What a Mental Model Is

Look at a subway map for a second. The lines are clean, straight, brightly coloured. Stations sit at tidy intervals. It’s beautiful — and it’s a lie. The real tunnels curve, the distances between stops vary wildly, and the geography is nothing like those neat parallel lines.

And yet the subway map is one of the most useful objects ever designed, precisely because it lies. It throws away everything you don’t need (exact distances, real curves, what’s above ground) and keeps the one thing you do need: which stop connects to which. That trade — drop detail, keep the useful structure — is exactly what a mental model is.

The definition

Here’s the precise version, and it’s worth reading slowly:

Three words in there are doing the heavy lifting:

• Simplified. A model is not the real thing. It’s a stripped-down version. That’s a feature, not a bug — the simplification is what makes it fast to use.
• Representation. It stands for something. A map represents a city; the model of “supply and demand” represents how prices move.
• Carry to new situations. A good model isn’t stuck to where you learned it. Learn it about subways and it works on circuit diagrams, org charts, and family trees.

The word “compressed” is the one to hold onto. A model takes something huge and complicated and squeezes it into a picture small enough to fit in your head and fast enough to use under pressure.

A worked example: a thermostat

Let’s build one model from scratch so you can watch the compression happen.

A house’s heating system is genuinely complicated: a boiler, pipes, a thermostat, the outside weather, how well the walls hold heat, where the windows leak. If you tried to hold all of that in mind at once, you’d never make a decision.

So instead you use a tiny model: “It’s a loop. If the room is colder than the target, the heat comes on; once it’s warm enough, the heat goes off.” That’s it. That sentence is the model.

Notice what it threw away — the boiler’s inner workings, the exact physics of heat loss, the gas pressure. And notice what it kept — the one relationship that lets you actually predict and control the thing: cold → heat on, warm → heat off. With that model you can answer real questions (“why won’t the house warm up if I crank the dial to 30?”) without knowing a single thing about plumbing.

Map, model, territory — sorting the three

People mix up three related things constantly. Let’s separate them cleanly, because the whole rest of the course leans on the distinction.

A fact is a single data point. A model is a machine for generating answers. You can memorise the fact “the blue line has 14 stops,” but the model “subway lines connect stations in sequence” lets you read any transit map you’ve never seen.

The pitfall: mistaking the model for reality

Here’s where beginners trip. Because a good model is so clean and useful, it’s tempting to forget it’s a simplification at all — to start believing the subway map is the city. The whole next lesson is devoted to this trap, but plant the flag now: a model is always less than the thing it describes. The moment you forget that, the model stops helping and starts misleading.

When to reach for a model

So when do you actually deploy one? Whenever a situation is too complex to think about in full — which is most of the time.

• Use a model when you need to decide or predict fast and can’t afford to understand every detail. The thermostat loop lets you act without a plumbing degree.
• Use one to transfer understanding from a domain you know to one you don’t. Recognise “this new thing is basically a feedback loop” and you import everything you already know about loops.
• Be more careful when the stakes are high and the situation sits near the model’s edge — where the dropped detail might be the detail that matters. That caution is the subject of the next lesson.

Recap

You now know what a model is. Next: the single most important thing to remember about every model you’ll ever use — that it is not the thing itself.