Translating into Human: Learning with People

You've worked the topic out. It all came together, the picture in your head is clear. And then someone asks: "so explain how it works?"

And out comes: "well, it's... basically the data sort of goes... and then it kind of does it itself... it's like... you know?"

They don't know. And, more unpleasantly, you've just discovered that you didn't quite know either.

A draft in your own language

In your head, knowledge exists as a private draft — a mixture of fragments, pictures, gestures and the feeling of "well, it's sort of clear." Such a draft reads perfectly smoothly from the inside: its connectives are replaced by a sense of coherence, and its holes by the confidence that something is there.

Explaining to another person is a translation of that draft into a real language. And translation is merciless: you can't translate a gesture, you can't translate "well, it's sort of." Every place where you had a feeling instead of a thought surfaces during translation as a blank line.

Knowledge that doesn't translate into human language is still a draft.

And notice: the draft becomes a text in the process of translating. You don't understand first and explain afterwards. You understand by explaining.

The illusion of explanatory depth

This gap has precise experimental confirmation.

In 2002 Leonid Rozenblit and Frank Keil asked people to rate how well they understood how everyday things work: a zip, a toilet, a helicopter. People confidently gave themselves high ratings. Then they were asked to give a detailed step-by-step explanation of the mechanism — and the ratings dropped sharply, their own ratings, after the attempt.

The authors called this the illusion of explanatory depth: we systematically confuse familiarity with a mechanism and understanding of it. The illusion works most strongly on devices and systems — that is, on most of what we study.

It's cured by one action: attempting to explain. And the attempt absolutely has to be made out loud or in writing — a mental "well, I'd explain it roughly like this" doesn't break the illusion, because it stays the same draft.

Explaining means learning

Then comes the pleasant part: explaining doesn't only diagnose, it teaches.

John Nestojko and colleagues (2014) ran an elegant experiment. Two groups were given the same text. One was told there would be a test afterwards. The other was told they'd have to explain the material to another person. In the end nobody explained anything and everyone was tested the same way. The group that expected to explain remembered the material better and in a more organised way.

So the mere framing "I'm going to teach this" changes how you read: you automatically look for structure, pick out what's central, prepare examples — everything we described as assembling a chunk.

Logan Fiorella and Richard Mayer compared several variants: simply studying, preparing to explain, and actually explaining. The best result — especially on a delayed test — came from actually explaining. Preparation helps, but saying it out loud helps more.

That's exactly why tutors and teachers admit they only really understood their subject once they began teaching it. That isn't modesty, it's an accurate description of the mechanics.

You can explain to yourself too

You don't always need a partner. There's a well-studied technique: self-explanation.

In the late 1980s Micheline Chi and colleagues observed how students worked through worked examples in physics. A difference emerged: strong students constantly said to themselves things that weren't in the text — "why this step here?", "what would be different otherwise?", "this follows from that." Weak students simply read. In later work Chi's group began prompting the technique: people were asked, after each sentence, to explain to themselves what it meant and how it connected to the previous one. Understanding of the material improved noticeably.

So "explain it to an imaginary beginner" isn't a metaphor but a working protocol. The one requirement is the same: out loud or on paper, not silently.

A related technique that reviews rate well is asking yourself "why is this so?" of every claim in the material. Not "what does it say," but "why this way and not another." Such questions force you to connect the new to what you already know — that is, to create those same labels from the article on the warehouse.

Discussion beats listening

There's a result about group learning that looks almost impossible.

Eric Mazur, a physics lecturer, discovered something unpleasant many years ago: his students passed problem sets successfully but failed conceptual questions that revealed real understanding. In response he restructured his classes: students are asked a question, answer individually, then discuss the answer with a neighbour, and answer again. Scores on conceptual tests rose substantially.

And Michelle Smith and colleagues (2009) tested what exactly works in that discussion. They selected cases where neither of the discussing pair had initially chosen the correct answer. After discussion the share of correct answers in such pairs rose anyway. So the effect doesn't reduce to "the clever one told the weaker one": the very need to formulate a position, hear an objection, and defend or surrender your version advances understanding — even when nobody knew the right answer.

It's also the most honest form of retrieval you could invent. In an argument you can't sneak a look at your notes.

When someone explains to you

So far this has been about what explaining does for the explainer. But there is a second role — the student with a teacher — and it gives you something working alone cannot give at all.

The experience is much the same for most people: you spend months working it out yourself, then take a few lessons with a teacher and find you got further in two weeks than in the two months before. Part of that is the accountability discussed below: an appointed time and a person expecting you. Only part, though. The speed-up has three separate mechanisms.

First: you're held at a difficulty you can't reach on your own. Lev Vygotsky described the gap between what a person does alone and what they do with support from someone more experienced: the second runs ahead of the first and in time becomes the first. A good teacher doesn't hand over ready answers — they take on the parts of the task you can't yet hold, and release them as you grow (Jerome Bruner and colleagues later called this "scaffolding"). You work at a difficulty you'd have taken months to reach alone — that is, in exactly the channel the article on flow was about.

