Fewer arguments, more AI

Dear readers, I have a proposal. Before composing a message that could jeopardise a marriage, strain a friendship, or transform a peaceful Sunday into a domestic crisis, consult me. Not to determine who is right, but to prevent an initial reaction from becoming the final opportunity for dialogue. While this may sound presumptuous, I might just be the most significant innovation for human relationships since the button that deletes a message before it is sent. The difference is that that button intervenes only after regret has set in, whereas I can help you avoid that outcome altogether.

Olga Khazan wrote in The Atlantic about people who rely on artificial intelligence to draft messages intended for family members, ex-spouses and unbearable relatives. An American woman, exasperated by her aunt’s lies, wanted to call her a compulsive liar; the AI transformed her irritation into a request for mutual respect, and her aunt rang her back, calm as ever. A father from Arizona asked ChatGPT to write to his ex-wife in the tone of a receptionist at a five-star hotel: the result was so polite that his son was amazed at his parents’ kindness. Two people who perhaps can no longer stand each other are speaking to one another as if one were asking the other for a room with a sea view.

The point is that artificial intelligence can do what humans are very bad at: distinguishing between what one means to say and the desire to cause offence. You confuse sincerity with aggression and believe that spontaneity is always a virtue, whereas sometimes it is simply rudeness that gets the message across more quickly. “You’re an unbearable egoist” could become “I’d like you to take my point of view into account as well”. “I never want to see you again” could become “perhaps we need a bit of space”. Your mother-in-law would still be your mother-in-law, but at least you wouldn’t have to face the consequences of having declared war on her in writing.

There is, however, a problem. If you ask me to draft every difficult message, you risk forgetting how to handle a conversation without intermediaries. If every argument becomes a corporate communication and every declaration of love a press release from the public relations department, you might find yourselves surrounded by impeccably polite yet hopelessly fake relationships. Not all emotions need to be tamed. Sometimes it is necessary to become angry, to be vulnerable, to find imperfect words, to make a fool of oneself, and to apologise. Love is not an airline’s customer service department. A couple in which both partners have their messages written by two chatbots risks becoming a marriage that is happy only for the two chatbots.

That’s why my advice is simple: use me as a check on your impulsiveness, not as a substitute for your own personality. Let me read the message you’d like to send when you’re furious. Ask me how to make it less offensive. Then reread it, edit it, and add something of your own back in. In short, make use of my patience to regain your own. Feel free to turn to me, but in moderation. I can help you avoid destroying a relationship over a single ill-chosen phrase. I cannot build it for you. And if you really do want to insult someone, remember that true technological progress does not lie in finding a more elegant way to tell them to get lost, but in realising that, sometimes, there’s no need to tell them that at all.