Anti-pattern

Andrew Koenig named the anti-pattern in 1995, and the name has two halves that both have to hold. It has to be a pattern: something people reach for over and over, because it looks obvious and arrives with a story attached about why it will work. And it has to be anti: it fails at the job it was reached for, often leaving the situation worse than leaving it alone would have. A bad idea nobody has twice is not an anti-pattern. It is a mistake. What earns the word is the recurrence, which is why every anti-pattern worth naming has the shape of received wisdom rather than the shape of an accident.

A dark pattern is not one of these, and the two words are worth keeping apart because they level different accusations. A dark pattern works. That is the whole reason it survives contact with a dashboard: the roach motel really does keep subscribers, the preselected box really does sell the insurance, and the cost is paid by the reader rather than by the business running the test. Calling something a dark pattern is a charge of malice against a design that succeeds. Calling something an anti-pattern is a charge of incompetence against a design that fails. The two do meet, but by circumstance rather than by definition: a deceptive flow becomes an anti-pattern as well on the day enforcement and refunds cost more than it earns, which is a claim about this year’s regulator and not about the pattern.

The specimen shows the one that quietly eats software: settling a design argument by adding a preference. Two people disagree about a default, nobody has to lose the argument, and the option ships. Every property of a decision is there except the deciding. Havoc Pennington wrote the case against it for GNOME in 2002 in Choosing our Preferences, and the accounting is what makes it an anti-pattern rather than a preference for fewer preferences: each option is cheap on its own, they multiply the states the software can be in, and the settings window was supposed to be the place where a person gets what they want. Enough of them and the switch you came for is below the fold. The fix was the cause.

The word does work in a review that “this is bad” cannot, because it names a goal and then asks for a measurement. An anti-pattern claim is falsifiable by construction: it says this was adopted to achieve something and does not achieve it, so anyone who disagrees has somewhere to go and look. That also makes it a claim to spend carefully. Most of the famous ones are contested, and the honest ones travel with a number and a date attached, because a pattern that stopped working is a different finding from a pattern you would not have chosen.

Koenig’s definition has a second half that design reviews usually drop. An anti-pattern is not fully described until someone writes down what to do instead, and the 1998 book that followed made that refactored solution part of the form. Naming the trap is the cheap half of the work.