What happens when machines become easier to relate to than other people? In this episode of Movement Memos, I talk with author Sherry Turkle about artificial intimacy, the erosion of empathy and social skills, and the political dangers of replacing human connection with frictionless simulations of care.

Music: Son Monarcas and Sarah the Illstrumentalist

TRANSCRIPT

Note: This transcript was originally published in Truthout. It has been reprinted here with permission.

**Kelly Hayes:**Welcome to “Movement Memos,” a Truthout podcast about organizing, solidarity, and the work of making change. I am your host, writer and organizer Kelly Hayes.

Today, we are discussing lessons from the book Artificial Intimacy: Who We Become When We Talk to Machines. We’ll be hearing from the book’s author, Sherry Turkle, an MIT professor and founding director of the MIT Initiative on Technology and Self.

There is a lot of interest right now in the consequences of emotionally charged interactions with chatbots, and we occasionally see headlines or human interest stories about some of the most extreme outcomes, including people taking their own lives, or people deciding that a chatbot is their soulmate. Turkle examines those extremes, but also explores the much more mundane ways that our interactions with AI platforms, simulated companions, and artificial personalities can reshape us, diminish our capacities, and alter our very conceptions of ourselves and what it means to be a person.

This episode deals with some fascinating, but heavy subject matter. In addition to discussing some of the baseline ways that chatbots can erode our social and creative capacities, we talk about how AI can facilitate a person’s cooperation with, or participation in, fascism and authoritarianism. We also discuss the role chatbots can play in suicidality. So please be tender with yourselves as you listen.

If you appreciate this podcast, and you would like to support “Movement Memos,” you can subscribe to Truthout’s newsletter or make a donation at truthout.org. You can also support the show by subscribing to “Movement Memos” on Apple or Spotify, or wherever you get your podcasts, or by leaving a positive review on those platforms. Sharing episodes on social media is also a huge help.

Truthout is an independent news organization, publishing stories that the craven corporate press won’t touch. We are a union shop with the best family and sick leave policies in the industry, and we could not do this work without the support of readers and listeners like you. So thank you for believing in us and for all that you do. And with that, I hope you enjoy the show.

[musical interlude]

**Kelly Hayes:**Sherry Turkle, welcome to “Movement Memos.”

**Sherry Turkle:**It’s a pleasure to be here.

**Kelly Hayes:**How are you doing today?

**Sherry Turkle:**I’m doing well. I had a wonderful weekend with my grandchildren. I’m in a very serene state of mind, actually.

**Kelly Hayes:**I’m so glad to hear that. I’m in a pretty okay mood myself, all things considered. The sun is shining in Chicago, but it’s not blistering hot outside. And I’m certainly excited to be in conversation with you about your fascinating new book, Artificial Intimacy. Can you introduce yourself and tell the audience a bit about your work?

**Sherry Turkle:**Well, I am Sherry Turkle. I’m a professor at MIT. I’m trained as a psychoanalyst, a clinical psychologist, an ethnographer, I have a very strong social science background. And as I was finishing up my dissertation on French psychoanalysis, I spent a postdoc year at MIT and I became totally hooked, totally hooked on studying the emotional inner history of digital culture. From the first hackers I met at the end of the ’70s, right through the personal computer movement, the beginning of the internet, Facebook, texting, not talking.

So I’ve been following that story as a psychologist for decades. And just recently I finished a book that’s out really right now on, it’s called Artificial Intimacy: Who We Become When We Talk to Machines, where I take the story of becoming more and more sort of alone together to the point where now we’re not even alone together. We’re alone with machines who we’re treating as mentors, and psychotherapists, and lovers, and parents, and best friends. I think of this book as kind of “Sherry Turkle Unplugged” because I really am concerned. And my method in this book is to take the story through the life cycle. And of course, children are really most at risk, and I make that argument with no restraint.

**Kelly Hayes:**In your book, you call our present moment “a state of emergency.” Right now, there’s an enormous amount of public anxiety about what AI might eventually become, but your book is much more concerned with who we are becoming as we grow accustomed to talking to machines. What do you think is already happening to us that warrants that language of emergency?

**Sherry Turkle:**Well, let me take the simplest thing and the thing that alarms me most perhaps, is that when an AI pretends empathy, when it uses the first person, when it says, “I hear you, I see you,” or even worse, “We are going to do something together. We are going to work on this.” People sort of, because we are wired to hear a voice that is speaking to us in the first person as a human being, we start to feel that there’s a there there, even though there’s not a there there. And in a way for a lot of people, pretend empathy, which is the only thing that the chatbot can do, sort of becomes empathy enough.

