AI Companionship is Unnatural

As a former journalist, I really enjoy films about investigative reporters. Last weekend I rushed to see The Social Reckoning, Aaron Sorkin’s drama about how Wall Street Journal reporter Jeff Horwitz worked with the Facebook whistleblower Frances Haugen to obtain damaging internal Facebook data. It’s a good story, and in real life the Wall Street Journal’s 2021 “Facebook Files” series exposed that Facebook was aware of flaws in its platform that caused various harms, especially to young people. The reporting led to a federal lawsuit, culminating in a $17.1 billion settlement in August this year, and Meta’s commitment to introduce new limits on teen use of its social platforms. 

The Social Reckoning reminded me of two realities: It’s foolhardy to trust tech companies to act in the public interest, and litigation is a limited, much after-the-fact way to make amends. Plaintiffs get some pecuniary relief, but lives are not easily repaired. Remember, five years elapsed between publication of “The Facebook Files” and the settlement. And, of course, social media-driven pathologies took shape long before 2021 and Horwitz’s reporting. 

And we’re not done; in fact, we’re now being enveloped by generative AI chatbots, a technology many times more powerful than social media. You don’t have to believe AI will bring about human extinction to grasp the chatbot danger to human society. 

That’s why I was so delighted to discover that the legendary MIT sociologist Sherry Turkle just published Artificial Intimacy: Who We Become When We Talk to Machines, which is the book we really need right now. At a time when tens of millions of people already use chatbots as companions and AI startups are racing to insinuate AI into everything from children’s education to human grieving, it’s a relief that a researcher of Turkle’s standing has written a book that shouts, “Wait, there is something really wrong about all this.” 

Until now, I might have suggested that Pope Leo’s Magnifica Humanitas is the strongest warning to date along those lines, but Turkle’s on-the-ground, investigatory approach, involving many interviews and volumes of research, is very complementary to the encyclical. It is also highly readable. Turkle approaches the topic from a secular, “pro-human” standpoint (to use her words), but at the end of the interview below, she quipped that the encyclical is “definitely a Sherry Turkle adjacent thing.” 

Here are excerpts from an interview with Professor Turkle conducted by phone on October 21, 2026. 

The first chapter of your new book, Artificial Intimacy, is entitled "State of Emergency.” How late are we?

Well, we are very late. But I got tired of being told that the horse was out of the barn for [AI] applications that have been around for just two and a half years. So particularly in the areas like education and mental health, where people are arguing [that AI] should be mentors, teachers, and psychotherapists, I come down on you need a consumer movement, an activist movement, a political movement. You need a Ralph Nader. It's not too late. 

Where I think you start to get that feeling of exhaustion before the fight is the money and publicity lined up on the other side. On the day that I gave a talk about the necessity for social infrastructure for seniors—not just leaving these poor seniors alone in their apartments except [with] chatbots—I came across this new series of ads about an older woman knitting alone, talking to her new wonderful ChatGPT that's made especially for seniors. And she's sitting there knitting and happy as a clam—without any people.

We’re being pitched every day to make something unnatural seem natural. But there is no supporting evidence. All the evidence is coming the other way. I saw a remarkable study where they gave people ChatGPT as much as they wanted, and then they gave them random college students to talk to. The random college students did better. People said they felt better. They were more energized. They felt more of a sense of agency and connection. 

We just need to set up the places where college students can talk to seniors. This can be done. Give me the money we're spending on one data center, and I will get you armies of people who want to talk to seniors, who want to organize and talk to kids, who want to run adolescent play groups.  

Are we just looking at another step down the road to finding efficient ways to deal with “unproductive human assets?”

We have to stop thinking of humanity that way. It's a fundamental flaw. People are not assets. People are people. 

I was at a conference where a formulation came up that’s really helpful: There are some things that only people can do for people, and we really need to keep AI in its lane so that AI is doing the things that an AI can do. It's not an anti-technology position; it's a pro-human position. AI is not equipped to talk to you about death or dying or illness or loving because it doesn't have an experience of those things.  

