Errors Are the Switch That Triggers Deep Learning
— What did you say, the last time your child got a question wrong?

We live in an age of unprecedented intolerance for error. On social media, everyone projects perfection. In the workplace, one misstep can mean one performance review. In schools, scores and rankings tell students every single day — "You're not good enough."

Education ought to be the last domain where mistakes are permitted. Because learning, at its core, is the process of constructing understanding through error. Yet the reality is that our classrooms are becoming more and more afraid of being wrong — teachers fear that a student's wrong answer will make their lesson look less than perfect; students fear that saying the wrong thing will bring ridicule; parents fear that a red cross means their child is falling behind.

In Part 2 of this series, I told the story of a parent who messaged me on WeChat. Her child's homework was covered in red crosses, and her hand was trembling as she typed. In that piece, I addressed the question of "stagnation" — why your child's scores not moving doesn't mean they're not learning. Today, I want to answer a different question: what about the red crosses themselves?

After twenty years as a principal, I have become convinced of one thing: we have been thinking about error entirely backwards. Error is not the enemy of learning. Error is the switch that triggers the brain's deepest learning mechanisms.

In 2012, Manu Kapur of Singapore's National Institute of Education conducted an experiment that sent shockwaves through the global education community. His question was deceptively simple: what happens if you let students fail first, and only then teach them the correct answer?

He divided ninth-graders into two groups and gave them the same mathematical concept to learn.

Control group: The teacher first explained the concept and formula clearly, then students worked on problems. This is the most traditional instructional approach — and exactly what every parent imagines when they picture a "good classroom."

Experimental group: The teacher explained nothing. They handed the students problems directly — problems involving concepts the students had never been taught. The students had to rely entirely on intuition, guesswork, and fragments of prior knowledge. They had to throw themselves against the wall with nothing to go on. The result was predictable: nearly every student failed. Some failed spectacularly.

Only then did the teacher step in to explain.

It sounds like a colossal waste of time. Let them flail around randomly, and only then teach the correct method — this goes against every instinct we have about education. But when the final test results came in, they silenced everyone:

Kapur named this phenomenon "productive failure." His explanation: those who struggled before receiving the correct method had given their brains a kind of preparatory rehearsal for understanding. Their errors told them "what you know, what you don't know, and which of your intuitions are wrong" — insights that no amount of teacher explanation could ever produce.

Error is not waste. Error is exploration.

In the YouAI University AI Literacy course we introduced at our school, there's a student named JAM. He's not the kind of "model student" who follows all the rules — his mind is filled with black holes, singularities, and multiverses, and his thoughts jump at lightning speed. But when he first used an AI generation tool, his expression came across as fragmented, low in logical coherence.

One lesson's theme was "Universe-Themed Creation." JAM tried to use AI to recreate the black hole in his imagination. His first prompt went like this:

The image AI produced was utterly mediocre — even showing distorted color blocks that completely failed to capture the sense of awe he was describing.

Any teacher eager for results would have handed him a "standard prompt template" right then — subject + details + style + lighting. Write clearly. Write in structure. JAM could have plugged in the template and produced a polished space image in three minutes. But that would have been a "correct" image — not JAM's image.

His teacher made a decision that seemed nearly irresponsible at the time: don't correct him. Don't teach the template. Let him keep experimenting his own way.

Over the next class sessions, JAM went through a complete learning cycle.

The Frustration Phase. The images kept coming out wrong, and he grew increasingly restless. His prompts were either too vague or too chaotic, and each AI return was more bizarre than the last. Measured with the behavioral ruler — his error count was the highest in the class.

The Turning Point. Because no one judged his "wrong images," he grew bolder. He started stuffing his accumulated astronomy knowledge — accretion disks, event horizons, spacetime curvature — helter-skelter into the prompt box. Not to "write the right prompt," but to see what would happen if he threw it in. Every wild experiment was a cognitive variation.

The Breakthrough Phase. To make the AI understand him, he spontaneously looked up the precise definitions of these astronomical terms and observed how word order affected the visuals. Those earlier "wrong images" became the most valuable feedback — they showed him the boundaries of the AI: what kind of description produced what kind of effect. He was no longer guessing blindly. He was constructing his own system of "universe creation logic."

While other students were still mechanically applying the formulaic prompts the teacher had given them, JAM was already demonstrating remarkable transfer ability. He could not only use AI to draw the black hole in his imagination, but could explain why his prompts worked the way they did. What he learned wasn't "how to write a good black hole prompt" — it was a transferable understanding model applicable to any creative theme.

Kapur's experiment proved this at the behavioral level. Neuroscience has provided evidence at an even deeper layer.

Neuroscientist Clay Holroyd at Columbia University discovered that when a person makes an error, the anterior cingulate cortex releases a signal called error-related negativity (ERN). This signal is essentially an "attention redirection" command — "Alert! Something's wrong here! Adjust your strategy!"

Even more critically: the strength of the ERN signal is positively correlated with the quality of subsequent learning. In other words, the more "alert" the brain becomes upon making an error, the deeper the learning. The people whose ERN responses are strongest when they make mistakes are precisely the ones who ultimately achieve the deepest mastery.