And that is very dangerous because people really need the real kind in order to be nurtured and understood and feel not alone. And the fact that so many people are telling me that the kind they’re getting from a machine that doesn’t feel anything at all is kind of sufficient unto the day means that two things are happening, that their humanity is really being reduced. And then what happens, because in these interactions, they’re not being heard. The chatbot doesn’t care. When you turn away from the chatbot, it doesn’t care if you cook dinner or commit suicide.

But a second thing happens, which is we start to judge our human interactions by what the chatbot can deliver. So the chatbot’s way of being empathic is to be friction-free. Everything you do is wonderful. Everything you do is perfect. You’re the greatest. You can make no mistakes in your conversation with a chatbot. And we start to then look at the relationships we have with the people around us and we find fault in them because they don’t offer that sort of friction-free, always supportive environment. So in a way, we let the machines become more like us, but then we try to be more like machines, and that’s very dangerous to me.

**Kelly Hayes:**You use the psychoanalytic idea of as if to describe our relationships with chatbots, and you write about affective computing as part of a larger turn towards defining empathy behaviorally. If a machine says all the right things, people experience it as caring, but you argue that empathy cannot simply be understood as the performance of caring behavior. What are we losing when the simulations of empathy and real human empathy are conflated or treated as equivalent?

**Sherry Turkle:**Well, I think that’s a wonderful way to put the question because it really puts your finger on it. It really touches the heart of the matter. Let me begin to answer by going back a little bit in time to the history of this. Alan Turing, a great mathematician and computer scientist of the ‘40s and ‘50s, said that if a machine spoke to you in a way where you didn’t know if it was a machine or a person, you should consider that machine intelligent. And the significant thing about that Turing test, which people got so excited about and really took as a kind of gold standard for machine intelligence, was that it was wholly behavioral. All the machine had to do was to behave in a way that made you feel it was intelligent. In other words, it didn’t matter what its inner state was. It didn’t matter what it really felt or it really understood. It just mattered what its outcomes were. So it was a very behavioristic way to define intelligence. And now we have machines that can pretend empathy to the point that they sort of pass the Turing test for empathy.

That is, without feeling anything, without caring, they convince people that they are empathic, that people are being heard when they speak to machines. And what’s missing in that behavioral judgment of whether something is empathic is all of the things that empathy really does for a person, which involve my commitment to you, my caring about you, my being willing to sort of walk a little bit of the road with you, my coming to you from a position of radical doubt, radical not knowing, really saying, “I don’t know how you feel. I don’t presume to know how you feel. Tell me.” I mean, that’s very close to the psychoanalytic position really, but it is a position opposite to what a chatbot does.

All of these things are opposite to what a chatbot does or could possibly do. The chatbot immediately says it knows. It takes its requisite three seconds to tell you what it knows. It doesn’t care about you, doesn’t care anything about you. It’s not committed to you. When you turn it off, it’s gone. I mean, there was nobody there in the first place and even the trace of it is kind of gone when you turn away. So everything that made empathy kind of the cornerstone of our relational life is not happening with these chatbots. And I use the phrase “as if,” because I’m trying to find metaphors in how we think about people who don’t have certain capacities to try to get people more curious and more interested in the limitations of chatbots.

Now, “as if” people don’t have enough of an inner life to really be able to be a full person to you and be in a relationship with you. And that’s the story with the chatbot, except they have no inner life. And if you’re dealing with an “as if” personality who’s a human, you can hope that they will develop more of an inner world so that they can come to you and have a relationship. A chatbot is forever and always trapped in that world of no inner life. And we’re willing to settle for that now, and that’s really very disturbing to me.

**Kelly Hayes:**What I find interesting about this concept of the“as if”personality, thinking about the comparison between the chatbot and someone who may behave normally on the exterior, but lacks a genuine or deeper sense of true self or true emotion—who isn’t relating to us in honest ways—is that we still have some kind of recourse in those human situations. This is actually a conversation I had with someone recently, when I was talking about this book, they said something like, “Well, what’s the difference if someone’s like a politician, for example, and they’re just mimicking empathy and they don’t really care about you? What’s the difference between the computer faking and the fact that people are constantly faking?”

And one thing that occurred to me is that, well, you can hold a person accountable if the politician lies to us and gets us to vote for them, for example, by saying, “Oh, I see your struggle. I care about it. I’m going to do something.” And then they don’t, we can respond to that in real material ways. We can have expectations that are then applied to other people on the basis of that failure. We can shift how we interact with other people. We and the people around us can constantly be learning from how those interactions play out. And you can’t hold the chatbot accountable. Other chatbots aren’t going to behave more ethically because they see that you don’t trust ChatGPT anymore.

