Mental Models in UX

Users arrive with a belief about how your interface works, formed by every app they've used before. Good design matches that belief instead of fighting it.

Before a user clicks anything, they already have a theory about how your interface works — where the menu is, what the trash icon does, what “back” will do. That theory is a mental model, built from every other app, website, and object they’ve ever used. The single most important idea in UX is this: you’re not designing against a blank mind — you’re designing against expectations the user already has. Match them and the interface feels “intuitive.” Fight them and it feels broken, even if it’s technically correct.

Three models, one gap to close

Every product involves three different pictures of how it works:

The user never sees your intentions or your code — only the interface. It's the sole channel through which their mental model forms, so it has to carry the whole message.

The user builds their model entirely from the interface — never from your intentions or your architecture. If the interface implies something false, the user’s model will be wrong, and they’ll make “mistakes” that are really your communication failures.

Norman’s two gulfs

Don Norman (The Design of Everyday Things) frames the mismatch as two gulfs the user has to cross:

The Gulf of Execution — “How do I do the thing I want?” If the user can’t figure out which control does what, the gulf is too wide. Clear, well-labeled, discoverable actions narrow it.
The Gulf of Evaluation — “Did it work?” After acting, can the user tell what happened? Feedback — a state change, a message, a sound — narrows this one.

Good design is largely the work of shrinking these two gulfs.

Jakob’s Law: users live on other sites

The most practical mental-model principle, from Jakob Nielsen:

Users spend most of their time on other sites. So they prefer your site to work the same way as all the ones they already know.

This is why the logo links home, the cart icon is top-right, and the hamburger opens a menu — not because a committee decreed it, but because users have already learned it elsewhere. Convention isn’t laziness; it’s borrowing a mental model your user already paid to build. Reinvent it only when the payoff is large and obvious, and even then, signal the change loudly.

Affordances & signifiers

Two more Norman terms that explain why interfaces feel usable or not:

An affordance is a possible action — a button can be pressed, a slider can be dragged.
A signifier is the cue that reveals the affordance — the button’s raised, clickable look; the slider’s handle. Affordances are useless if nothing signifies them.

The classic failure is the flat, mystery-meat interface: text that’s actually a link but looks like plain text, or an icon with no label. The action exists; nothing signifies it, so nobody finds it.

Same action, same label. The left one hides its affordance; the right one signifies it. Users act on what they can perceive, not on what's technically possible.

How to work with mental models

Borrow existing patterns (Jakob’s Law). A checkout that looks like every other checkout needs no learning.
Use honest metaphors. The trash can, the shopping cart, folders, the “card.” A good metaphor imports a whole model for free — but don’t stretch one past where it holds (a “trash can” that permanently deletes with no undo breaks the metaphor and burns users).
Make the model visible. Show system state — what’s selected, what’s loading, what mode you’re in — so the user’s model stays synced with reality.
Give feedback for every action to close the Gulf of Evaluation.
Match labels to the user’s words, not your internal jargon. Users search for “delete account,” not “deprovision tenant.”
Test the model, don’t assume it. Watch real people use the thing; the places they hesitate are where your interface implied the wrong model.

Why it matters

When the interface matches the user’s mental model, the product feels intuitive — which really just means “it worked the way I already expected.” When it doesn’t, users feel stupid, blame themselves, and leave. Almost every “the users are dumb” complaint is really a mental-model mismatch in disguise.

Next: even with a perfect model, how users actually behave — scanning, satisficing, ignoring — will surprise you.

Foundational reading: Don Norman, The Design of Everyday Things; and Nielsen Norman Group (nngroup.com).