Lost in yesterday’s coverage of the Apple Watch was a small software feature that, when demonstrated on the stage of the Flint Center, earned brief but vigorous applause from the audience. It was the watch’s ability to scan incoming messages and suggest possible responses. The Verge’s live-blogging crew were wowed:
The example Apple presented was pretty rudimentary. The incoming message included the question “Are you going with Love Shack or Wild Thing?” To which the watch suggested three possible answers: Love Shack, Wild Thing, Not Sure. Big whoop. In terms of natural language processing, that’s like Watson with a lobotomy.
But it was just a taste of a much more sophisticated “predictive text” capability, called QuickType, that Apple has built into the latest version of its smartphone operating system. “iOS 8 predicts what you’ll say next,” explains the company. “No matter whom you’re saying it to.”
Now you can write entire sentences with a few taps. Because as you type, you’ll see choices of words or phrases you’d probably type next, based on your past conversations and writing style. iOS 8 takes into account the casual style you might use in messages and the more formal language you probably use in Mail. It also adjusts based on the person you’re communicating with, because your choice of words is likely more laid back with your spouse than with your boss.
Now, this may all turn out to be a clumsy parlor trick. If the system isn’t adept at mimicking a user’s writing style and matching it to the intended recipient — if it doesn’t nail both text and context — the predictive-text feature will rarely be used, except for purposes of making “stupid robot” jokes. But if the feature actually turns out to be “good enough” — or if our conversational expectations devolve to a point where the automated messages feel acceptable — then it will mark a breakthrough in the automation of communication and even thought. We’ll begin allowing our computers to speak for us.
Is that a development to be welcomed? It seems more than a little weird that Apple’s developers would get excited about an algorithm that will converse with your spouse on your behalf, channeling the “laid back” tone you deploy for conjugal chitchat. The programmers seem to assume that romantic partners are desperate to trade intimacy for efficiency. I suppose the next step is to get Frederick Winslow Taylor to stand beside the marriage bed with a stopwatch and a clipboard. “Three caresses would have been sufficient, ma’am.”
In The Glass Cage, I argue that we’ve embraced a wrong-headed and ultimately destructive approach to automating human activities, and in Apple’s let-the-software-do-the-talking feature we see a particularly disquieting manifestation of the reigning design ethic. Technical qualities are given precedence over human qualities, and human qualities come to be seen as dispensable.
When we allow ourselves to be guided by predictive algorithms, in acting, speaking, or thinking, we inevitably become more predictable ourselves, as Rochester Institute of Technology philosopher Evan Selinger pointed out in discussing the Apple system:
Predicting you is predicting a predictable you. Which is itself subtracting from your autonomy. And it’s encouraging you to be predictable, to be a facsimile of yourself. So it’s a prediction and a nudge at the same moment.
It’s a slippery slope, and it becomes more slippery with each nudge. Predicted responses begin to replace responses, simply because it’s a little more efficient to simulate a response —a thought, a sentence, a gesture — than to undertake the small amount of work necessary to have a response. And then that small amount of work begins to seem like a lot of work — like correcting your own typos rather than allowing the spellchecker to do it. And then, as original responses become rarer, the predictions become predictions based on earlier predictions. Where does the algorithm end and the self begin?
And if we assume that the people we’re exchanging messages with are also using the predictive-text program to formulate their responses . . . well, then things get really strange. Everything becomes a parlor trick.
Image: Thomas Edison’s talking doll.
I find communicating via mobile to be kind of a pain. But with two small kids and all the driving/pickups/carpools, I’m constantly texting other parents.
Better predictive texting would help — I’d be happy to get “I’ll pick M1 and M2 up after gymnastics” with just a couple keystrokes.
But it would also tend to reduce more serious communication to the creative and emotional level of emoticons. And there’s that whole privacy issue. Right now, I block as many trackers as I can while browsing — so am I going to let Google and Apple know that my daughters go to gymnastics, or that we regularly run out of milk, or…?
It’s like the technologies we’re creating are reducing us down to the binary … turning us into mindless drones.
Much ado about nothing. predictive text just makes typing faster/more efficient, because generally typing on tiny keyboards is a big trade-off between accuracy and speed. Android has had 3rd party apps such as swift key and swype that have been doing this for years, and they’re pretty great (and one of the few things i legitimately miss about the android OS).
basically, you can fire off a message along the lines of “I’m on my way” in ~ 5 key strokes instead of typing it all out as you currently have to do in iOs. Think of it more along the lines of an auto correct that you don’t spend twice the amount of time undo-ing its inappropriate corrections.
This is a brilliant commentary. Because of the AV disaster I missed the predictive text demo but I had a feeling it was coming and like you, I’m concerned with how it will affect things in the long run.
I’m old enough to remember turning “spaghetti” into “pasta” because on a typewriter, a spelling mistake was unfixable and who the heck could remember how to spell “spaghetti?”
Of course, language itself is a filter: maybe some day Apple will get it’s “haptic” or “taptic” shit together and really channel the brain sans-language.
How long before we replace the reply completely and the software automates it, puts “pick up the kids” on our calendar, drives our car to the daycare center, and texts the daycare worker to bring out the kids, all with no interaction from us? Not that “Her” was a good movie, but the notion that we can outsource our human interaction to computers (or to an automated service using humans, but not us), is incredibly disquieting.
The problem with technology that molds us to fit into it is a real one. Only the example with predictive text on mobile devices is not very accurate.
When I reply to SMS (or any other short message) I create an answer in my mind and then type it. Lets say that without predictive feature I would need 25 keystrokes to type my answer. So now: if my device predicts correctly I type the answer in 5 keystrokes, if it does not I type it in 25 keystrokes. But, it does not affect my answer, or my way of thinking.
Of course I might be wrong in a way, that if it does not affect me, it does not affect the others, especially that I consider myself “an Amish” compared to my technology loving friends. But to be honest short messages usually do not contain anything except crucial information. So there is not really any area to change our way of thinking in it. The real change was when we introduced instant short messaging – that was a point when we traded eloquence and openness of the language, for pure information and binary question/answering.
So as far as I agree that technology changes the way we think, and more often it flattens our thought process instead of enriching it, this time I do not think there is any damage. Unless we move predictive texting to a point where there will be no need for an input from a user to have proposition of an answer.