The Door That Argues With You

You have met this door. It is glass, it is handsome, and it has a vertical metal bar running down its centre on the side you are approaching. A bar is a thing you pull. You pull it. It does not move. You pull harder, glance around to see who noticed, and then push, and it opens smoothly, because it was always a push door, and someone put a handle on it.

You will probably conclude that you were being careless. Almost everybody does.

You were not. You were reading the object correctly and the object was lying. A flat plate says push because there is nothing to grip. A bar says pull because a hand closes around it. These are not conventions you learned; they are properties of your hands meeting properties of the world, and a designer who overrides them with a symmetrical bar on both sides has produced an object that requires a label to explain its own operation. The presence of a sign reading PUSH on a door is not helpful signage. It is a defect report, printed and mounted at eye level.

This is the founding observation of a whole discipline, and the reason it is worth an essay is that we are now surrounded by objects with no physical form at all, whose designers have vastly more freedom and correspondingly fewer excuses.

Four things an object owes you

Start with affordance: what the thing physically permits. A chair affords sitting. A slot affords insertion. Affordance is a relationship between an object and a body, not a decorative choice, and it exists whether or not anyone designed it.

Then signifier: the visible clue announcing where the affordance is. Affordance without signifier is invisible capability — the elegant handle-less cabinet that nobody can open, the hidden gesture in an application that ninety per cent of users never discover, the flat rectangle that might be a heading or might be the button you need. Designers of the last decade removed shadows, borders and colour contrast in pursuit of cleanliness, and in doing so removed the signifiers. A great deal of contemporary interface design consists of making things beautiful by making them mute.

Then feedback: confirmation that your action registered. The lift button that lights. The click that a silent touchscreen has to counterfeit. Remove feedback and you produce the most reliable behaviour in the whole field: the user presses again. The pedestrian crossing button, pressed eleven times. The order submitted four times because the page showed nothing for six seconds. Nobody who does this is stupid. They are doing precisely what an unresponsive system instructs them to do.

Finally mapping: the correspondence between controls and effects. Four burners on a hob, four knobs in a row along the front. There is no physical reason the arrangement should be arbitrary, and yet after a century of manufacture, a large fraction of cookers still require you to memorise which knob is which — a fact you rediscover annually in an unfamiliar kitchen, generally while something burns.

The lie of user error

Which brings us to the phrase that has done more damage than any other in this field.

When an operator makes a mistake, the investigation concludes human error, the operator is retrained or dismissed, and the system is left unchanged. This is satisfying, cheap, and nearly always the wrong diagnosis. If one person in ten thousand mishandles a control, that is an individual. If a quarter of trained professionals make the same mistake in the same place, that is not a population of careless people. That is a control that is difficult to use correctly, and the error rate is a property of the design.

The classic industrial cases are studied precisely because the operators were not incompetent. Control rooms have been found in which an indicator showed that a command had been sent rather than that the valve had actually moved — feedback that confirms the wrong thing. Hundreds of alarms have been found to activate simultaneously during an incident, with no prioritisation, producing a wall of noise in which the important signal is statistically undiscoverable. Identical switches performing opposite functions have been distinguished only by a small printed label. In each case the ensuing report named operator error, and in each case the honest finding was that the room had been designed by people who had never had to read it under pressure.

Software inherits all of it. The most persistent example is the mode error: the same input producing different results depending on an invisible state. You type a sentence into an application that has silently returned to command mode, and your keystrokes are interpreted as instructions. You have not made a mistake; you have been given an interface whose current meaning is not displayed.

When the design works against you deliberately

Everything above assumes incompetence. Some of it is not.

There is now a documented catalogue of patterns engineered to exploit the very tendencies good design respects. The subscription that takes two clicks to start and a telephone call during business hours to end. The consent dialogue in which "accept everything" is a large coloured button and "reject" is grey text of lower contrast, placed one screen further away. The confirmation worded to make declining feel like an admission — No thanks, I prefer to pay full price.

These are not failures of craft. They are craft, deployed against the user, by people who know exactly how attention and friction work, and they are the reason regulators have begun writing interface requirements into consumer law — a development that strikes designers as an intrusion and is, more accurately, a bill arriving.

What good design actually feels like

It feels like nothing.

That is the difficulty of the whole discipline and the reason it is chronically underfunded. Excellent design is invisible: nobody notices the door they opened correctly on the first attempt. The evidence of success is an absence — of hesitation, of support calls, of the small daily humiliations that accumulate into a general belief among ordinary people that they are bad with technology.

They are not bad with technology. They have been handed several thousand badly labelled doors and told the problem is them.