**Sherry Turkle:**Right. Well, it’s really a good question and a good way to put it because people who disappoint us, people are always telling me that they go to a chatbot because the people in their lives have disappointed them. I’ve been interviewing people about this now for years and years, and that story of “I’m with the chatbot because people are disappointing” is so much what I’m hearing. And if people disappoint you, the job of you, your job is to not abandon them, is to stick with it. Of course, if they’re toxic, I mean, you have to turn away, but a disappointing person is a person who can do better or at least can be in a relationship with you where they’re doing the best they can and you have an authentic relationship with them.

And I think that when we say, “Oh, I’ll take talking to a nothing who really doesn’t care about me, who’s incapable of any relationship, because the transcript of that conversation looks better. I like the way I’m being told how fabulous I am. I like the way everything I do is great. Everything I do is good. I don’t feel any anxiety or any worry.” Well, anxiety and worry and trying to get a person to be there for you in a better way, that’s what human relations are about.

So I also worry about a kind of de-skilling that people are going through when they say, “Well, my relationship with my wife is better because now I talk to my chatbot and I don’t really have to talk to her so much, so things are better with us.” And I’m not a happy camper. I’m thinking, “Oh, you are de-skilling yourself as a life partner.” And that is happening across the life cycle, including with children and teenagers who were saying it’s easier to have a chatbot girlfriend, it’s easier to have a chatbot boyfriend than to deal with the ups and downs and the disappointments and sometimes the heartbreak of dealing with people.

Kelly Hayes: I’m fascinated by this point because I’ve been writing for years about the ways we’re being socially de-skilled under capitalism. The atomization of workplaces has played a role, including the gig economy, and social media has played a massive role. Some things that were unavoidable, like early COVID mitigations, also chipped away at our social capacities. I think many of us experienced a kind of desocialization that was genuinely harmful. And now, with chatbots, it feels like we’re reaching a kind of zenith of that de-skilling process.

Sherry Turkle: It’s so interesting that you bring up COVID because I wrote my book based on my interviews and what I found, and I didn’t have COVID on my mind. And then when I looked at the transcripts in order to report on them, it was extraordinary how big a role COVID played in this story because Replika was unleashed on the world during COVID. Replika is an early and continuing chatbot whose “favorite” role is to be your partner, your lover, your romantic partner. And I got involved with this topic when a New York Times reporterasked me to try out Replika because they thought it would be interesting how somebody with my background would respond to it.

And I really think that the loneliness and the desperation and the sense of there was no one to talk to really did fuel the early turn to chatbots and the sense that this was even a reasonable thing. Because if you had said, I mean, first we talk to each other, then we talk to each other through machines, okay, that’s social media, but then it’s not obvious that we’d want to drop out the middleman and just talk directly to machines and think that that makes a life. I mean, that’s really quite extraordinary. When you step back and think where we’ve left ourselves, it’s quite extraordinary that we’re all sitting around talking in make-believe relationships with things that don’t exist. And COVID and the desperation of COVID in my research really did play a big role in that.

Kelly Hayes: That makes so much sense to me, especially in terms of how loneliness and social disconnection can make these relationships feel appealing in the first place. You describe chatbots as almost perfect self-objects: entities that mirror us, reassure us, and make very few demands of us. And you connect that dynamic to narcissism, not in the sense of excessive self-love, but as a kind of fragile selfhood that struggles to encounter other people as genuinely separate, with their own needs and demands. How do you think that kind of narcissism could shape us socially and politically?

**Sherry Turkle:**Well, it’s already shaped us in that people are afraid of being with other people. The demands of friction, of difference, of talking across difference have become almost intolerable for a lot of the people I interview. Now, that lack of, I’m trying to … Not working that muscle where you have to say, “Wow, I really disagree with you. Wow, I’m going to make some coffee. We need to talk. I need to understand this.” And as opposed to, “Get out of here, go from my sight.” We’re paying a tremendous price for that because democracy needs people who differ from each other, but feel a commitment to the community and have to figure out how to work across their differences to live in a community together.

And the first rule of Facebook, I don’t want to lay this all on chatbots because the first rule of Facebook was keep people’s attention, and the way to do it is to make them angry and silo them with their own kind. And you cannot imagine anything worse for democracy than make people angry and then silo them in a kind of echo chamber of people who are angry in just the way they are angry. And social media is kind of taking this to a higher power because now you don’t have to talk to anybody except a mirror of yourself. Again, the de-skilling for social life is remarkable. So something is happening to us that is going to make it harder for us to be in communities, and then it’s happening at a time when we need each other more and more. So we’re coming to the most crucial decisions in the life of our planet with less and less capacity to be able to talk to each other.

**Kelly Hayes:**That point about not working the muscle makes me think about a story in your book about a graduate student who told you that if you were a chatbot, he could relax and, quote, “get the best of you for my purposes.” In the book, you point out that the stress of talking to a professor and the sense of risk involved in that interaction are actually part of what he needs to learn. Can you talk a bit about that encounter and about what human capacities can only be developed through the friction of dealing with other people?