Another study comes from Michelene Chi at the University of Pittsburgh. She had students "explain to themselves" while learning physics concepts — not listening to a teacher, but thinking aloud as they went. She found that the students who exposed more of their own misconceptions during self-explanation actually scored higher on the post-test. Chi called this the "self-explanation effect" — it was 40% more efficient than teacher explanation.

These two findings point to the same conclusion: errors trigger the brain's deep learning mechanisms. When you give a wrong answer, your brain hasn't wasted a response. It has completed a full neural cycle: detect problem → activate attention → prepare correction. Every time this cycle runs, your cognitive structure becomes more robust.

At this point, a critical question emerges: if error is merely raw material, why is it that in some classrooms errors propel people forward, while in others, errors make people retreat further and further?

The answer: error itself is neither good nor bad. It's the feedback that follows the error that determines whether it becomes a learning resource or a psychological wound.

I've seen this scene play out far too many times. A student gathers their courage, stands up, and gives a wrong answer. The teacher furrows their brow: "Wrong. Sit down." The class falls silent. That student does not raise their hand again for the rest of the semester. Their brain didn't just learn "that answer was wrong" — it learned "in this class, being wrong means losing face." And the signal of "losing face" activates the amygdala's fear circuitry, not the prefrontal cortex's learning circuitry.

Now, change the scene. Same question. Same wrong answer. The teacher pauses: "That's an interesting angle. Can you tell us how you arrived at that?" The student begins explaining their reasoning. Halfway through, they stop themselves — "Oh, wait, I got it wrong." They discovered the contradiction on their own. The teacher smiles: "You found it yourself. That works far better than me telling you."

Same error. Two responses. Two completely different learning outcomes. The difference lies not in the error itself. The difference lies in what kind of feedback was waiting for it.

This principle maps directly onto the algorithm I've been returning to throughout this series — Variation → Selection → Retention.

Errors are "variation" — they are the source of cognitive diversity. Without variation, selection has no material to sift through. The teacher's response is the "selection pressure" — it determines what kinds of cognitive attempts get encouraged and what kinds get eliminated. A punitive response aims at eliminating "errors," but what it actually eliminates is the capacity to produce variation itself. An exploratory response selects for "structures of understanding" — your direction may be off, but your reasoning has value. Keep going.

A classroom that treats errors as enemies to be exterminated doesn't just kill wrong answers. It kills students' courage to generate cognitive variation. And learning without variation is like evolution without genetic mutation — nothing remains but stagnant water.

At this point, a practical question is unavoidable: How do I know whether my approach to errors is right? How can I tell whether I'm tolerating errors — or unconsciously punishing them?

Behavioral science offers an extraordinarily simple standard: watch the frequency of behavior. Don't look at your intentions. Don't trust how you feel about what you're doing. Look at what the child does after your response — does the behavior increase or decrease?

If you read the story in Part 2 about the parent who messaged me on WeChat, you already know my response back then. Today, I'm taking that response one step further — from "don't panic" to "change the way you see error."

When your child brings home a homework page covered in red crosses, here are the three things you should never say:

"You got it wrong again?" — Getting this problem wrong is not the same as "you're making the same old mistake again." Don't personify the error. An error is an outcome, not a character flaw.

"How can you not know something this simple?" — What's simple to you may not be simple to your child. Moreover, "simple" is a hindsight perspective. The reason you find it simple is that you already understand it. Your child is still in the process of understanding.

"Everyone else gets it — why don't you?" — Comparison is the fastest confidence killer. Every child's path of cognitive variation is different. Some need three errors to reach understanding; some need eight. The one who needs eight may not be learning any worse than the one who needs three — their cognitive structure may actually be more robust, because they explored more side paths.

And here are the three things you should say instead:

"This mistake is interesting — what were you thinking?" — Shift the focus from "outcome" to "process." Behind every wrong answer hides a valuable line of reasoning.

"Take another look — can you spot what's off?" — Don't give the answer directly. Let the brain's ERN signal run its full cycle. Self-discovered error, self-corrected — that is the highest-quality learning there is.

"You got this one wrong, but the direction you were trying was right. Just adjust this part." — Localize the error; don't let it spread. "Your effort wasn't wasted. You just need one more step."

In my twenty years as a principal, the students who worry me most are not the ones with "poor grades." Not the ones with "weak foundations." It's the ones who are afraid to make mistakes. Their homework is flawless. Their test scores are stable. They never say a single unnecessary word in class. They live inside an absolutely safe zone — a zone free of red crosses, and free of real learning.

Conversely, the students who make me think "this kid is going to be remarkable" have never been the ones with the most perfect scores. It's the one who dares to say out loud, in class, the "wrong answer" that everyone else was thinking but no one dared to voice. It's the one who chases the teacher after an exam, clutching their marked paper, demanding to know "why did I get this wrong?" It's the one whose homework has the most red crosses — but next to every question, in their own handwriting, is a reflection note.

The raw material of learning has never been "the correct answer." The raw material of learning is the complete cycle: error + feedback + correction. The correct answer is merely the endpoint of that cycle. Mistake the endpoint for the starting point, and learning ceases to exist — all that remains is memorization.