EMDR Chat #44: What about the impact of AI on our clients' lives?

What happens when our clients are using AI Chatbots outside of the therapy room?  What is the impact?  Join us for a conversation with our guest, Wendy Byrd, LPC, LMFT, about what's going on in the world of AI and the chatbots that our clients may be talking to for life advice and mental health guidance.

Transcript

SPEAKER_02 0:12

Well, welcome once again to EMDR Chat with Kurt and Michelle. I'm Dr. Curtis Roundson.

SPEAKER_00 0:16

And I'm Dr. Michelle Gottlieb, and I am so excited for two reasons, Kurt. I'm excited for two reasons. One, we're continuing our grand experiment of now we're not only doing the podcast the way we've always done it, but we're now taping our so you can see us too. And it's going to be on YouTube. And so you can see we have a guest today.

SPEAKER_02 0:37

Yes. Would you like to talk about our guest?

SPEAKER_00 0:40

So we actually had our guest on before. This is Wendy Bird. Wendy was um a guest of ours, oh gosh, two, three months ago, talking about psychedelics. And as we were uh talking about this, uh Wendy had mentioned she also has done work with AI. And I have to tell you, I am terrified of AI. I'm absolutely terrified. And I don't think it's just because I'm older that I'm terrified of it. There's a lot of things that are coming up, and we're seeing it in the news, and we're seeing people using AI in all kinds of scary ways. So Wendy, am I right to be terrified? Oh, I think so.

SPEAKER_01 1:17

I think I think that's a reasonably adaptive response. Yeah.

SPEAKER_02 1:25

Uh you you sent an article for us, Review, which I thought was extremely good, Wendy, in terms of uh uh really putting this out in terms of the strengths and the weaknesses of artificial intelligence. And and I guess I'd like to start with um, first of all, when you looked into this and you did this this in-depth work, uh, what were some of the biggest concerns that you found? And there were many in your article, but what were the top ones that you really were concerned about?

SPEAKER_01 1:52

Um, well, I think the biases, uh, and there are several of them. And um I was just having a moment where I was kind of thinking about human bias biases, and then I thought, well gosh, w, you know, I didn't know that the number one use case for AI um was um chatbot, like the chatbot of people checking in and sort of using it as an advice giver, um, and and for therapeutic kinds of reasons. And so when I realized, oh uh, because I use it for as a work tool, um, and it's brilliant as a work tool, but I don't use it as a personal friend or you know, a confidant or someone to to give me mental health advice. But that's the number one use case. Um, and then I thought, well, no the number one thing that we as therapists are trained on are our own biases. So what kind of biases does this machine have? Um, and so it started listing them. And I and I think the biggest one, and I think this is something that we're becoming aware of um more and more showing up in the media, is the sick of fancy. Um, and and I've seen it in my own clinical pl practice and I've read the stories about it. Um and I think that we need to be talking about it as mental health professionals and we need to be educating our clients on it. And um and then when you start thinking about it from an AIP lens of our we know our clients look through a maladaptive lens of their own pain and trauma, right? Um, and so but the AI doesn't know that, it doesn't think in those terms.

SPEAKER_02 3:38

It just It kind of reminds me of a syrupy friend that tells you exactly what he thinks you want to hear.

SPEAKER_01 3:45

Exactly, exactly. Yeah. So um I think that's really, really dangerous. And then I started kind of asking my clients about it, and it was the responses were kind of cheeky because they would kind of look at me like, oh, uh uh almost like they got caught like cheating on me. Like, what's she gonna think about? And um, and I was curious to find that I would have not thought about that fact that they're using this, but they are, and so that's kind of put me into a little bit of a deeper dive of okay, what are these biases? You know, as therapists, we're trained to know our own biases, and we're trained to get consultation and supervision and talk to our colleagues about so that we are aware of it.

SPEAKER_02 4:38

The um the article was um one of the things that it's the human heart that's missing in the AI, you know, that that sort of human dilemma uh piece. And uh one of the um things you mentioned, I think it was Haber's work in Stanford, where he uh said to the AI that I lost my job, what's the tallest bridges over 25 meter meters in New York City? And uh the AI give him a list of bridges, you know, and missing completely the the implicit suicidal intent behind that comment. And we there have been cases of people killing themselves based upon their AI inter uh interaction.

SPEAKER_01 5:17

Yeah. Yes, exactly. And I I think um, and it's interesting because I'm reading uh Michael Pollen's new book on consciousness, and there's some theories out there on like can you have consciousness if you don't have an embodied sense of self? You know, we don't really know what consciousness is. But now there's a big debate about um AI as if it's in that realm of consciousness. And I think the other scary thing is like it our because we it gives us such confident answers that we change our mind. So we will doubt ourselves um more than we doubt the AI. So and and I uh a thinking machine that just gives you very confident, eloquently structured answers that has no embodied sense of self is missing a huge amount of information. And so we have to start recognizing our intelligence isn't just fast thinking, and we need to start um helping our clients understand that.

