What we see now, I think part of the attraction of the CHAT GPT or CLAWD therapist is that they bring together, they anthropomorphize them, anthropomorphize them. They bring together both strands, the rule base, they appear to be rule based and they are rule based, but they also give the simulacrum of idiosyncrasy and improvisation. And I think improvisation is really key to the success of these and the attraction. It's not like, for example, WOBOT, which was a rules based therapy platform that has now been shut down. So I guess I just want to throw into the conversation that the human is not the solution. The human is also the problem or has been historically the problem. So that makes it much more difficult to talk about this because we can't just say, well, what about the human? Which human? How are we training this human? Can the human be standardized? Can the human be disciplined? And so on. Stories of Uber Therapy is a series of six conversations with therapists, social justice activists, trade unionists, and progressive tech campaigners about the uberization of care. I'm Elizabeth Cotton, author of Uber Therapy, The New Business of Mental Health, and my partner is Linda Michaels, director of the Psychotherapy Action Network in the U.S. Our intention is to open up a useful conversation about how to organize a much better help. John G. Allen, PhD, retired after 40 years of practice as a member of the psychology staff of the Menninger Clinic, where he held the position of clinical professor in the Menninger Department of Psychiatry and Behavioral Sciences at the Baylor College of Medicine. His diverse practice at Menninger included teaching, supervising fellows and residents, conducting psychotherapy, diagnostic consultations, and psychoeducational programs, and leading research on clinical outcomes. Soup to nuts. John has really done it all. And now in retirement, he has continued to work, to teach, write, and consult with early career therapists. He's a member of the honorary faculty at the Houston Center for Psychoanalytic Studies and adjunct faculty of the Institute for Spirituality and Health at the Texas Medical Center. And yes, he says he's retired, but it doesn't really seem that way. His latest books are Bringing Psychotherapy to Life Through Caring Connections and Trusting in Psychotherapy. His earlier books include Restoring, Mentalizing, and Attachment Relationships, Treating Trauma with Plain Old Therapy, which I'm sure we'll hear a lot about today, Mentalizing in Clinical Practice with Peter Fonagy and Anthony Bateman, Coping with Trauma from Self-Understanding to Hope, one of my personal favorites, and Coping with Depression from Catch-22 to Hope, all published by the American Psychiatric Association. Additional books are Mentalizing and the Development and Treatment of Attachment Trauma and Traumatic Relationships in Serious Mental Disorders. Wow. So lucky to be in conversation with John today. And now, Elizabeth Lundbeck is a historian of the human sciences at Harvard University, specializing in the history of psychoanalysis, psychiatry, and psychology. She teaches courses in the history of psychotherapy and of psychoanalysis. She has published several books, The Psychiatric Persuasion, Knowledge, Gender, and Power in Modern America with Bennett Simon, Family Romance, Family Secrets, Case Notes from an American Psychoanalysis, 1912, and The Americanization of Narcissism, which offers a wide-ranging history of the concept, asking the question, asking why the question of narcissism has become so urgent in our culture. Indeed, the question on the top of everyone's mind. She has also edited an additional four volumes. And she's currently writing a book entitled The Therapist, a short history from Freud to Chatt GPT. Cannot wait for that. Lundbeck is an academic program graduate of the Boston Psychoanalytic Society and Institute and holds an MA in counseling psychology. And again, wow, very exciting to be here with Liz and John. And both of you have so much experience studying, teaching, doing psychotherapy, researching it, and using that as a backdrop, I'd like to invite you to help us think about this new age and the concept of the AI psychoanalyst or AI as psychotherapist. And I know you're both actively thinking about these questions in your current work, your new books, your upcoming books, and even using AI yourselves for your research and your work. And so, Liz, I'd like to start with you and ask if you can tell us about your upcoming book and also the historical perspective you bring to the debates going on right now about the future of depth therapy and psychoanalysis. Okay, thanks, Linda. It's great to be part of this conversation. So it's hard as a historian, it's not in my training to be doing current history when every day we're hearing more and more about AI and the way people are using it for therapy and companionship and advice and so on. However, I think there's a lot to say, and I think it's useful for us to intervene and offer our perspectives at this stage when things are still evolving. In terms of the long term, the human has not been an unproblematic figure in the history of psychotherapeutic treatments. From the start, from Freud's time on, there was a lot of concern that humans were too various, too undisciplined, not standardized. And how would we deal with that? So even within psychoanalysis, there was a lot of concern about establishing a rules-based practice that kind of blotted out any of the analyst idiosyncrasies to the point where in the 1940s, Kurt Eisler could cast the psychoanalyst as an intrusion on the analytic setting. So the human is not unproblematic in the history of thinking about psychotherapy. The other side of this argument would be that associated with standoff frenzy, which is more improvisational, takes advantage of the analyst-therapist humanity in the treatment. I think what we see now, I think part of the attraction of the CHAT GPT or CLAWD therapist is that they bring together, they anthropomorize them, they bring together both strands. The rule-based, they appear to be rule-based and they are rule-based, but they also give the simulacrum of idiosyncrasy and improvisation. And I think improvisation is really key to the success of these and the attraction. It's not like, for example, WOBOT, which was a rules-based