Autocorrect is meant to help me, but I find it more confounding than useful, as it seems to have a mind of its own. Why does it add suffixes to words I have chosen, or correct spelling to change a word’s meaning from the one I intended? These rogue edits often occur without my being aware they have happened. I know I should check more carefully before sending messages into the ether world, but I often don’t want to become bogged down in vigilance when I am writing a quick note, or I simply am thinking of other things.
Some examples of unexpected “corrections.” Jon is turned into Jonathan; why does autocorrect assume there are no men named simply Jon? How did “where on earth” become “whee on earth?” Does autocorrect think I need more fun in my life? Then there is “mahatma” instead of “morning.” I have no idea where that one came from.
I texted my daughter to ask about the health of her cat, Belle, who had seemed lethargic for a few days. “Well how is Belleville doing?” I inquired. Autocorrect had changed the cat’s name to that of the town in which I had grown up. “I imagine it’s as sleepy as ever,” my daughter replied with a touch of sass. She thinks my hometown is boring.
She has her own autocorrect problems, however. In a text telling me about the spring garden she was preparing she related that she had planted boy chop. Turns out it was bok choy.
My favorite example derives from an email exchange I had with my friend Charlotte. Her father was not doing well and had just been moved to a higher level of care in his assisted-living facility. I asked how he was faring in his new environment. “Much better,” she replied, “now that he has a nude coming in twice a day.” Of course, she meant to say it was a nurse who came in twice a day. Autocorrect somehow filled out the rest of “nu” with its own preferred ending.
“Well, no wonder,” I responded. “That would perk any guy up.”
Autocorrect has been around since the early 1990s, developed by Microsoft to correct spelling errors in its Word program, using a dictionary for comparison. With the dawn of the smart phone — requiring fat fingers to type on a tiny screen — the spell-checker moved to other platforms and became more sophisticated, anticipating context and what a word should be. That’s when trouble truly blossomed. Goodness knows how many inadvertent changes it has made to texts and emails. In deciding it knew the word you were planning to type, it often misjudged and made a nonsensical substitution. Sometimes funny, sometimes not.
Autocorrect, now a generic term that also applies to the predictive iteration, runs on both a local device — phone, tablet — and the cloud, that mysterious location in the ether-sky that holds so much of our information. It uses language models plus AI to determine what should be typed. And sometimes it guesses context incorrectly, though the grammar may be spot-on. A famous example lies in the 1957 book by linguist Noam Chomsky, Syntactic Structures: “Colorless green ideas sleep furiously.” The sentence is grammatically correct but semantically absurd.
I suppose we can enjoy the goofy things autocorrect comes up with, as they can add levity to otherwise boring correspondence. But they also can lead to misunderstandings. For example, it once changed “I’ll bring napkins” to “I’ll bring weapons” and “I love you” to “I love tofu.”
The problem often arises with sentences that are short, making it difficult for an algorithm to discern a context. So, we humans presumably can do a better job of checking the meaning of a message — if we take the time to do it — and avoid inane predictions of the word to be typed if we use longer sentences.
Autocorrect requires care, just like most of life. I suppose it is an asset — but only in the right context. I am going to try to make vigilance my watchword, and I will try to write longer sentences. Maybe. Sometimes I’m just in a hurry. Right now I’m off to a party and bringing chips and guano. Oops, make that chips and guac.