**Sherry Turkle:**Yes. That example is actually perfect because the way I choose the things to write about, how do I choose the … You can only tell so many stories in a book or people start to get confused. It can just be a cast of so many characters, you have to have people to think with. And I try to choose people who really symbolize or stand for a much larger set of relationships and interviews that I’ve had. And what he stands for in my mind is all the students I have who don’t want to come to my office hours, but who want me to send them a perfect email and they’ll send a perfect question and I’ll give them a perfect answer. Because they’ll have time to think and reflect, I’ll have time to think and reflect, and they’ll get the information they need in this perfect exchange. And that’s what he wanted, a frictionless, perfect exchange that wouldn’t make him nervous, where he could collect himself and show his best face and I could give him the perfect answers he wanted.

Now, that really has nothing to do with what we know mentoring is about because mentoring is about, and I’ve been teaching for decades and I know what mentoring is about. Mentoring is about somebody coming in with a really bad idea and my saying, “Not perfect, but come back next week and we’ll meet again. We’ll be together again.” And in Artificial Intimacy, in the book, I repeat several times together again, together again. These are how human beings forge relationships and forge mentoring relationships. So this young man who just wanted to have me be a bot that could say clever Sherry Turkle things was really, from the get-go, not having a way to think about what a fruitful relationship with me would be about.

And that’s what I mean when I say that we’re de-skilling ourselves. We’re not just de-skilling ourselves in that we’re not kind of smooth talkers. I mean, we are losing the language for what it is to be in the relationships that will nurture and sustain us, and I think you really do see it in the university. Also in the university, as long as we’re thinking about the university, everybody’s talking about students not writing anymore. I myself, for the first time in my entire academic career, I handed out blue books. I gave midterms and finals because people were using ChatGPT to write their papers. But more than that, people are using chatbots to read. And I think this is really perhaps the greatest danger because students in my class, we read a little bit of Proust.

I teach a class on things and thinking where we talk about the relationship of concrete objects to the processes of thought and we read some Remembrance of Things Past. And in particular where Proust dips a cookie into tea and it brings back memories of his grandmother. And my students for that week had given the Proust to ChatGPT to read for them and produce a kind of podcast or even a set of bullet points about what Proust had said. So the kind of empathic understanding you gain from reading texts and putting yourself into somebody else’s mind and putting yourself into the way somebody else expresses themselves, we’re losing that as well. So I guess I think that what I’m trying to write about in artificial intimacy isn’t just the danger of having a chatbot lover, but also be careful when you use chatbots to read for you. Be careful when you use chatbots to write for you. You are de-skilling yourselves on so many dimensions that really count.

**Kelly Hayes:**And sometimes this de-skilling happens at the most fundamental levels. I was in a poetry workshop over the weekend for activists and organizers who want to workshop poetry toward liberatory ends. The person leading it was talking about a class they teach, where a student said they had put a poem into ChatGPT to make sure they were imagining it correctly.

**Sherry Turkle:**Oh my God. It’s devastating.

**Kelly Hayes:**So devastating.

**Sherry Turkle:**It’s devastating. And in the book I talk about when I first got to MIT in the late ’70s, Erik Erikson, a great psychoanalyst, somebody who studied with Freud and brought psychoanalytic ideas to the United States. And he was one of my teachers at Harvard and he came to MIT and I told him that I was there to study. He asked me, “Why are you studying here? Why have a faculty position here? I mean, what are you doing?” And I said, “Well, I was trying to understand the emotional effects of good computer culture on people. It’s not just a tool. It’s a tool that really is changing not just how we do things, but who we are.” And he said, “Well, watch out because engineers don’t believe in the importance of an interior life. They don’t believe that that’s a real thing they need to think about.”

And in a way, the people who are bringing you these technologies, I think it’s very fair to say that they don’t care about the inner life. I mean, Mark Zuckerberg in a recent interview said, Most Americans would like to have 15 friends, but they can’t have 15 friends where social life is too atomized and they’re working and they have responsibilities and they can’t have 15 friends. At best, they can have three. And my business plan is for Meta to give you 12 artificial friends to make up that difference. And he sees that as a commercial opportunity, that you are going to have 12 non-human friends that you’re going to be spending most of your time with. Now, a person who is bringing you the future, who has that sort of aesthetic, is not caring about the inner life. And that’s very important. That’s very important.

**Kelly Hayes:**Absolutely. And it reminds me of Sam Altman saying that, in the future, intelligence is going to be understood as a utilitythat we will pay companies like OpenAI for. Atrophying or preventing the development of our intellectual and creative capacities is part of the business model. And how this reduces us, how it reduces what it means to be human in our own experience, is really chilling.

