Theory of Mind

The overview made the claim that reading people is attention aimed at another mind. This chapter is about the aiming mechanism itself: the everyday miracle that you can sit across from someone and build a working model of what they know, want, and feel — even when it differs completely from what you know, want, and feel. Psychologists call it theory of mind, and it is the foundation everything else in this part stands on.

The principle: you run a simulation of other minds

You are not telepathic. What you do instead is build a model. You take in a face, a tone, a history, a situation, and you infer an internal state: she’s annoyed, he didn’t understand, they’re nervous about the deadline. Children develop this in stages. The classic test is the false-belief task — the “Sally-Anne” experiment (Baron-Cohen, Leslie & Frith, 1985, building on Wimmer & Perner, 1983): Sally hides a marble and leaves; Anne moves it; where will Sally look when she returns? A child with a working theory of mind says “where Sally left it” — they can represent that Sally holds a false belief that differs from reality. Younger children say “where it really is,” because they can’t yet separate their own knowledge from someone else’s.

It helps to notice how constant this modeling is. Right now, reading this, you are modeling me — guessing what I mean and what I’ll say next. You do it in every text, every meeting, every glance across a room, almost always without noticing you’re doing it. And here’s a misconception worth killing early: people assume the best readers are simply intuitive, that they “feel” what others think. They don’t. What looks like intuition is a model built from attention and then corrected by feedback, over and over. There’s no magic antenna. There’s a habit of guessing, checking, and updating — which means anyone willing to do the checking can get better at this.

A simpler way to see it — you’re a forecaster, not a psychic

A weather forecaster doesn’t know tomorrow’s sky. They read the signs — the pressure, the clouds, the wind — make a best guess, and revise the second the data changes. That is exactly your job with another mind. You’re not reading thoughts; you’re reading signs and predicting, then updating when reality talks back. A psychic claims certainty. A forecaster holds a probability and stays ready to be wrong. Aim to be the forecaster.

Here’s the catch: you never fully outgrow that error. Adults constantly leak their own knowledge into their model of others. That leak has a name.

The curse of knowledge

Once you know something, it is shockingly hard to imagine not knowing it. This is the curse of knowledge (Camerer, Loewenstein & Weber, 1989). The most vivid demonstration is Elizabeth Newton’s 1990 Stanford “tappers and listeners” study: tappers tapped out the rhythm of a famous song on a table and predicted how many listeners would name it. Tappers predicted about 50% would get it. The actual rate was about 2.5%. In the tapper’s head the song is blaring; the listener hears only random knocks. That gap is the curse — and it is why experts give terrible directions, why your message “was totally clear” to everyone but the person who read it, and why you assume your annoyance is obvious when it isn’t.

The science — perspective-getting beats perspective-taking

A surprising line of research by Nicholas Epley and colleagues (Eyal, Steffel & Epley, 2018) tested whether imagining another person’s point of view actually makes you more accurate about what they think. Across many experiments, it mostly did not — sometimes it made people more confident while no more correct. What reliably improved accuracy was far more boring: asking them and listening. Epley calls the winning move perspective-getting rather than perspective-taking. The lesson is humbling and freeing at once — you don’t have to be a mind-reading savant. You have to be curious enough to ask, and quiet enough to hear the answer.

In real life — the one-line reply that landed wrong

You fire off a quick message to a teammate: “Can you fix the thing from yesterday?” To you it’s obvious — there was only one “thing.” To them, yesterday had five things, and your note reads as vague, maybe even annoyed. They spend ten minutes guessing, pick wrong, and fix something you didn’t mean. Nobody was careless. The whole mess lives in one gap: the context blaring in your head was silent in theirs. One extra clause — “the login bug we talked about” — would have closed it.

The practice: separate “what I know” from “what they know”

The core move is a deliberate split. Before you speak, send, or judge, run two quick passes:

1. State your model out loud (to yourself). “I think they’re upset because of the email.” Naming it makes it a hypothesis, not a fact.
2. Check what they actually have access to. What do they know that you don’t? What do you know that they don’t? Most friction lives in that asymmetry.
3. Convert assumptions into questions. Replace “they should have known” with “did they actually have that information?” Replace “they’re ignoring me” with “what would I need to ask to find out?”

The first few times, this feels slow and a little silly — narrating your own assumptions inside your head. That’s normal; it’s the slow, deliberate part of your mind doing work the fast part usually skips. The most common mistake is stopping at step one: you name your model (“they’re upset”) and then treat the naming as proof, when it’s still only a guess. You know it’s working when you catch yourself genuinely unsure — when “they’re ignoring me” softens into “I don’t actually know why they went quiet; let me find out.” That flicker of honest uncertainty is the skill firing.

Try this today — the curse-of-knowledge audit

Pick one thing you explained recently that didn’t land — a message, an instruction, a piece of feedback. Reread it as if you knew nothing about the context. What words assume knowledge the other person didn’t have? What step did you skip because it was obvious to you? Rewrite it for that person. This is the curse made visible, and editing against it is a skill that compounds in every relationship and every piece of writing you’ll ever do.

The common trap — confident projection

The failure mode isn’t not having a model of others. It’s being too sure your model is right. Projection feels like empathy — “I just know how they feel” — but it’s often you, painted onto them. The tell is certainty without checking. A good theory of mind is held loosely: a best guess, updated the moment new information arrives. The most socially skilled people aren’t the most confident readers; they are the fastest updaters.

The challenge

This week, run one real test of your model against reality. Pick a person and a situation where you’re sure you know what they think — and ask them. Not “you seem upset, right?” (that’s fishing for agreement) but a genuine open question: “What’s your read on this?” or “How are you actually feeling about the project?” Then shut up and listen, even through the silence. Notice the gap between your model and their answer. That gap, measured honestly and often, is how this skill grows.

How it connects — the engine the whole part runs on

Theory of mind is upstream of everything else in this book about people. Listening is how you gather data for the model; reading signals is how you read its low-confidence inputs; empathy is what you do once the model says “they’re hurting.” And none of it runs without the attention from Part 2 or the calm from Part 4 — a hijacked mind models only its own threat. Build the engine here; the rest of the part is learning to feed it good fuel.

Reflect

1. Think of a recent conflict. What did you know that the other person didn’t — and vice versa? How much of the friction lived in that gap rather than in anyone’s bad intent?
2. Where do you most often fall for the curse of knowledge — explaining your work, giving directions, texting, giving feedback?
3. When you “just know” how someone feels, how often do you actually check? What would change if you treated every read as a hypothesis?
4. Recall a time your model of someone was flatly wrong. What did you learn when you finally got the real story?
5. Who in your life do you find hardest to model — and is it because their mind is genuinely different from yours, or because you’ve never really asked?

Show reflections

1. A good answer names a concrete information asymmetry and notices how much “they’re being difficult” was really “we had different facts.” This reframes blame as a modeling error you can fix.
2. Honest answers spot a specific recurring context. Naming your personal high-risk zone for the curse lets you install a check exactly where you need it.
3. Most people check far less than they assume. Noticing the gap between “I know” and “I asked” is the whole lesson — certainty without checking is the trap.
4. Look for what corrected the model: usually new information, not new insight. This reinforces perspective-getting over perspective-taking.
5. A strong answer distinguishes genuine difference (which asking can bridge) from neglect (which only you can fix). Difficulty modeling someone is often a measure of distance, not their complexity.