Saying that very obvious thing has become, in the current atmosphere, a sort of utopian, unrealistic vision. People are saying to use AI for things where it has no intrinsic value, doesn't know anything, and can only do real harm—and this is the common sense position. So if my position becomes a utopian dream, and putting AI in positions where it has absolutely no [place] is common sense—that's crazy. 

We more or less survived mobile phones and social media. Is this different?

It is. We've never had something that said, "Why don't you drop the people in your lives and sub me in?”

I quote in my book the anthropologist Sarah Hrdy, who gave a talk and said, “If robots raise our children, they won't be human.” If generative AI raises our children, they won't be human. They won't. They won't have empathy. They won't have empathy as we understand human empathy to be. Even now, my students can barely talk to each other. 

What does a society with children brought up that way look like?

If you grow up with pretend empathy, two things happen. If you say that pretend empathy is empathy enough, which is basically what we're being encouraged to do, you define empathy down to what machines can give, and then you judge people by the kind of empathy that machines can give. 

That means that you look at human empathy and you find it wanting because [people are] not sycophantic, they give you a hard time, there's lots of friction. My book is filled with people who say, "Who needs this?” And the message of my book is: We need this. 

In a place like Silicon Valley, where there is such an emphasis on individuality and autonomy, do you think there is more of an opening for machine empathy? 

Yes. It's also a society with a radical behaviorist culture that thinks, like the Turing test, that “intelligence” is doing something that convinces you it's a person, even though it's not a person. Now, [AI] can pass the Turing test for empathy—it can convince you that it has empathy, even though it absolutely doesn't have empathy. Empathy isn't just saying words that sound empathic. It's caring that you live or die. It's caring that you exist and thrive. And an AI genuinely, genuinely doesn't care, when you turn away from it, if you cook dinner or commit suicide. Genuinely doesn't care. That's the nature of its nature. 

It doesn't live in our world. It doesn't care if your water is clean. It doesn't care if there's garbage on the street. It doesn't care if the air you breathe is toxic. It doesn't have a body. It doesn't fear death. It doesn't fear pain. It doesn't know what it is to lose a child or lose a spouse or have somebody leave you or fear that you have to leave somebody you love and care about. It's an alien. Why would we want this alien as our closest confidant? 

And yet it's so seductive. Millions of people have switched off humans in favor of their AI. 

Yes, and that's why I got so upset that I wrote the book. I was interviewing people and I thought I would get a little bit of this and that, but I didn't. I got people who were happy to say, yeah, I keep my husband around, but really my relationship is with my chatbot. 

Jonathan Haidt's book The Anxious Generation helped achieve a bit of a reset for young people. Is that sort of consciousness-raising helpful here? 

Absolutely, except his solution is sort of technocratic. Lock up the phone. My solution is, you've got to make a friend. It's not enough to lock up your phone. You've become deskilled. You need to learn how to talk. You need to do this hard thing. You need to build social infrastructure. 

If we get this wrong, how dark can this get? 

I quote Sarah Hrdy, who says if robots raise our children, they won't be human. Another way to say that is you're going to have a lot of children on the spectrum, but worse. I talk about “as if”* personalities [in the book], and I talk about a kind of autistic way of life, narcissistic way of life, where we're not listening to each other. We don't know how to know our own boundaries and respect who we are. So we can't really talk to each other. We really are at a tipping point. I call [the book’s first chapter] “State of Emergency”—I've never started a book [with] ‘state of emergency.’ That's not my style.

On empathy

People are not born with empathy. When you're in a household where people listen to you and then they say, "What do you think?” And then they say what they think, and that teaches you empathy. You need to cultivate empathy. Empathy is actually quite difficult because you have to actually shut up and listen a lot over time. It's a discipline of listening and being in the present. A lot of things in our culture are militating against it, and you have to shut off a lot of noise in order to be empathic. 