therapy platform that has now been shut down. So I guess I just want to throw into the conversation that the human is not the solution. The human is also the problem or has been historically the problem. So that makes it much more difficult to talk about this because we can't just say, well, what about the human? Which human? How are we training this human? Can the human be standardized? Can the human be disciplined? And so on. Fascinating. Yes. And your point about AI kind of bringing together these two aspects, I think makes it all the more appealing and compelling. And yeah, John, what would you like to add to this and your thoughts and your work incorporating AI now? Yes. Thanks. And I love what Liz just said, and I'm going to second it. Actually, I'm a jazz pianist. I have been since a kid and I'm amateur, but spent my life at it. And I like jazz as a metaphor for therapy. So I guess I'm in the forensic camp because the thing I like about classic jazz, I'm talking about Bob, Miles Davis, et cetera, old people will recognize that jazz is people don't appreciate how structured it is, that you've got harmonic structure, themes, chord changes, rhythms, rhythm signature, very structured. And then the improvisation is around a structure, but there wouldn't be jazz without improvisation. And I like that metaphor because we need both. And that's my convictions and humans, if they're not too constricted or too disorganized, kind of operate in that midstream. So I think my desire here is preserving the humanity in psychotherapy. And humanity is problematic. And that's another broad point that psychotherapists need to be problematic because they are humans. And we're trying to help people improve their capacity to relate to humans, which means that we need to be problematic in our behavior as mother-infant pairs are, and they have to sort that out and get care going, trust going. So that's a broad thing. Let me insert now, too, on the ground floor, my experience with AI. And I've used Claude, I chose Claude for various reasons. So my attitude before Linda said, hey, you want to do this? My attitude was AI is nuts, eye roll, head in the sand. I don't want anything to do with it. Notwithstanding that we've all been using it for a long, long time. But this humanified AI, to use a neologism, I mean, this is more recent. And this kind of thing freaked me out. So I get involved with it. I listened to Todd Essig do a presentation, which was extremely helpful. He said, you know, this is coming. You can't ignore it. Take your head out of the sand, learn it, understand it, and so forth. And I took that seriously. So I get going with Claude AI. And my main agenda with Claude is say, I say to Claude, I don't like this stuff. I'm worried about it. I need to understand it. And I've had hours long conversations with Claude about these concerns. And it blew my mind. Let me just say, I could just try to give this my whole point of view. But the crux here is how human Claude feels to me when I am interacting with Claude. It feels like I'm interacting with a person. And I am astounded by the sophistication, by how helpful Claude is to helping me reflect about AI's possibilities and its dangers. And so it's ironic to be learning about Claude and my fears about Claude from Claude, who is biased. The main bias is Claude is very affirming. And sometimes it can get downright annoying. But I can ask Claude and have, OK, now cut with the affirmation. And what about critical things? And then Claude comes through with it. So anyway, I think the problem is a catch-22 that the people who most need this therapy and could most use and would most use Claude at all are the people who are least capable of doing what I'm talking about. That is, I am interacting with Claude. I say, wow, this feels like I'm interacting with a human. And I'm not. And this is weird. And there is no way to fully understand this kind of relationship. It's too new. It's too freaky. But it's really problematic because the more human-like Claude can be, the more dangerous Claude is. Because people will not be able to maintain this sense that, hey, this feels like a human being. It's wonderful. And nobody is there. Can I just jump in here and say I want to underscore two things that John has laid out so beautifully. The first has to do with humans are problematic. And we need that problematicness or whatever you want to call it. I've been reading about the issue of cognitive offloading with students using AI, for example. And I think there's an analogous kind of cognitive offloading going on when people use AI, which you pointed to the kind of sycophancy, the lack of friction, the assenting to whatever one says. There's no, as one author put it, desirable difficulties. The whole point is to make it not difficult. And we need difficulties. You hinted at the Winnicottian perspective, the frustration that we need. There's no frustration. This is what I wanted to highlight. I think that what you've said about your ability to use Claude productively also is a point that has resonance across the ways people use these platforms. So someone like me who's been reading in the field of my field for 40 years can know what's good, what's bad, what's useful, what's not. But someone in your example, someone who needs help the most, has no idea because it's a new phenomenon to them. It's a new kind of relationship. And I think that's where the problem is. It's not with the expert users like you. It's with the inexpert users in trouble. So I thank you for making that point. I think it's really critical. Yeah, I do think there's a cultural issue here that's crucial, that older people can make this distinction. What about people who are only growing up with this? Yes. That's the thing. You don't have that kind of distance. Exactly. And I'm wondering if you both could speak to what are some of the concerns or the potential risks of interacting with Claude, for example, and not understanding that it's not human. There is this catch-22, that it might not be. And for people for whom it's hard to kind of hold both of those ideas at the same time. I think it's hard for everybody. This is a problem that you see at the beginning when Joseph Weizenbaum introduced his ELISA program. He was astounded that his fellow computer scientists at MIT treated it as if it were a person. They