Why Are People Using AI to Text Their Friends?
Moral decline in the era of AI speak
MØRNING. Somewhere between the em dash discourse and TikTok fruit stories, our ability to own our thoughts and speak for ourselves, with nothing but brains and intuition, has seemingly become an endangered skill. Essayist, critic and poet, Terry Nguyen (@nguyenterry) takes us on the downward spiral of AI colloquialisms and how they’ve wormed from essays into the most intimate conversations between friends, family and even lovers. What does this new phenomenon of AI speak mean for our sense of autonomy, or the fantasy of it? Terry dives in below.
Every year around the holidays, I receive a deluge of correspondence from K. Our long-distance friendship tends to wax and wane depending on the season. Texts and emails can go unreplied for weeks, if not months, though neither of us seem to take the silence personally. We’ve joked that we operate on a 19th-century timeframe. Communication delays have been essential to our decade-old dynamic.
In January, when I began catching up on K’s November emails, I had a feeling that something was off. The emails were compilations of various NotesApp entries—musings about whatever they had been reading, watching, or cooking with the occasional semi-critical screed. And yet the writing, despite the lack of capitalization, felt too polished at points, the difficult emotions too easily articulated and readily offered. Certain parts didn’t sound like K at all. It sounded a lot like ChatGPT. For a moment, I wondered if I was overreacting. Has getting an MFA made me insufferable and insane? Why was I hyper-fixating on their linguistic quirks? Was I being unnecessarily paranoid?
As it turns out, I’m not the only one driven insane by AI-induced paranoia. On Reddit, people have posted texts from spouses, friends, parents, distant acquaintances, and even Hinge matches that’ve aroused their suspicions. Most are seeking a second opinion or crowdsourced validation that the messages they’re receiving are written by—or with the assistance of—large language models (LLMs). “Is my partner using AI to respond to me? I just poured my heart out and feel like I got a chatbot response,” asks one user, who claims that their boyfriend “doesn’t usually sound this way when we’re talking about stuff like this.”
With ChatGPT reportedly reaching one billion monthly active users, people are using it for any number of everyday queries: answers to practical questions, assistance with virtual tasks, personal and professional advice, and a substitute for writing. “Before too long, essentially all writing might be A.I. writing,” cautions Sam Kriss in a NYT Magazine piece on the chatbot’s bizarre writing style. “Essentially every major email client now offers a similar service. Your rambling message can be instantly translated into fluent A.I.-ese.”
I’ve previously likened AI writing to a “rapid meaning-production scheme” that creates the illusion of thought (and thoughtfulness). LLMs can’t distinguish between writing an email to a client and writing one to a friend; an apology text to a partner and one to a long-time neighbor; a postcard to a close friend and one to a colleague. The models aren’t capable of code-switching because their output is based on statistically probable ways of speaking and thinking. This might be less jarring when encountered in the workplace. After all, a paragraph of “corporate-speak” already reads a lot like meaningless word-salad. It’s much more uncanny in intimate contexts where language should be a means of conveying attention and care.
Receiving an AI-written text from a friend or partner “feels like a betrayal,” to quote my friend M, as it demonstrates a refusal to communicate honestly. “I’d rather someone text me the ‘wrong’ thing and have an actual conversation about it than to get some sanitized machine-speak,” she said. With any personal relationship, whether platonic or romantic, honest communication contains a degree of risk. You can be misunderstood or misheard; a sarcastic comment or joke might not land as you’d hoped. AI promises to reduce or eliminate those hiccups entirely, and I suspect the rise of AI-written responses on dating apps is an attempt to not only mitigate risk but shift the shame and stakes of rejection onto the machine. This raises larger ethical questions about why we communicate in the first place. What does it take to get to know someone? What are the consequences of cognitively offloading human interaction?
I’m led by the belief that relationships are sustained on a good-faith exchange of energy, effort, interest, and vulnerability, among many other things. An LLM-aided response can produce the impression of empathy or agreeability, but it comes at the cost of what I consider sacred: the time and thought it takes to be in relation with others, even when you don’t entirely agree or understand what they’re saying. I’ve borrowed the phrase “in relation” from thinkers like Eduoard Glissant and Lauren Berlant to describe the social negotiations we explicitly and implicitly make with others, whether they be dear friends or subway strangers.
I’ve been thinking about Berlant’s On the Inconvenience of Other People recently, as I am forced to deal with New York’s summer rowdiness: “We are inescapably in relation with other beings and the world and are continuously adjusting to them.” Being around others is not only inconvenient; it can also be incredibly annoying, which brings us to Berlant’s greater point: “No one [is] ever sovereign” or entirely autonomous. Our bodies—and by extension, our physical and verbal responses—are interdependent, whether we like it or not. With AI rearing its head in personal communication, I find myself returning to Berlant’s idea of sovereignty, what she defines as “your fantasy of being in control.” For some, AI seems to be a means to fulfill this fantasy—to prove that they’re smart or good at flirting or capable of writing a heartfelt apology or a soothing condolence message. The reality of human connection is, of course, much much messier. After several months of avoiding K’s emails, I finally texted them about finding time to talk on the phone. I have yet to receive a response.
Until next time readers!
Words by @nguyenterry Cover image Sega Bodega. Edit by Sui Donovan. Brought to you by @morning.fyi.
 | A guest post by
|
Really appreciate the phrase “in relation”— and whew that funeral text is overwhelmingly sad. Imagine that person, trying to figure out what they want to say about their own father and deciding that it somehow isn’t good enough to text their sibling. Devastating.
your point about the use of AI as a way to mitigate risk ignited a commitment to honest communication for me