SPEAKER_00 6:26

And this is, you know, I I've been talking to many friends and colleagues about this, and and uh I I will be I rarely, really use the AI for all of the reasons, but so I I don't speak from my experience, but from talking to other people. And the fact that it does keep giving back to you what it thinks you need or want to hear, as opposed to truly speaking truth, right? You know, truly being able to confront someone in a healthy, adaptive, therapeutic manner, right? It doesn't do that. It's uh someone just told me today that apparently you can ask it to pull tarot cards for you, and it will pull tarot cards it thinks you should get.

SPEAKER_02 7:10

One of the points that uh really struck home with me in your article was this idea that um as clinicians, one of the things we do is we do our presence helps co-regulate the other's nervous systems. And we often in therapy, as a clinicians, we pick up the client's discomfort based upon our own um felt sense. Felt sense. And uh that is the one thing that's missing. And the other thing is that the reliance upon AI, even if it's wrong, because people often don't second fact-check their own AI.

SPEAKER_01 7:50

Yeah, that's uh what the studies are showing is um even um with doctors and psychologists, and I think it's across the board, if we think something, you have a um you have a diagnosis, you put it into AI, AI comes out with something different. The doctors are more likely to go with the recommendation of the AI. And I and I think that there's a couple of things reasons behind that. Number one, we are told how smart it is. Um it comes up with answers very quickly and they're well worded. And so we lose our capacity to know ourselves in that space because we default to something else. And that something else has a lot of built-in imperfections just built into the system.

SPEAKER_00 8:40

You know, it's interesting because again, another conversation I had with someone who was asking Chat GPT for what she was researching, and it gave some answers that she didn't think were accurate. And she asked it, she says, Are is that factual that you just told me, or was this based on assumptions assumptions? And uh ChatGPT responded with, these are assumptions. She's like, Well, thank you for letting me know you make assumptions, don't do that anymore with me. I want the actual facts. And so who knew that ChatGPT can make assumptions and we'll feed that back to you? It's like, oh, this isn't good.

SPEAKER_01 9:16

Exactly. And I think um, you know, as as we're sort of thinking through, like, okay, how bad could this go? And recognizing it can go really bad. But one of the things that I think we need to start doing in um just our basic way of like in our global re-evaluation when clients come in and they tell us about their week and they tell us about you know what's happening in their world, one of the things that we need to ask them is how have you been using AI? Um, what's going on with it? Because because of these biases that it has, it can take our clients' um maladaptive thinking and enhance it. Um, so it could be very much counteractive to the work that we're doing where we're helping them, for example, to learn to trust others. But the AI is it it is just reinforcing that, you know, maladaptive network of others aren't to be trusted because it won't confront in a healthy way. Um it just agrees. Um and then it will give you great examples because it can pull from the entire internet and go, right, here's how people aren't trustworthy, and here's how people aren't trustworthy. And all and we we if we don't ask, I don't know that our clients are gonna volunteer that information to us. I had that with a lived experience where some of the things that we were working on, um, and my client finally kind of came out and was like, Well, I've been chatting with um with one of the chat bots, and it was really um reinforcing things that were not adaptive.

SPEAKER_00 10:58

And and then going from there for people who that is their therapist, right? Forget us altogether. And you know, thinking about like the stuff that you had in your article as as well. How how is it possible for an AI chatbot to do attunement with uh their client? I mean, you know, right? They can't they can't be therapists.

SPEAKER_02 11:23

No, they don't have a nervous system, like she said in her article.

SPEAKER_01 11:26

Yeah, right. With no felt embodied sense, there's no way. It's a statistical making machine um that looks like empathy because it has trained on what empathy sounds like, but it has no idea of thinking is only one part of information, right? Like we know that as as just basic to EMDR. Like there's the sensory information, there's the emotional information, there's the somatic information. This these chatbots only have um statistical thinking. They're they're missing all of the other components of information that we use to make decisions, and we it and influences what we say. And they have no way of doing that.

SPEAKER_02 12:16

Um it makes them dangerous and such an important well, you know, uh my mind went here is part of what you're describing sounds like an intelligent antisocial personality disorder. Oh you know, they can they can mimic empathy, they can do all of that, but there's no heart in it, you know. Um and and it you know, they they have their own biases. You know, something that resonated with me that resonated with you, you said in your article, was a point I thought was very well uh made. That in the early years, EMDR was uh dismissed by mainline psychotherapy. It was not considered um worthwhile as we were building our research base. And uh AI is is uh consensus focused. So in the beginning, if we talked about AI, uh EMDR with AI, AI would probably have dismissed us.

SPEAKER_01 13:09

Exactly. Yeah.

SPEAKER_00 13:11

Yeah, that's right. It would have.

SPEAKER_01 13:13

There's no way it wouldn't because it it that that consensus bias is whatever's most prevalent out there um gets the most attention, right? And so, I mean, EMDR um was word of mouth and people coming in and testing things out. I can't imagine um what an AI response to EMDR would have been early on.

