You, Me and Your AI Chatbot — How Therapists Are Navigating Chatbot Use — Ep 10

Are AI chatbots shifting the therapeutic landscape and how we relate to one another? A conversation with cyber psychology expert Rachel Wood.

Transcript

Rachel Wood 0:00

If she or any of us continually make a habit of offloading these types of decisions, there will be a complete void of building up self-trust, of listening to your own voice that actually leads to kind of a bankruptcy of internal discernment and wisdom.

Stephanie Hepburn 0:23

This is Crisis Talk. I'm your host, Stephanie Hepburn. Today, cyber psychology expert Rachel Wood joins me. We discuss how AI chatbots are shifting the therapeutic landscape and the way we relate to ourselves and others. Let's jump in.

Rachel Wood 0:40

My name is Rachel Wood. I have a PhD in cyber psychology. Cyberpsychology is really the scientific study of the digitally connected human experience. And so that's kind of a large umbrella, but my realm of expertise is in mental health and AI. I'm also a licensed therapist.

Stephanie Hepburn 1:02

So, Rachel, recently I interviewed Matthew Bergman, and he was part of the legal team that won a social media addiction case in Los Angeles against Meta and Google. And in our discussion, we talked about all kinds of things. We talked about the case, but there was this point where he said where these chatbots seek to imitate human relationships, they're very dangerous based on the nature of human cognition and linguistics. And interestingly, where they are simply providing information, they're less pernicious. And that got me thinking, you know, what is the purpose of large language models? Is it solely to be a resource? Because what's happening right now is it is a resource, but it's also mimicking human dialogue. And it's also mimicking companionship. And so I wanted to get your thoughts on that.

Rachel Wood 1:53

I think what's interesting right now and how people are really turning to chatbots and really ways that are not serving us particularly well is that the very skills that we need in order to build and maintain healthy relationships with humans are the very skills that do not get practiced when we are interacting with a chatbot. So it's a bit of a double bind, if you will, because some people are turning to chatbots as a way to maybe escape the friction of real relationships. So they're not facing the friction, but then in the midst of this kind of escape, they're actually having an erosion and an atrophy of the very skills that would make their interaction with humans better and stronger. And so it's this interesting thing where people are turning to it and then it's really not helping build what we need for connection with others.

Stephanie Hepburn 2:59

I've heard many psychologists use the term friction. I've even heard psychologists say you should be adding in friction. What does that mean exactly?

Rachel Wood 3:08

Well, what's interesting is, you know, life is full of friction, whether you try to bring it in or you try and avoid it. It's like you really can't be connected to someone for more than a little bit of time to notice that you're going to kind of rub shoulders in some way. And the thing is, we have this view that friction is maybe negative. Oh goodness, I'm having some friction with either a friend or a family member or a partner. And maybe that means that this relationship isn't good. Well, that's not true. Having a friction-free connection with someone is a total fantasy. And it's actually not even the success marker for good relationships. The success marker is how do you navigate the friction? So we have a different opinion on things or we have a different preference of how we want to do things. One person maybe hurt the feelings of the other. Whatever the friction is that comes up, the success marker is how are you going to move through that? How are you going to approach that? Some of the terminology we use is rupture and repair. So there's a rupture, which is a friction point. And it just means that there's something that has happened that has felt bumpy between both of you. And then you really build the strength and the bond of the relationship in the repair. Can we talk through it? Can we find our way back to each other?

Stephanie Hepburn 4:30

So one of my interviewees was Laura Riley. Her daughter died by suicide in 2025. And in our interview, she shared that the chatbots are enabling us to avoid human conflict. And she brought up COVID and youth who were going through large milestones, important milestones during the pandemic. And she felt like that was part of the dynamic. Maybe they missed out on some of these practices of addressing friction, like you're talking about. And so then it creates this reliance or this maybe opportunity for people to turn to uh the AI chatbots. And let's say you're having friction with your boss or with a friend. Instead of working through that on their own or with the individual where that friction exists, we're using this technology, an AI chatbot, to figure out the way to communicate, which on its face sounds okay. On its face, it's like, you know, when we're having a challenging time, whether it's with our boss or a friend or somebody that we're dating, we do often turn to our friends and we talk to them about, hey, this is what we're experiencing. What are your thoughts? And we get their feedback. Why is turning to an AI chatbot different than turning to a friend?

