning, then we do not use it.
The failure of discernment has a name. Thought laundering is presenting AI-generated judgments as your own to gain authority. This is not cheating; cheating is what the outcome frame calls it, because that frame can only see a rule broken and a result that was not earned based on its value system. Thought laundering is something else, it is the substitution of judgment, used afterwards as credibility and the accrual of educational or cultural capital. Which returns us to who decides what gets written down, which is about power, and that needs its own article. Assessment confers authority through the record and documentation. Thought laundering claims authority without one. Both are questions about who gets to be believed, and neither is settled by the young person themselves.
Which is why no rule will settle this. A rule can only reach the moment of submission, and by then the thinking has already been given away or it has not. What we are asking young people to do is notice the difference themselves, in this moment, in this place. They may not always get it right, but discernment is a practice, and it is forward looking. It cannot be assessed because it is how we enter the encounter, shaping the encounter and the possibilities that emerge from it.
The grammar of school demands that everything be assessable. This is what capitalism does everywhere: everything commodified, everything given a value, everything measured. Discernment does not submit to that demand, which does not make it invisible. You know it is happening because of what follows. Not in the moment and not on a rubric, but over time, in what comes together, in the quality of what a young person does next and what they decide to do with what they know.
Content returns
This is where content returns, and it returns as more than application. Expertise is what lets you see what AI has handed you. As a historian, I know a great deal about currents in history, say the French Revolution, Fascist ideology or the advent of capitalism. So when AI gives me something on it I can tell almost immediately whether it is any good, where it is thin, and what it has quietly got wrong. Put me in front of organic chemistry and I have none of that. I could get through it, but I would not know what I was looking at.
So content matters, and it matters here in a specific way: it sharpens discernment. It makes you quicker at seeing whether AI has saved you time or cost you something, whether it is worth having as a thought partner, and whether you are better off doing the thing yourself. Expertise lets you put the laptop down sooner.
None of which means you must be an expert before you can discern. Discernment does not begin with expertise, it begins with the question of life-giving possibilities, and anyone can ask that from the first encounter. What expertise does is hone it. The novice is not crowded out, they may just be slower and more reliant on the encounter itself where the expert can discern more quickly. Which is an argument for going analog first, not against it.
AI as ordinary
All of this pulls AI down from the godly to the normal. The whole point of AI is to save time in order to close our laptops and go and do other things. Much like the washing machine, or really the whole set of household machines that arrived alongside it, which did release time, and released it on a scale that reshaped who could take paid work. But the machines freed the hours, they did not decide what the hours were for. The market was waiting to absorb them, and largely did.
Which is where saving time has to be distinguished from efficiency as the system means it. Efficiency in the extractive register frees hours in order to produce more of the same, faster and at greater scale. That is Jevons, and it is what AI offers education. Here the hours are released into encounters, into participation in place with human and other-than-human life, into deepening reciprocity. They do not go back into the machine. Discernment is what decides between the two. What needs to be decided now is whether we let capital absorb the time AI saves and capture us for higher extractive productivity.
AI is not the end, it is a means to other things and to richer lives. It is not a way to build an edge in the market, which would only continue to individualise.
The students said it first
None of this is something I came up with on my own.
At Green School, we talked to many students. They told us they thought more deeply and more richly without computers and without AI. They preferred pen and pencil. They said it let them engage more deeply with the material, and work in a stream of consciousness, moving into artwork and flows and pictures and diagrams in ways they could not otherwise. The materiality helped. So did the conversations.
That is a preference, but there is more to it than that. They also told us they did not want to do anything that short circuits their thinking, and that AI helps them in many ways but will always come second. OK, not all the students, but in high enough proportions that we are onto something. These are the spaces for life-giving possibilities that open when students co-design the learning experience.
That is discernment, even if young people may not always use that word. It is how they enter the encounter.
What we cultivate is thinking and feeling and embodiment and relationships. AI may or may not help with that, and it is never the outcome. It is a way to offload what we do not want to do, so that there is more time for what we do.
The point of AI is discerning when not to use it, so that life-giving possibilities can open up.
Coconut Thinking 