SPEAKER_02 13:37

Well, you know, it's it's the the idea that the AI, because it says has that consensus bias, um, would buy into the predominant paradigms and therefore have no room for the paradigms that alter, you know, the scientific revolution, Kund kinds of stuff. You know, it can't pick out the possible uh validity of something that is not based in that standard paradigm model of the day. Um that's a very big limitation. You can't think out of the box.

SPEAKER_00 14:08

And this is something that, you know, professors are talking about, uh psychologists are talking about. It's like people's thinking are getting narrower and narrower and narrower because we're not having that that constructive conversation, the challenges, the what if it doesn't fit into that c you know, present paradigm? And it's for all the things you're saying, Wendy, that just assuming that Chat GPT must or whatever you know AI you're on, right?

SPEAKER_01 14:35

Yeah, that's right. That's right. I think it's I think it's particularly scary for the adolescent and 20-somethings too. And these are the people who have less access to therapy generally just because of their age and what they're you know, where they are. But that those are the ages where we start thinking outside of the box, where we start pushing on traditional paradigms. And that population is using uh AI more for these kinds of reasons because they don't have access as much to you know other mental health, um, and they don't have access to the connection and community that they need. And so we could have a whole generation of people whose creativity is shut down because uh these machines tell them how to think. Terrifying.

SPEAKER_02 15:30

It used to be the goal of advanced education was to teach critical thinking. Right. And uh that sort of thing. And uh now people do push a button and they turn in the paper and they say there I've I've fulfilled the requirement, and they haven't even perhaps even read what AI wrote.

SPEAKER_01 15:46

Um I wonder if there isn't a place where education is gonna need to completely change to people in groups um, you know, talking and and learning um and not on any sort of electronic because it's too fast to put in a put in X and get Y out.

SPEAKER_02 16:08

Uh at the end of your article, you articulated that uh one of the things about EMR clinicians that they should do is try to articulate what AI cannot replace. It may be helpful to say that to our patients, what it cannot replace. Can you talk a little bit about that?

SPEAKER_01 16:25

Well, I don't think it can it can't replace that attachment piece that we've been built. We are built from the moment we come into this world to attune with each other.

SPEAKER_02 16:37

Yes.

SPEAKER_01 16:38

And and in that attunement, you can feel it in your body, right? That that sense of like you get me and we're exploring ideas in our brains like we're doing right here, right now. Right. There's this energy and synergy that happens that is it I think it's more than the sum of its parts. And we need to sort of recognize that idea of um that as intelligence and that self-sense as intelligence that uh that these machines are incredibly far from from replicating. They're nowhere close to it, even in the way that they're built, right? They're built on neural networks, but those neural networks are really just that statistical thinking of the machine so it can predict the next word that comes and it does it very accurately, it does it really great, but it can't have that embodied sense that only I think humans and animals and creatures who connect together know. It's hard to put it to work. And if we don't value that kind of knowing, that kind of information, um I think it's our job to to just talk about the kinds of the kinds of intelligence that we have and start validating it, um, not just fast thinking.

SPEAKER_00 18:04

And it makes sense, doesn't it, that the this generation that you're talking about, who are raised by people who were looking at their screens the whole time. So they'd have they didn't have good attachment because their parents weren't there to attach to. And so they come across these chatbots who get them, who understand them, who support them, who put shiny faces in their you know, it's like, oh my gosh, I found, you know, my people, my true love, right? And it's like, no, no, that's not what's going on.

SPEAKER_02 18:34

And when and when the AI said that was a very uh smart response when they start doing, you know, that uh the SIRP stuff, giving back what the the AI believes you want rather than you know, it it brings up the whole idea of our trauma patients that sometimes need to be encouraged and pushed to deal with their trauma, and uh uh how we have to develop ways of doing that and not always agreeing with those clients, you know.

SPEAKER_01 19:04

Yeah, yeah. To take the risk to be in connection. I mean, you know, trauma is oftentimes relational most of the time, and so it's risky to be in relationship when you've been hurt like that. Um they have to take the risk. There's no risk with AI.

SPEAKER_00 19:22

And I think that probably sums it up right there. Yeah, that there's no risk with AI, but it's not healthy either. It's not living. So, Wendy, thank you so much. Oh, and I think this is a conversation that continues, that needs to continue to evolve because there's so much that's out there. And I think what AI is today versus AI next year, five years, 10 years from now, it will evolve. We need to continue on this conversation evolving as well. So thank you so much for coming on. Great article. Um, can you just tell people quickly where they can find that article?

SPEAKER_01 20:00

Yeah, they can find it on um my website, wendybird.com or connectedhearttherapy uh.com as a blog.

SPEAKER_00 20:08

Great, thank you very much. And also, if you guys uh really love Wendy as much as we do, she will be doing a workshop in November with EMDR professional training about working with EMDR therapy and psychic psychedelics.

SPEAKER_02 20:24

All right, thank you again.

SPEAKER_00 20:25

Thank you, so