You Have a Picture of How the Thermostat Works. It Is Probably Wrong

A person interacting with anything complicated builds a simplified internal representation of how it behaves, and uses that representation to predict what will happen, to decide what to do and to explain what went wrong. The representation is not a copy of the system and is usually much simpler, frequently borrowed by analogy from something already familiar. It does not need to be correct to be useful, and a wrong one that produces the right actions in ordinary circumstances survives indefinitely because nothing contradicts it.

Several incorrect pictures are held very widely:

• •That a thermostat set higher makes a room heat faster
• •That a lift button pressed repeatedly arrives sooner
• •That closing unused programs meaningfully speeds a computer
• •That a plane door could be opened in flight
• •That heavier objects fall faster
• •Each produces sensible-seeming actions and is simply untrue

The heating example is worth taking apart because it shows how such a picture forms. A thermostat is a switch that turns heating on until a temperature is reached and then off, so the setting controls the destination and not the rate. Many people instead hold a picture borrowed from a tap or a throttle, in which turning it further increases flow, which is how most controls they encounter actually behave. The borrowed picture produces a harmless error most of the time and a wasteful one in a cold house, and nothing in the thermostat's appearance corrects it.

The analogies people reach for are not random and knowing the sources explains which errors recur. Previous devices supply most of them, so a new control is assumed to behave like the nearest familiar one. Physical intuition supplies others, including the belief that a heavier object falls faster, which survives despite being taught otherwise. Language supplies some, since calling something a memory invites assumptions about how it stores and loses things. And instructions supply a few, though written explanations are the weakest source, because a picture formed from use overrides one formed from reading almost every time.

The practical response is to design so that the picture people form is the correct one, which is a large part of what interface design is. A control should suggest what it does by its form, so a dial that sets a level looks different from a switch. The system should show its state, since a picture cannot be corrected without feedback. The behaviour should match the analogy the design invites, or the analogy should be avoided. Where the true mechanism is genuinely complicated, the interface should offer a simpler one that is consistent rather than leaving people to invent their own.

People run machines and systems from a simplified internal picture, usually borrowed by analogy, which survives while it produces workable actions regardless of whether it is true. The thermostat treated as a throttle rather than a switch is the standard example. Design that makes the correct picture the obvious one, and that shows the system's state, is what prevents the invented alternatives.