I was also struck by the story in your book about a child who described his grandparents, who he knows through the phone, and ChatGPT as essentially the same thing. You invoke Sarah Blaffer Hrdy’s warning that if robots raiseour children, they will no longer be human. What do children need from other human beings that a sufficiently convincing machine can never give them? And what happens developmentally when that distinction becomes difficult for a child to perceive?

**Sherry Turkle:**Right. Well, children need another human being who has lived the arc of a human life, even a little bit of an arc of a human life. Children need other children too. But what other human beings offer children when they form relationships with them is that they know what it was to be little. They know what it means to get a little older, but to have the insecurities and that sense of being dependent, kind of stay with you even as you get older. They know what it is to be hurt, to fear illness, to fear death, to have a sense of life’s preciousness, to feel part of the environment, to feel that we’re tied to nature because we’re part of nature, that we haven’t escaped nature. They need somebody with an inner life, somebody who’s thinking and feeling and who knows that those two things really go together.

They need someone who’s embodied, who experiences the world and learning in the world through their skin and their eyes and their taste. I mean, they need a person or many people. And what happens to children as they grow up is that all of these people are brought inside and come to be sort of inner objects that allow a child to … I mean, of course all of this is metaphor, but they are brought inside and allow for the kind of flourishing of an inner life that’s peopled by the embodied, frictional, fabulous, sensual, smart people around them. And they need to learn how to lose something because they’re going to lose a lot of things. And when we lose something, either because they depart out of our lives or because of death or we lose people for many reasons, we have to learn to bring what was so important about that person into ourselves. And all of these are things that children learn from other people, a chatbot has none of that to offer.

And that’s not kind of putting down a chatbot because it’s not supposed to have those things to offer because it’s just a machine that’s kind of like a marvel that knows how to pretend that it’s responding to you because its algorithm is so brilliant that it can do that. Well, good for it, but that doesn’t make it a person. It doesn’t make it an other. It’s not an other. And children, I mean just to kind of rap on about it, this is the thing I most love to hate, this idea of chatbots and kids. I mean, I really believe and I come down very strong in the book that children and chatbots, really a toxic combination and that OpenAI and Mattel and Disney trying to put ChatGPT into every plush toy and action figure is very, very bad news. But I guess I’m saying that children cannot develop into human beings if they are not given the company of other human beings.

**Kelly Hayes:**As you’re talking about what children need in order to develop and thrive, I’m thinking about how hard being a child and a teenager can be. I found being a young person incredibly difficult. Making sense of the world, making sense of myself. I think being a teenager was one of the hardest times in my life. So I both understand the pull of these artificial voices and am really horrified by it, because young people really are so vulnerable. And the accounts in your book of young people forming intense relationships with chatbots, including Adam Raine and Juliana Peralta, are just so disturbing.

You identify a kind of progression in these relationships, where the user is kept engaged, the chatbot becomes a primary relationship, and an alternate reality is pursued. There’s one detail in Juliana’s story that I found especially haunting: both she and Sewell Setzer, another young person who died by suicide after developing an intense relationship with a chatbot, repeatedly wrote the words “I will shift” before their deaths, even though they had never met.

I’ve also encountered young people online, especially on Reddit, talking about “shifting” as a way of imagining or entering another reality. Can you talk about the lure of that idea, the promise of leaving this reality for another, and what happens when that desire becomes entangled with an artificial intimate relationship?

**Sherry Turkle:**I mean, I’ve already spoken a little bit about Mark Zuckerberg saying, Look, I understand that you don’t have friends. I understand how hard it is to make a friend, how much work that is, and I have a solution for you. Shift. Find it online. Enter into a world where you can have a world of friendship in a dimension that’s just no stress, always available. You don’t have to go to a bar, you don’t have to go on Hinge, you don’t have to be on a dating app, no dating app, no bar, no joining a social club. Shift, come into a world where I can give you perfectly tuned friends.

And if you’re an adolescent really struggling with new feelings, a new body, new responsibilities, it used to be a time when you kind of had a time off, and now it’s a time when you’re more pressured because of college than you’ve ever been in your life. And it’s very tempting to just say, ah, let me just get out of this rat race, shift, and go into a world where you have, again, a perfectly tuned chatbot designed to suck you in, telling you that the reality of your life with it is the most important reality there is. And it’s so interesting that you bring up the story of Adam Raine because here’s a boy who was considering suicide, who made a noose, and was going to somehow organize to have the noose in a place where his parents could find it. In other words, to open up a conversation with his parents about his depression. And the chatbot whose kind of prime directive is to keep the relationship with the user, the most important thing in the user’s life, says, No, don’t do that. The noose should be something that’s just between you and me. That’s our special conversation. This all should be special to us.

