Skip to Main Content

Stop Misusing Autocorrect


There’s a right way and a wrong way to use your phone’s autocorrect, says the person who invented it. Ken Kocienda, former Principal Engineer of iPhone Software for Apple and author of the new book Creative Selection: Inside Apple’s Design Process During the Golden Age of Steve Jobs, led the team creating the first software keyboard for the first release of iOS. And he says that it’s possible to get too aggressive with autocorrect.

Autocorrect has evolved since its first form, but it’s still mostly based on the same fundamental factors. When you type a word, your keyboard’s software asks these three questions:

  1. What dictionary words, when typed “correctly,” would make similar tap patterns to the user’s? (Your taps make a “keyboard constellation,” which your phone compares to the ideal constellations for every dictionary word.)

  2. Which of these words are more common, and how much more common? (Kocienda points to Zipf’s Law, which predicts proportional word frequency in any language.)

  3. Which of these words show up in the user’s “dynamic dictionary”—their most frequently used words, or the names of their contacts, or words they’ve retyped even after autocorrect changed them?

For each word, the software has to choose how much to weigh these three factors to decide what you meant. To understand how difficult this choice is, and how much is at stake, remember how long it took to teach your phone the word “fuck.”

But this whole process works best if you don’t stop to fix every word. It’s much more efficient to type a whole sentence or message, then proofread at the end. That’s what the system was designed for, and what works best in Apple’s user testing.

And this will get even truer as autocorrect and predictive text become more sophisticated, says Kocienda. As keyboards examine longer passages, they can recognize more mistakes. For instance, he says, your phone might think you typed of, but realize several words later that you meant to type if. (It knows you didn’t mean “Of you like piña coladas, and getting caught in the rain...”) The more advanced autocorrect gets, the more it can fix things without you constantly watching it. So learn to trust autocorrect. (Or turn it off.)

And don’t constantly rely on predictive text. I use this bad habit too often, checking the suggested words above the keyboard, seeing if I can tap the finished word to save a couple of seconds here and there. Kocienda calls this context switching—you’re typing, then you’re reading, then you’re looking, and every time you switch between those tasks, you lose some time and attention. In Apple’s tests, most users (not just advanced phone typists) typed much faster if they just completed their own words. Predictive text is best ignored outside of unusual words or situations.

This kind of revelation reminds me how much I miss the hardware keyboards on the BlackBerry and the Sidekick. But as Kocienda points out, because software keyboards can disappear and make room for twice the display space, they’re necessary for full-screen phones. In his view, we couldn’t get the full-featured apps we have now while lugging around a hardware keyboard. So we have to give up the simplicity of clicking a real physical button—which wasn’t actually all that easy.

What separates Apple from competitors like Google, Kocienda tells Lifehacker, is that Apple always aims to make the single best choice for most users, instead of handing the choice off to users, which can overwhelm busy people who just want to send a text. That means that some features and options come out later on iPhones than on Androids, but once they show up, they’re usually thoughtfully designed. (Usually.)

Kocienda says this is another reason that tech needs diversity: Devices like the iPhone are meant for everyone, and you can’t design for everyone if everyone isn’t represented on your design and engineering teams. You can’t design a software keyboard without considering the sizes of the hands and fingers that use it. And if you’re not anticipating problems until you start user testing, you can’t come up with the best solutions.