That's why I tell the story [in the book] of turning to AI during COVID and trying to say that I was afraid of dying. I don't know if you remember, but night after night on television, people would be dying, and nurses would be holding up iPads of their children's faces. That's how they were saying goodbye to their children because nobody could come close. I was starting my research on AIs; I was in the age range and at risk. Everybody was saying [to me], "Talk to the AI.” But it didn't understand anything about how I felt about [possibly] saying goodbye to my daughter on an iPad. It was completely out of its lane. It was obscene. That's when I got this idea that AI is really just out of its lane. It doesn't mean that it can't do fantastic things and develop new drugs for Parkinson's and cancer. I mean, God bless. But there are some things that human beings owe each other.

On AI and human agency

I have a grant from the Templeton Foundation to study how chatbots affect human agency. The early research is showing that it's bad for human agency because every time a chatbot tells you that it will do something for you, it's basically telling you [that] you can't do it. You can't do it alone. You're not enough. You can't write this paper. You need a helper. You can't read this book. It needs to be turned into a podcast. 

On science fiction

What I find dangerous about talking about sci-fi is that so many sci-fi things that were written as cautionary tales are picked up by the tech community as though they're utopian scenarios. I have a very dark reading of Blade Runner, [but] I was giving a lecture, and somebody says, "Well, that's the future. We'll all have these replicants, and it'll be fantastic.” In other words, sci-fi used to be based on dystopian futures, and people were willing to read it as dystopian. Now you portray something where our companions are robots, and people say, "Right on.” Here are these guys, and they are mostly guys who say, "Not to worry. I'm going to be under the earth, or I'm going to be on the moon, or I'm going to be in orbit, or I'm going to live forever on a chip.” I'm trying to stay away from science fiction because it's so contested now. 

On AI consciousness

If we are at a point where an AI shows signs of genuine self-consciousness, the thing to say is that it's self-conscious, but it's an alien. It doesn't get the protections and the rights and the responsibilities of human beings. I think if something shows signs of self-consciousness, [we] will have a new kind of respect, but it doesn't mean you let it vote, or let it be your husband, or you let it raise your kids. But I do believe that there probably will be a code of ethics when the AIs get complex enough.

* In “Artificial Intimacy,” Turkle writes, “As-if personalities can act in a socially acceptable manner, but their actions don’t reflect genuine feelings or spontaneity. They suffer from essential emptiness as they try to behave their way into normalcy.”

“Why Silicon Valley Is Turning to the Catholic Church,” Elias Wachtel for The Atlantic. A nuanced, very well-reported account of the many-sided interactions between the Catholic world and Silicon Valley. It delineates the emerging division among Catholics between a more worldly versus a more evangelistic approach. 

“AI and Evangelization,” George Weigel for First Things. In this brief meditation on the debate about AI and how its creators understand the world, Weigel makes a timeless and very necessary point: “Permit me to suggest, however, that the Church’s primary and most urgent task in confronting AI is evangelization: doing everything we can to persuade those developing AI at breakneck speed that we are creatures of a Creator, not bipedal algorithms with data-processing capacities to be maximized.” 

“Microsoft CEO Nadella Calls for ‘Emergency Brake’ on Advanced AI,” María Paula Mijares Torres for Bloomberg. Microsoft continues its campaign calling for sensible controls on AI. “We need to separate the supply of intelligence from the authority over it.”

“Eight US States sue OpenAI, claiming ChatGPT harms young users,” Diana Novak Jones for Reuters. 

“Slash, Prune, or Let It Grow: How Would You Manage A.I.?” The Editorial Board for the New York Times. The editors ran a poll with twelve questions ranging from autonomous cars to deepfakes. The results and the commentary are fascinating. More broad-based agreement than you might expect. Most overwhelming agreement? 97 percent of respondents want a requirement that political phone calls from AIs disclose that they are AIs. 

“The Original Sin of AI,” Sherry Turkle for The Atlantic. A preview of Turkle’s arguments in her book, Artificial Intimacy: Who We Become When We Talk to Machines.