The Crisis Was Never AI: Two Well-Intentioned Mistakes That Broke Learning

Let me start where I always start, because the frame matters more than the argument. Ask almost any teacher, any principal, any parent who has been paying attention, and they will tell you the same thing: something has been wrong with formal education for a while now. Students disengaged. Results flattened or fell. The diploma stopped meaning what it used to mean. This is not a new complaint and it is not an AI complaint. It predates ChatGPT by decades. So when we now say "AI is going to destroy learning," we should at least be honest enough to notice that learning was already in trouble — and to ask what actually put it there.

Because if we misdiagnose this, we will spend the next ten years banning tools and policing prompts while the real disease goes untreated. The machine did not cause the crisis. Our own design did. And I can point to the two design decisions I believe did the most damage — both made sincerely, both defended by good people, both, in the end, devastating.

The first mistake: we fragmented knowledge into competencies

At some point we decided that instead of teaching integral, holistic knowledge, we would teach discrete, measurable competencies. And I want to be fair to that decision, because it came from a genuinely good instinct. For years the criticism of education was that it rewarded knowing over doing — that students could recite and not apply. So we said: let us focus on what students can actually do, on demonstrable skills, on outcomes we can see and check. That sounds sensible. It sounds like rigor.

But watch what happened in practice. To make competencies measurable, we had to make them small. And to make them small, we had to break the whole into pieces. We fragmented learning into isolated micro-competencies, each one defined, assessed, and checked off on its own. We lost the picture of the whole. We produced students who could demonstrate a specific competency on a specific assessment and could not, when it mattered, integrate, synthesize, or apply that knowledge holistically in a real context.

Here is the deeper failure, and it is a structural one, not a matter of any individual teacher's effort. When you optimize a system for measurement, you get measurement — not necessarily the thing you were trying to measure. We told ourselves the competency was a proxy for understanding. Over time the proxy became the target, and the understanding it was supposed to stand for quietly went missing. We optimized for measurement, not for deep learning. And a system reliably gives you what you optimize it for.

The second mistake: we replaced practice with beautiful one-off projects

The second decision is subtler, and I say this as someone broadly sympathetic to the idea behind it. We decided learning should be constructivist — that students should build their own understanding through meaningful, motivating experiences connected to their own reality. As a principle, I do not disagree with that. Making, exploring, connecting to the real world: these are real and durable ideas about how people learn.

But in practice — and this is where the good idea curdled — we let it slide into an extreme. We took "learn by constructing meaningful experiences" and turned it into "do a complex, creative, impressive project, once, and move on." We built beautiful things. Memorable experiences. Presentations that looked like the future. And then the student did it once, and turned the page.

There was no time for practice. No space for repetition. No opportunity for mastery through extensive iteration. We had traded the perfecting of technique for the staging of experiences — and in that trade we quietly deleted the very mechanism by which humans actually get good at anything. Because there is no version of this where it is not true: without extensive practice, there is no real mastery. One brilliant project is a memory. It is not a skill.

Two good intentions, one hollow result

Put the two together and you can see the shape of what we built. One reform fragmented knowledge into pieces small enough to measure but too small to add back up into understanding. The other stripped out the repetition that turns understanding into ability. Neither was malicious. Both were championed by people who cared. And together they produced a generation of students who passed through experiences that looked like learning and left without the thing itself.

They went through gorgeous "constructivist" experiences that were often shallow. They demonstrated "competencies" on assessments and did not retain them. They graduated without genuine command of the fundamentals. They looked like they had learned — and had not developed real expertise. That gap, between the appearance of learning and the fact of it, is the whole crisis in a sentence.

Socrates was right — about us, not about AI

There is an old line here worth sitting with. Socrates warned, in Plato's Phaedrus, that writing would give people the appearance of wisdom without the reality of it — that they would seem to know much and know nothing. He was wrong about writing. But he was pointing at something real: the illusion of knowledge is a thing that can be manufactured. You can build a system that produces the sensation of having learned, without the substance.

And here is the part I most need people to hear. We built exactly that system — and we built it before AI existed. The illusion of learning is not a side effect of the chatbot. It is the direct output of educational designs we chose, on purpose, with good intentions, over the last several decades. Socrates was right that the illusion can be manufactured. He just had the wrong culprit. It is not the tool that manufactures it. It is our flawed design.

Why this reframe changes everything

I am not making a historical point for its own sake. I am making it because the diagnosis determines the treatment. If you believe AI caused the crisis, your instinct is defensive: ban it, detect it, wall it off, and try to preserve the system as it was. But the system as it was is precisely what was failing. Preserving it is not a cure. It is a way of protecting the disease.

If, instead, you accept that the crisis is ours — that we fragmented what should have stayed whole, and deleted the practice that builds real ability — then the question about AI becomes a much better one. Not "how do we keep it out?" but "can this finally help us fix what we broke?" Can it help us restore holistic understanding instead of atomized checkboxes? Can it give every student the extensive, repeated, well-supported practice we stopped making room for? Those are the questions worth years of our attention. And you can only get to them once you stop blaming the machine for a wound we inflicted ourselves.

The honest starting point is not comfortable, but it is liberating. Education was not working. We know why, and it was not artificial intelligence. Which means the fix is within our reach, because the cause was within our control all along.

A working draft. This essay lays out my position; a forthcoming revision will weave in sourced research — the competency-based-education literature, the constructivism and constructionism scholarship, and the Socratic Phaedrus argument — with full citations, and will replace directional claims with verified figures where they exist. If you have work I should read before then, tell me. — Carlos Miranda Levy

Four perspectives