LLOS.ai
How Things Work
● How Things Work · Module 5 · Constraints

Constraints

You’ve flipped the USB three times and it still won’t go in. The plug isn’t stupid — the one that only fits one way is just kinder.

You can signal a control, map it, give feedback and hand over the right model — and a tired, distracted human will still reach for the wrong thing. Modules 1 to 4 help people recover from mistakes. This one is about stopping them at the source: designing so the wrong action is hard, awkward, or flat-out impossible.

A constraint deliberately limits what you can do, to steer you to the right move — a plug that only fits one way, a microwave that won’t run with its door open, a button greyed out until it’s safe. The strongest are forcing functions: the system simply refuses to let you fail. You’ll feel a keyed plug, an interlock, and the gap between a design that warns you and one that prevents you.

~8 min3 felt labswarn vs prevent10-question bankglossary · FAQ
The journey — how things work
signalsaffordance
mappingmatch the world
feedbackdid it work?
modelsthe story
constraintsno wrong moves
errorforgive
Same promise, step five: when a design confuses you, it’s the design’s fault — not yours.
What you’ll be able to do
01 · Make it impossible

The plug that only fits one way

The most reliable way to stop a mistake isn’t a label, a warning, or a well-trained user. It’s a shape — a plug keyed so it physically cannot go in wrong, a SIM tray with one clipped corner, a three-pin plug that only seats one way. The design does the remembering, so you don’t have to.

Try both. First an old symmetric plug you can jam in backwards, then a keyed one that refuses every wrong angle. Feel the difference between being careful and being unable to fail.

The keyed plug
Get the plug into the socket. Rotate it, then insert.
▶ Start on Unkeyed and fumble. Then Keyed — you can’t get it wrong if you try.
Plug and socket
forced wrong 0
02 · The forcing function

The microwave that won’t run with the door open

A physical constraint stops the wrong fit. A forcing function stops the wrong order: it won’t let step B happen until step A is safe. A microwave that refuses to run with its door open is the classic — an interlock. You can’t irradiate yourself even if you try, because the unsafe sequence is simply not available.

Heat your food twice. First on a machine with no interlock — leave the door open and hit Start. Then on one with the interlock, and feel it refuse.

The interlock
Open the door, then press Start — and see what each machine allows.
▶ Start on No interlock and get careless. Then Interlock — the unsafe move becomes impossible.
Door: closed
unsafe starts 0
03 · The strongest guard

Warn, or prevent?

When you can’t design the mistake out entirely, you have two options. Warn — pop up “Are you sure?” — or prevent — make the dangerous action genuinely hard to do by accident. They feel worlds apart. A warning is a speed bump you learn to fly over; prevention is a wall.

Delete everything, twice. First behind a polite dialog you’ll click straight through. Then behind a guard you have to mean to open.

Delete everything
Wipe the data. Notice how hard — or easy — each design makes it.
▶ Start on Warn and click through. Then Prevent — try to do it on autopilot.

This permanently deletes everything. Are you sure?

The craft of the closed door

Make the mistake impossible, not just discouraged

Constraints sound like restrictions until a keyed plug or a door interlock quietly saves you from yourself. Named clearly, they are the strongest tool in the kit — the one that stops the error instead of cleaning it up.

Design the wrong move out of existence.

There are four ways to limit what a person can do. Physical constraints use shape: a plug that only fits one way, a SIM with a clipped corner. Logical constraints use reasoning: one screw left, one hole left, so it must go there. Cultural constraints use shared meaning: red means stop, a padlock means locked. And the strongest, the forcing function, uses sequence: it will not let step B happen until step A is safe.

Forcing functions come in three flavours. An interlock forces an order (the microwave won’t run with the door open; a car won’t start out of park). A lock-in keeps something going so you can’t stop it by accident (“you have unsaved changes — really leave?”). A lock-out keeps you out of a dangerous place (a bar on the basement door so no one flees a fire the wrong way).

And the golden rule when you can’t design the error out completely: prefer prevention over warning. A warning is a speed bump a tired brain flies over; a good constraint is a wall. So the working test is blunt: could a distracted, exhausted person still make this mistake? The closer the answer to “no, not even if they tried,” the better the design — and the fewer apologies it will ever have to make.

Cast your vote

Which design prevents the error?

Five face-offs. Same risk, two designs — one merely asks you to be careful, one makes the mistake hard to make. Pick the safer, then see why.

Round 1 of 5Score 0
Seen in the wild

A constraint that saves you daily

The best constraints are invisible — you only notice them the day one is missing. Here are two that do their work in silence.

SIM tray
One clipped corner

A SIM card’s cut corner is a physical constraint: it drops into the tray one way and jams every other. You cannot insert it wrong even in the dark — the shape remembers for you.

Fuel nozzle
The wrong nozzle won’t fit

A diesel nozzle is too fat for a petrol filler neck. It’s a deliberate physical constraint that quietly prevents a very expensive mistake, with no sign to read and no training required.

Where constraints lead next

Constraints stop most mistakes before they start. But no design catches everything — a slip will always slither through. So the final move isn’t prevention, it’s grace: when the error does happen, catch it gently and let the person undo it. That is the last bridge: error — designing for the mistake you couldn’t prevent.

Take it further

The real references

Constraints, forcing functions and poka-yoke come straight from Norman and the Toyota factory floor. Four worth your time. (Opens in a new tab.)

The vocabulary

The Constraints glossary

Six words that turn “how do we stop this mistake?” into a precise choice. Tap a card to flip it.

The nuance

Questions you actually ask

Busting the myths

What people get wrong

Myth
“An ‘Are you sure?’ dialog protects users.”
People habituate: by the tenth time they click through it without reading, on autopilot. A warning you always confirm is no warning at all — prevent the action, or make confirming it genuinely deliberate.
Myth
“Constraints frustrate power users.”
A good constraint is invisible to correct use — the keyed plug never slows the person plugging it in right. It only blocks the wrong move. Bad constraints nag; good ones just quietly remove the cliff.
Myth
“If they ignored the warning, it’s their fault.”
Design for the tired, distracted human, because that’s who shows up. ‘User error’ is almost always a design that left the error possible. Close the door instead of blaming the person who walked through it.
Myth
“More confirmations mean more safety.”
Pile on dialogs and you breed confirmation fatigue — each one means less. One well-placed forcing function beats five ‘are you sure?’s, and doesn’t train people to click blindly.
Train the eye

Read the constraint

Ten questions in seven formats, tagged by thinking level — each with layered hints: a nudge, the reasoning, then a deeper connection. A wrong answer opens a door, never a dead end. Served from the question bank, never hardcoded.

Loading the question bank…
Question 1 of 10
Multiple choice
Keep going

Where this connects

▶ Play the labs