Curse of knowledge in teaching

Why can someone who learned a skill last year explain it better than the person who has taught it for thirty years?

We assume teaching quality tracks subject mastery: know more, explain better. It is such a natural assumption that we build hiring around it. Yet almost everyone has met the counterexample — the graduate student whose explanation lands, and the eminent professor whose does not, on exactly the same material.

The lazy diagnosis is that the professor cannot be bothered. But the pattern shows up in people who care enormously about teaching, so the assumption worth questioning is a different one: that knowing a thing leaves you able to imagine not knowing it.

Try the imagining yourself. Look at a written word in your own language and attempt to see it as a shape rather than a meaning. You cannot. The meaning arrives before you can intervene. Reading was once effortful and is now unavoidable, and no act of will restores the earlier state.

That is the whole phenomenon in miniature. What happened to reading happens to every skill practised long enough: the separate steps get chunked, then run automatically, and once automatic they are no longer available for inspection. You did not forget them — something worse. You cannot examine them, because they no longer occur to you as steps.

Now ask what a teacher does when judging whether an explanation is clear. Having no access to the student's mind, they simulate: they consult their own reaction and adjust. But the starting point of that simulation is a mind in which the answer is already present, and adjustment away from an anchor is famously insufficient. So the estimate lands far closer to the expert's state than the learner's — and it does so smoothly, with no sensation of error.

Elizabeth Newton's experiment at Stanford made this measurable. Tappers tapped out a well-known tune on a table; listeners guessed which. Tappers predicted around half the listeners would get it. About two and a half percent did. To the tapper the melody is unmistakably there, playing in the head, and it is genuinely impossible to hear the taps as a listener hears them — a bare, arrhythmic knocking. The tapper is not careless. The tapper is unable. Camerer, Loewenstein and Weber had earlier named the effect in economics, showing that people holding private information cannot fully discount it when predicting the judgements of those who lack it — and that incentives to get it right reduced the error without removing it.

So follow that into a classroom. Which parts of an explanation get skipped? Not random parts. Precisely the parts that have become automatic — which, for the expert, are the parts that were hardest longest ago. The foundational move, the notational convention, the reason this step is even legal: these vanish first, because they are the most practised. What survives is the material still effortful for the expert, which is the advanced material. An expert's explanation is therefore not simply too fast; it is systematically hollowed out at the base.

And the recent learner? They are not smarter about pedagogy. They simply do not have to simulate. Their own confusion of last year is still in episodic memory, retrievable as an experience rather than reconstructed as a theory — so they know which step cost them a week, and they stop there.

Three refinements, because the conclusion is easy to overstate.

This is not expertise reversal, though they are cousins. Expertise reversal is about the learner: guidance that helps a novice can hinder an advanced student whose own knowledge conflicts with it. The curse of knowledge is about the teacher — a metacognitive failure to model another mind. One concerns which method suits whom; the other concerns why the expert cannot tell.

Nor does it mean the novice is the better teacher. The recent learner has an accurate model of difficulty and little else — a thin repertoire of alternative explanations, no map of where misconceptions lead, and a decent chance of passing on their own half-formed version. The expert's advantage is real and lies exactly where the novice is weakest. What the comparison exposes is that these are two separable competences which we have been treating as one.

And it resists insight. Being told about the curse does not lift it, any more than being told about an illusion straightens the lines. What works is external evidence rather than introspection: watching a novice attempt the task, collecting the wrong answers, asking someone to explain it back.

• Why hindsight bias — the sense that an outcome was obvious once known — is the same mechanism aimed at the past.
• How instructional designers substitute measured evidence of difficulty for the teacher's intuition.

Every term the collection defines is gathered in the glossary.