Lately I’ve been noticing that one of my email services has a mind of its own. Not a great mind, mind you. But an officious little mind that thinks it knows what I want to say better than I do. When I open an incoming message in this program, there are often three little oblongs lying at the bottom, like a selection of tranquilizer capsules, each containing a brief bland word or set of words I might use to answer the message.
A student revising an essay writes to ask whether a sentence is grammatically correct. The little capsules offer me the choice of answering “You are correct,” or “Correct,” or “No, it’s not.” All of which sound robotic, curt, and indifferent.
The choices are inane. “Thank you!” the program suggests I could say in answer to someone’s note, apparently assuming I am incapable of coming up with this phrase on my own. Or, if “Thank you!” doesn’t seem like quite the right wording, I could hit the capsule that says “Thank you so much” or the one that says “Many thanks.” Sometimes the syntax is stiff and weird, like something a butler would say in a bad movie. “All the best with your writing endeavors.”
By this point you may be saying, “Duh.” This suggested-reply feature — called, in a classic example of Orwellian language, “Smart Reply” — has been around for years. I guess I’ve ignored it, the way you can ignore a fly buzzing around the room, until suddenly you find the buzzing intolerably loud and distracting.
I recently became curious about how this program would respond to different kinds of messages, and if there might be ones whose tone or subject matter would rattle it. So I sent myself a series of emails. The first one was neutral, asking for advice about windsurfing — could I help the sender find a teacher and equipment? “Yes, I can help you,” the program suggested I might answer. Or “No, sorry.” Or “Where do you live?”
What if the message were a little more personal and breezy? “Hey, Joan,” I wrote. “I miss you. Any plans to visit Tulsa?” The program was noncommittal (“No plans yet”) but glibly affectionate (“Miss you too”). By this point I felt as if I were engaged in a battle of wits with a Magic 8 ball.
I started to bait the program. Could it handle strong emotion? “I am still very angry and sad about what you did,” I wrote, “and don’t know if I can get over it.” The program offered no suggestions on how to reply to that one. Nor did it want to involve itself in anything to do with politics. It also wisely stayed silent when I asked myself to marry me, but when I wrote, “Shall I compare thee to a summer’s day?” it offered the options of “Yes, please do” and “Not really.”
This ridiculous feature is something you can ignore — as I did for years — or turn off. But now that it has started to really bug me, I’m finding I don’t want to turn it off. It keeps me awake. It carries a whiff of the much more subtle and frightening ways that AI is now trying to imitate how we write. In its cloddish robotic way, Smart Reply is a reminder of the importance of language, and of precise, individual expression in language.
The program’s voice is presumptuous, ignorant, and lazy. I am not sure how an inanimate thing is able to sound contemptuous, but this one does. It doesn’t care how the writer thinks or what the recipient feels. It blows people off with a kind of staggeringly bland algorithmic sang-froid.
It’s a potent little example of how all our devices, with their assistive time-saving shortcuts, are relieving us of the need to actually think and communicate, and making us more distant from one another, more careless, more apathetic, more callous, more panicky and distracted, and — both in our private lives and in our public life — a lot more dangerously hasty.
Joan Wickersham’s latest book is “No Ship Sets Out To Be A Shipwreck.” She is a contributing Globe Opinion writer.