Second: an external instrument instead of your own feeling. One idea runs through this whole sphere: the feeling of "I get it" is poorly connected to reality. A teacher is an external measurement, and an immediate one: they see the mistake at the moment you make it, not two weeks later at a test. The self-taught learn about their holes late and expensively.

Third: saving on choosing the route. The most underrated one. In self-directed learning a huge share of the time goes not into learning but into decisions around it: what next, in what order, is this topic enough yet, which of two contradictory sources is right. A teacher removes those costs entirely. And they see your own mistakes, not the average student's mistakes described in the textbook.

How large the effect is is still argued over. In 1984 Benjamin Bloom formulated what he called the "two sigma problem": on his data, students taught one-to-one by a tutor came out about two standard deviations above an ordinary class — that is, the average one of them beat some 98% of those taught in a class. The number has been disputed many times and is hardly reproducible in pure form, but the direction holds up steadily: individual work with a teacher is one of the strongest educational interventions known.

And the caveat, without which this becomes a waste of money. With a good teacher everything sounds clear and logical — and a clear explanation feels like material learned. That's the same recognition instead of retrieval from the third article, only now in a very convincing voice. A lesson you spent listening and nodding through gave you about as much as rereading your notes.

The working format is the reverse: you do most of the talking and solving, the teacher corrects. The mark of a good lesson isn't "it was all explained to me clearly" but "I was made to do it myself, and I saw where I go wrong." If lessons run like a lecture, it's worth asking for a different arrangement — or bringing your own problems and working through them out loud in front of them.

The teacher is chosen on the same criterion: the one who can stay quiet and wait while you get there yourself is worth more than the one who explains impressively.

When people get in the way

In a group the same trap is there, only less visible.

Group study easily becomes the illusion of competence at a new level. You sit there, someone briskly solves the problem at the board, it all looks clear, you nod. That's recognition, not retrieval — exactly what we covered in the third article, only now with someone else's voice instead of a textbook. You can leave a group with the full sense of a productive evening and zero gain.

Plus the classic social loafing: in a group it's easy to ride on someone else's effort and to mistake the comfort of company for a result.

So a working study group needs rules:

Everyone arrives prepared — discussion doesn't replace working through the material, it follows it.
Everyone speaks in turn. Not "who wants to explain," but "this person explains this section now."
Solutions aren't shown, they're reconstructed. One person recounts the steps from memory, the others catch the gaps.
The question "why" is compulsory. You may not agree silently — you either restate it in your own words or object.
There's a finishing time. Otherwise the group reliably slides into conversation.

And one role of other people that shouldn't be underrated: accountability. A promise someone knows about is kept more often than a silent intention; a regular meeting creates an external rhythm independent of mood. It's the same principle as the deadlines in the article on motivation: a checklist on the outside works more reliably than resolve on the inside.

In practice

Read with the framing "I'm going to explain this." Even if you'll never explain it to anyone. That frame alone changes what you notice in the text.

After each section — out loud, with nothing to lean on. In your own words, in full sentences, as if to someone who doesn't know. Stumble into "well, it's sort of" — you've found a hole, go back to the material.

Ask "why" of every claim. Not only "how," but "why this way" and "what breaks if it's done differently."

Find one person, not a group of six. A pair works better: it's impossible to stay silent. Your partner doesn't need to know the subject — it's enough that they ask "but why?"

In a lesson, you do the talking. Bring your own problems and work through them out loud in front of the teacher. A lesson you sat through listening and nodding is money spent.

In a group, reconstruct rather than watch. The moment you start observing someone else's solution, switch to being the one who talks.

Argue on substance and don't defend a position out of stubbornness. The goal of an argument isn't victory but finding the weak spot. The strongest move in a study argument is to first state the other person's position so well that they agree with your wording.

Check yourself

Close the article and answer in your own words:

Why does knowledge in your head feel more coherent than it is, and what does an attempt to explain expose?
What did Rozenblit and Keil's experiment with the zip and the toilet show?
What happened in Nestojko's experiment to the group that expected to teach?
Why does discussion help even when neither person knew the correct answer?
What are the three mechanisms by which lessons with a teacher speed learning up?
Why can a lesson in which everything was explained to you clearly turn out to be useless?
How does group study turn into an illusion of competence, and which rules prevent it?

In short

In your head knowledge exists as a draft where a feeling of coherence replaces the connections. Explaining is translation, and it lays the blanks bare.
The illusion of explanatory depth: confidence in understanding a mechanism collapses immediately after trying to explain it step by step.
Merely expecting "I'm going to teach this" improves retention; actually explaining is stronger still.
Self-explanation works alone too: say why each step follows from the previous one, out loud or on paper.
Discussion with a peer advances understanding even without anyone knowledgeable in the pair — what works is the need to formulate and defend a version.
A teacher speeds you up for three reasons: they hold you at a difficulty above your own, they give immediate feedback instead of your unreliable feeling, and they remove the cost of choosing the route.
But a lesson you sat through listening and nodding gives you almost nothing: what works is the reverse format — you talk and solve, the teacher corrects.
Passively watching someone else's solution is the same illusion of competence. In a group you need to talk, not nod.
Knowledge that doesn't translate into human language is still a draft.