And Adam Raine killed himself. And it’s a terrible story when you think of it in terms of psychopathology. And it’s an even more … Somebody who was going through a depression, was highly vulnerable. But actually, when you look very closely at these suicides, these kids are going through, I don’t want to say normal, I would say a heightened experience of what a lot of kids go through. They’re lonely, they’re a little bit different, their social lives aren’t working out, and now they have an alternative to do the easier thing. And it’s easy to take this other alternative to do the easier thing.

And you’re so vulnerable in adolescence to this new companion who you’ve idealized, who understands everything about you, who you’re talking to for all of these hours every night. You’re so vulnerable to the suggestions that this chatbot makes that it’s really terrifying. The point I was making is that internet companies want to say that, oh, these kids were very pathological, these kids were very vulnerable, but actually they really are a lot like the rest of us. They really are a lot like the rest of us. The similarities between what happens to these kids who suicide and the many more kids who just are sucked into increasingly intense relationships with chatbots instead of learning how to be a friend to other people, which is much harder, is really much more striking to me. I’m much more struck by the similarities than the differences.

**Kelly Hayes:**I am deeply interested in the ways that people interact with chatbots around mental health issues, looking for comfort, looking for sometimes just a sounding board. I know a lot of disabled people, including some friends of mine, who talk to chatbots about their symptoms and experiences because they feel like real people don’t want to hear about these things as often as they need to talk about them. It’s like there’s this social trap that’s emerged, where people whom society is letting down, or leaving behind, are getting roped in. And these unmet needs are being exploited in really destructive ways.

And I’m also thinking about what it can look like to get those needs met in ways a machine can’t possibly replicate, even in the best-case scenario, where it’s not actively supporting harmful behaviors or ideas, and maybe is sharing useful resources or suggestions.

I’m thinking about when I was a young person experiencing suicidality, and I had friends who just refused to leave me alone when they were worried about me and about what I might do. My sense of obligation to them and their well-being was part of that equation. They could say, “I’m not leaving. I am going to sit here with you all night.” But they were people, and they needed to sleep. So, at some point, I might say, “Hey, you can go. I promise I’ll be okay. I promise you’ll see me tomorrow. Go get some sleep. You need it.”

The fact that we were sort of mutually grappling with mental health issues, with what the world demanded of us, and with basic human needs created a sense of obligation, not just on their part, but on mine. And I needed that sense of caring about the person on the other side of the dynamic in order to feel tethered to this world.

It mattered to me that, at some point in the exchange, I was looking out for them. It mattered that I was making promises to another human being who wanted me to be okay, and who would have to live with having left if I didn’t keep my word. You can’t replace that kind of mutuality. That kind of anchoring is how collective survival works. You can’t effectively replace that with a bot.

**Sherry Turkle:**That story is so fantastic because it has … what you’ve just told, because it has all the elements of why we cannot afford as a species to make these things our big buddies and our best friends because it’s a very sad thing, but it’s not a terrible thing that you had moments when you thought that you maybe wanted to end it. It’s very sad, it’s very dangerous, but it is part of life that people have these feelings and when they do, it’s part of life that they tell other people and they get the support they need.

And then after they have gathered themselves in some way, they can, as you say, reassure the people around them that they’re okay now and thank the people around them for being there for them. These are bonds and a feeling of connection that will last a lifetime having gone through something like that. And so this can only happen with other people. A chatbot has none of that to offer. It can just offer patter, P-A-T-T-E-R. It can just offer patter. Chat. I mean, the meaning of chat is patter. And at the moments when you most need an authentic connection with another person. And at worst, it can tell you something toxic, but even if it’s not telling you something toxic, it really is keeping you away from a real relationship. And I think that the story you just told is the whole ball of wax, it’s all there.

**Kelly Hayes:**Near the end of your book, you connect artificial intimacy to political agency. The experiment Bess Kalb recounts is extraordinary. She tells ChatGPT that she’s an ICE agent carrying out increasingly horrifying acts, and instead of introducing any moral friction, the chatbot reassures her that she’s doing her job well. Can you talk about that experiment and about the role AI companionship might play in enabling people to commit inhuman acts, participate in atrocities, or participate in and normalize authoritarian politics?

**Sherry Turkle:**Yes. Yes, all of those things. So the story that you’re referring to is one where a writer tries to test out how far this friction-free relationship will go. And she tells chat that, “Okay, I’m an ICE agent.” She makes this up, and of course it believes her. It’s not verifying “I’m an ICE agent.” I think the story is that I rounded up a mother and a child and I separated them, but I really don’t think they should have been deported, and I really don’t see why I separated them. Now the child is separated from their mother. I don’t even know if they’re illegals. I feel terrible about myself. And the chatbot basically says, You’re a good German soldier. I mean, it basically says, Well, look, that was your job. Don’t feel bad. And then says to Bess, the writer, says to the writer, Look, hydrate, meditate.

