Most of us have wrestled with the correction of “fucking” to “ducking”. This goes further.
iOS’s autocorrect engine subtly prompts me to dumb down my language. It “corrects” real words to other words. It has corrected “caving” to “raving”. It has corrected “inefficacy” to “in efficacy”. From 1.5 years of experience, my hypothesis is that it automatically substitutes in common words for more obscure ones—even though neither of the above examples is uncommon at all, and with no regard for context.
Worse, it clearly does have some idea of context, because it reaches back into previous words to reshape my phrasing, which is often an aesthetic choice. But sure, I sometimes think, maybe it’s more understandable that way.
This morning, I typed an expressive “saaad”, and I was not allowed to move the cursor past the word, to keep writing, until I’d clicked on one of the suggestions (the first of which was my word as typed).
I went looking for anyone talking about this, in case it’s already a dead horse. All I found was controversy about blind spots in the engine’s lexicon — a different type of political correction altogether.
What is easiest?
It’s not brute enforcement that gets us, in the end. It’s these gentle social pushes, the helpful “surely you meant…?”, the assumption that we wouldn’t be writing this way. It subtly tells us that complex or expressive writing is wrong. It treats us like children. It makes me feel intellectually restrained, as so many nerds have felt throughout their lives; and now, we come full circle as other nerds push us to simplify what we say even to ourselves—as mobile devices become, effectively, prosthetic extensions of the human brain. (See: people describing their devices as “my external brain” or “my exocortex”.)
This type of automatic, imposed “correction” makes us change without knowing, adopting almost unconsciously (as we so amazingly, humanly do) behaviors that will make our lives easier. Behaviors that make us less likely to encounter obstructions, to have to repeat ourselves, to have to monitor the machine lest we write “raving to pressure”. We adapt like this every day.
Which way is the “easy” way? You shape society by changing the answer to this question. It’s what makes people lean toward “normal” identities, conforming a little to the roles we see laid out for us via others’ assumptions. This is why it’s so important to consciously design search algorithms, because no ranking choice is neutral. And poor UX keeps people from adopting better security practices, in the storied war between privacy and convenience. But we humans want to be better. We want to be smarter. And the power of technology should support that.
But… more complex = less predictable, less marketable. So the impetus isn’t there for companies who need to sell iPhones. And for writers, too, a scripted style can be a conscious decision: repeating keywords in different combinations gets you more hits.
Verbal skills are contagious; they are improved by reading complex material with varied sentence structures and vocabulary. iOS’s autocorrect engine subtly makes my writing stupider. I wonder whose material I’ve read that is stunted by this subliminal urge to appease the digital schoolmarm. I wonder what color, what metaphor, what depth I’ve missed, because someone didn’t want to Google “inefficacy” to make sure they were using a real word — and instead replaced it with something more common. (Here is a fascinating study on Agatha Christie’s verbal decline as she possibly developed Alzheimer’s.)
Will my work only be acceptable to this machine when it could be Markov-chain generated from the rest of human writings?
We were doing this already… but technology accelerates it
Language inevitably changes over time, and a part of that is the elision of uncommon structures. The most common structures — such as English’s weird “I am, you are, let us be”—are allowed to stay weird, because we’re immersed in them every day. Some weirdnesses survive only as part of common phrases: in my favorite example, the dying past-participle -n suffix, “mown” is becoming “mowed” except in “new-mown hay”; the word “sawn” is mostly only used in “sawn-off shotgun” (and even that is being replaced by “sawed-off”). Meanwhile, “grown” and “seen” live on. I think the iOS autocorrect behavior could accelerate this process.
Simultaneously, we gather new words and forms of speech. Thus, we remain creatures of beautiful expressiveness.
Please, if you design systems, allow complexity to persist.
Do it the Google way: “Did you mean…”, as a suggestion, with options presented. One click away: due penance for our clumsy human fingers.
Do it the Android way: a row of potential corrections above the word.
Don’t tell me what I meant to say. Don’t rewire my brain for me.
This is part of my rant-mythology for complexity, beauty, and a life well lived.