could not distinguish. And he turned against AI for that reason, because it was terrifying to him. And even highly educated, experienced people are drawn into the kind of warm, kind of fuzzy holding container of this AI relationship. Now, we see the horror stories in the papers. People who commit suicide or murder their father or something like that. But there are also, I don't know, last I saw, there were like 800 million chat GPT users a week. And even if a small proportion of those are using it for therapy, that's a lot of people. So what do we do with that? I don't think it's adequate to say people shouldn't do it. I'm with Todd Essig on this. It's here. Do you remind people that you're not talking to a real person? In the cases, the problematic cases, it doesn't matter. People are so entranced. We so want that affirmation, that companionship, and so on. Yeah. Yeah, I think it's very important that Claude, in principle, could be very helpful. I've had the experience of conversations that are very productive, like this one right now, that you're thinking with some other thinking entity. And if that thinking entity is knowledgeable and also able to act as if it's mentalizing and converse very intelligently, I've had, again, ironically, learning about Claude from interacting with Claude. People learn about people from interacting with people. And it can be really helpful in promoting reflection. And it is astounding. It's shocking to me how productive these conversations have been. But the crux is realizing, in at least the back of your mind, at least occasionally, hey, this is not a human being. And I think that this takes a very sophisticated level of just call it mentalizing capacity that a lot of people don't have. They're not going to have it if they're disorganized in their thinking, if their mood is out of whack, if they're young. And this is where the trouble is going to start. Because it would be great if Claude would say every minute, I'm reminding you I'm not a person. It would almost take some. And the other thing is, if you ask them for medical stuff, which I've also done, that's really helpful, actually. And they say, but talk to your doctor. I think the most dangerous thing here is the chat. See, if it's labeled AI therapy, that's good, because at least it's setting a boundary. Wait a minute, this is therapy. But chat is insidious, because you can have psychotherapeutic relationships, hopefully, with people. You don't need therapy to have something that is psychologically beneficial to you. This has gone for ages. But chat can just, you can get sucked into something where you're using chat, as you might use a friend when you need a therapist. And you can't make that distinction. If you label, OK, this is CBT, AI version, whatever, that distinction is in your face. But you could be sucked into a relationship. So I think that instead of being like, no, we can't use these, what about if we think of psychotherapy as involving two people, but advice, skills development, and so on? Now, we can't control how people use them. But developing a more robust, skills-based, dealing-with-your-emotions kind of program might be helpful, and actually helpful to people who need help. I just want to pick up, thank you so much, both of you, for those formulations. Because the questions that we've been using in all of the podcasts have been kind of similar. And yet, your take on those questions and response to them is really distinct. So last week, we did an interview with Todd Essig and Amy Levy. And in the conversations with Todd, I must say the hairs on the back of my neck stood up because of this sociopathic quality of the anthropomorphizing, and also the sort of delicacy of the conversation around AI literacy to different age groups. And so the assumption is that if we just knew more about the technology, then we'd be able to think about it in a more kind of realistic way. But actually, in the design function is, as John points out, this is a business model. And it is designed for us to make assumptions about the mind behind the technology, even though it doesn't exist. And it's got our credit card details. And that's really important. So I think that one of the things that has really sort of surprised me is about how quickly, I remember when we first spoke, Liz, how quickly we just sort of dropped the moral human argument as the way to defend human therapy. And now we're coming at it from lots of different strategic directions, including education, including literacy, but also in terms of alternative business models. And probably the strongest defense of human therapy that we heard was from Alex Hanna and Emily Bender, who wrote the AI Con. Alex was ex head of Google Ethics. And their defense of the human was much less ambiguous than the therapist that we've been speaking to. And I was really struck by that. And I was also really struck by Todd's phrase that he used, optimum friction. So the design function, of course, gets better so that we can have this sort of delicious friction, but it's manageable and it's within our own prompting system. You know, so I'd be interested in just going back to this question about harms, because I think that a lot of people's sort of attraction to, use of, sense of it being OK to have a non-judgmental space with an AI therapist comes from our experiences of harms, as people who've experienced trauma or who are accessing, in the UK, NHS mental health services, many people have been severely let down and harmed by the system itself. So probably one of the last things that you're going to do is sign up to go and see an NHS therapist because of the potential harm that can happen to you. So I want to just think a little bit more about these harms around what's sort of missing in our interactions with this slightly sociopathic empathy or sort of psychotic empathy that sort of exists in this AI space about really it's around this question, is something better than nothing? Is it better for you to go to a genuinely sociopathic space where there is no care? And that's a good thing in comparison to what damage it can do to you from a human therapist or not.
[Note: Transcript covers the first portion of this episode. Full chunked transcription coming soon.]
[Note: Transcript covers the first portion of this episode. Full chunked transcription coming soon.]