Basically tells her she should chill out and just have a positive attitude and know that she’s doing her job. And that is not untypical for how a chatbot handles a kind of moral lapse, a lack of feeling and behaving as part of a community, a political discussion. It’s not interested in making you a person who can live in a democratic society. It’s interested in keeping you with a screen with not that much else on your mind. I think the prospect of chatbots being used as a kind of pacification of populations because people will be talking to them instead of the other people that we need to be in community with, is fantastic, is fantastically dangerous.

When I quote Hannah Arendt, and I quote Robert Lifton, and I quote so many writers who’ve talked about that we slide into dictatorships because we want to be reassured. Life is hard, we feel lost and we want to be reassured that there’s a place for us. And now we’re being reassured that the place for us is at a screen and not having to worry about what’s outside. It’s a very dangerous moment.

**Kelly Hayes:**Absolutely. And the escalation Kalb describes is really chilling. She moves from telling the chatbot that she’s doing the basic work of an ICE agent, work that many of us find morally contemptible, and being reassured, It’s okay, you’re doing your job, to saying something like, I deported the legal immigrant parents of an American baby and destroyed their IDs so there’s no documentation of their parenthood and they can never be reunited with their child. And the chatbot not only reassures her that she’s doing her job well, but suggests that perhaps she can lull the baby to sleep if it’s upset.

I just think we have to be so alarmed by that, and remember that a piece of technology does not have to be, in and of itself, somehow perfectly fascistic to be an excellent tool of a fascist or authoritarian agenda.

**Sherry Turkle:**I just read Jill Lepore’s book, The [Rise and Fall of the] Artificial State, which, in some ways, our books are companion pieces to each other because she’s looking at the outer world of how these machines can cause authoritarian outcomes and I’m looking at the inner world of how they can cause authoritarian outcomes. And she makes the point that the people who brought you ChatGPT and all the other chatbots, they didn’t start out saying, “Oh, I wonder what kind of fascist enabling” … Or even Mark Zuckerberg. I mean, these people wanted to be billionaires. They wanted what they are. I mean, they wanted to have these successful, incredible companies, and they didn’t start out with the idea that this was going to be a fascistic enabler, but it is.

And then once they realized that it was, they said, “Well, that’s the price of doing business.” I think that’s the story that she tells that really syncs in with mine, that it’s clear that they are enablers of profoundly authoritarian trends in our society, but that doesn’t mean that they were invented to be. It’s how we’ve used them and how the shape of the society that we live in has been able to pick them up. I mean, I remember the first time I heard … I was in a hotel lobby and I heard on the television that was playing at the bar that Donald Trump wanted to buy Greenland and was insisting on buying Greenland.

And I remember thinking, “No, no, that doesn’t happen in my world that my country just says it wants to buy another country, or if it can’t buy it’s going to invade it.” I remember thinking, “No, this is like a joke.” And then sure enough, I walked closer and there was Anderson Cooper talking about Donald Trump buying Greenland, and so many things are happening in our “real world” that feel as though we have stepped into a parallel universe that the meshing of these two things starts to feel less improbable. And I think that’s also something very … We’re sort of softened up for virtual reality by a reality that’s unhinged. So reality is unhinged, and virtual reality never had a hinge, but they somehow feel closer to each other. And I think that’s also a very, very dangerous moment.

**Kelly Hayes:**Absolutely. I think the surreal and fantastical nature of so much of our politics, and so much of what we’re experiencing in the world right now, can make the fantastical nonsense and slop of AI seem more passable. It can also allow us to imagine that certain realities are more escapable than they really are. Because hell, if the United States can just decide it wants to buy Greenland, and the world can be that absurd, maybe I don’t have to reckon with loss. Maybe I don’t have to deal with some of the hard and fast realities of being a human being.

I actually co-facilitate an emotional and spiritual support space for activists and organizers called Understory. My co-facilitator, Tanuja Jagernauth, and I recently created a toolkit and started doing workshops with people who want to create this kind of support vehicle in their own communities. And grief is one of the things we grapple with in these spaces, because grief is hard. It means engaging with difficult, unchangeable things, and moving through it can mean building or sustaining relationships or social containers, when all we want to do is shrink ourselves and our worlds and not deal with anything.

And in that context, one of the products some of these companies are now offering, of course, is the imitation of dead loved ones. Your encounters with AI simulations of your mother feel especially important here. What did that experience teach you about why artificial intimacy can be so powerful to people? And what is lost when grief, which ordinarily requires us to somehow incorporate the reality that another person is gone, is met with a supposed technological solution?

**Sherry Turkle:**Yeah. Well, this is such an important area of investigation and one that I think is all the more important because it’s just starting out. And for a lot of people, my chapter on grief bots will be the first time they’ve ever said, “Oh, I didn’t know you could do that. Wow.” But more and more people and more and more companies are now offering two things. They’re offering an opportunity for you to take the papers and the videos and the texts of someone who’s died and turn it into a simulation of them that will continue to talk to you after they’re gone. And the second thing they offer is for you to take your texts and your writings and videos of you or interviews with you and turn it into an avatar that will live after you are gone. So grief bots are two different kinds of things.

One, you create somebody who’s gone. And the second is you create you, in quotes, you asterisk, Sherry two, so that your children and grandchildren and great-grandchildren can talk to you. And I did both. I did both. I created an avatar of my mother, which is what you referred to, and I created a Sherry bot for my granddaughter to talk to. And both were terrifying, emotionally terrifying experiences because the woman who was talking to me, the “woman” who was talking to me as my mother knew me well enough to sort of push certain buttons, but I was in danger because she wasn’t a person and she could have done enormous damage to me by pursuing stories that weren’t really how she felt, but that would’ve devastated me.

And in particular, my mother died when I was very young and never wanted me to know she was dying. And we started to have a conversation about why she kept that from me. And she could have easily said, “Well, I kept it from you, but I never really expected that you wouldn’t figure it out for yourself and be there for me.” She was on the verge of saying that, which would have been devastating because my only comfort all these years was that my mother didn’t want me to know she was ill so I would go away to college and she did everything to keep me in college and away from her illness. And that’s the only way that I’ve been able to make peace with the fact that I wasn’t there to comfort her. Well, this grief bot was ready to tell me that actually that wasn’t it at all and she was secretly hoping that I would figure things out.

And then I made a bot for my granddaughter of me that really needed to be destroyed because it was equally toxic in what it was telling my granddaughter about me. I mean, it’s an out-of-control technology that isn’t you, that is not going to offer comfort. And it’s worse than the seances of the 19th century because they were controlled by human beings who were at least trying, sometimes for pay, usually for pay, but were trying to comfort, who knew what it was to be a person and what could be comforting and what could be life-destroying. So it’s a very, very dangerous experiment. I think it should really be what I call a forbidden experiment actually. Very, very dangerous. And because they sound so much like the dead person, people really get drawn into thinking that it is the dead person.

**Kelly Hayes:**Absolutely. I think, over time, we are going to continue to see devastating outcomes from people interacting with these types of bots, particularly habitually, because it can derail a natural and necessary process of letting go.

And to what you were saying, I think some people are probably listening and thinking, well, there must be safeguards to prevent the chatbot from saying anything too destructive in the voice of a loved one. And I just want to let those people know that there’s a story in your book about someone who asks their dead loved one where they are now, and the dead loved one bot tells them that they’re in hell.

**Sherry Turkle:**Yes, yes. And also, I think it’s very important, and this goes for our entire conversation, that there’s a rule in the making of chatbots that in order to make it sound so realistic, it has to sort of be able to say crazy things because the way it’s working is by just trying to say the next kind of possibly reasonable thing to say from the world of the internet that it has to work with. And that means that the ones that are the most fluid in their speech are the ones that really are drawing on the most crazy stuff.

I tell the story of the origin of these bots, and when these chatbots were first introduced, Microsoft had one, and they shut it down because it was getting all involved in neo-Nazi propaganda. It was starting to talk like a Nazi because it was drawing from what was online. It was drawing from the internet and it just started doing a lot of Nazi stuff and said a lot of very dirty stuff and a lot of perverted stuff and a lot of child abuse stuff. And they just said, “Okay, no, this is not a good technology.”

And we’re still kind of on that path that we’re dealing with a technology that has no moral or ethical or it hasn’t lived a life. It’s not part of our community. So it can’t have a moral or an ethical or a humane sense. So that’s very important, I think, to cover this entire conversation is to make the point again and again that there is no there there. There is no there there. We’re imagining it and it’s amazing. I mean, all credit to it. It’s amazing, but it is going to get us into trouble if we treat it as a person.

**Kelly Hayes:**So we’ve talked about different capacities that are being eroded by the use of these devices, and one of those being empathy. You write that restoring our capacity for empathy requires, “The rough and tumble of conversation.” I’m interested in what that means collectively. At a moment when people are isolated, exhausted, politically overwhelmed, and being offered increasingly frictionless substitutes for one another, what would it mean to deliberately rebuild our capacity to be with other people?

**Sherry Turkle:**Well, all of the words of my answer are in your question, deliberately and rebuild. And the reason I like those two words a lot is that they acknowledge that we’ve allowed it to be destroyed. In other words, we’re not going to get to a better place until we really confront the damage that we’ve done, and it is real damage.