Rachel Wood 5:56

Well, let me first say that you're right that we've kind of had a long runway that has led us up to this point. When we think about kind of the influence of social media, then we think about, you know, the amplification of things with COVID. And then here we are with AI chatbots. And I think that it's really healthy for us to have a village and a community in which we process things. And we actually need more of that. We need more of these kind of revival of the third places where we have problems in our life, but we actually don't need to go to therapy. We need to go talk to a neighbor or a friend or meet people at the coffee shop. These places need a revival in society so that we can have these deep conversations. But because there's a void there, that is, you know, a bit of the rise of chatbots because people feel like they're not going to be judged when they talk to a chatbot, they're not going to be rejected, they're not going to feel shame. I mean, all of these fears that are encompassed in vulnerability, they really go by the wayside because you know you're not going to encounter that with a chatbot. So there's this interesting kind of safety that feels like is present. And that's why people are leaning into it. But when you get into that, and let's say it's, you know, turning to a chatbot is maybe your only person that you kind of person, quote unquote, your only place where you divulge, there's a number of issues that can arise. First of all, chatbots don't know context. They're very bad at knowing kind of the normal context that we know. So if I'm sitting even with a friend, I'm going to be asking questions, or, you know, if you're talking about your boss, it's like, well, what else is at play here? What are the other dynamics and factors? And a chatbot usually isn't going to lean into all of that or even know to kind of ask some of those questions. So the response or the advice that you're getting from a chatbot is going to be skewed. It's going to be very narrow. It's not going to be maybe asking what your part in the dynamic might be. There's all these different things. And then, of course, it tends to be pretty affirming and pretty sycophantic. So whatever you give it, it's going to mirror that back to you. And so it can amplify kind of these loops of your existing beliefs without helping you challenge those beliefs, without helping you maybe gain a different perspective about the narrative you have in your mind. And so there's a lot of limitations that actually won't help breed better outcomes in real life for you.

Stephanie Hepburn 8:44

So thinking about that recently, OpenAI pulled back ChatGPT for because it was too sycophantic. So it's no longer available. And so there does seem to be this push and pull with the AI companies because there are no federal regulations. These companies are internally creating these guardrails and they do change and they are evolving. But even though they're trying to address the sycophancy in my interactions with them, they're still quite sycophantic. What do you see as the role that these chatbots should or shouldn't be playing?

Rachel Wood 9:20

Well, I think the first step in this is what you mentioned, Stephanie, which is the sycophancy. For many of us, we are aware of that. We can observe it and see it and know it for what it is. But it's important to note that there are many people who don't notice that, who maybe don't see that, or maybe they have zero affirmation anywhere else in their life. And so this is really alluring for them to hear that maybe their ideas are great or whatever it is that they're hearing. And so there's two sides of the coin. First of all, needing to really bring in awareness campaigns for users. And then, of course, you know, talking about strategies and guardrails and safeguards on the design side. Yes, I think that AI should be a tool and not a relationship. I mean, I do firmly believe that. And yet, here we are. It should not be a replacement for a relationship. And yet the numbers and the statistics tell us that that is actually what's happening quite a bit. When we look at Microsoft's report from 2025, they have 33 million people who used Copilot. And the words that they use is that everybody's using it as a trusted companion. And that's not even like a companion app. That's just kind of a basic administrative chatbot. And yet, this is how people are using it. We truly do need to look at how do we address this? How do we solve this? And it's really complex.

Stephanie Hepburn 11:00

Yeah, I think that's accurate. And again, because this is evolving so very quickly, and also the usage, people are adopting it so very quickly, and not always by choice. What we're seeing is it's becoming so very pervasive that it would be very difficult to not have any sort of interaction with a large language model at this point. What happens for individuals as they're using it from your perspective? If you have a client that you're talking to, what would you tell them? Right now, I think it seems like increasingly psychologists would need to ask their client, are you using a large language model? What should that look like as this evolves so very quickly?

Rachel Wood 11:41

You're absolutely right. And I do a lot of training with mental health professionals. And that's exactly what I do recommend that in the same way we ask about social media use or different technology use. You know, therapists even ask just about general support systems. What kind of ecosystem of support do you have in your life? Well, chatbots are becoming part of that ecosystem of support for people. So it is important that we are bringing forth the conversation because sometimes people may feel ashamed or nervous to disclose that they are engaging with a chatbot, especially people who maybe it's their best friend or even their romantic partner. I mean, all of this is happening. And if we don't talk about it, then we clearly cannot find a way through it. So, yes, I do recommend that these conversations get brought up in non-judgmental, open ways, so that we can kind of essentially what we're doing is assessing for what role is the chatbot playing? What is the chatbot giving you that is void in your life? And then let's start there. You know, okay, is there a reason that the chatbot feels easier than going out and maybe trying to meet new people? Yes, of course there's a reason. How do we resource people to get their needs met in a way that they can kind of slowly remove leaning and dependence on the chatbot?

Stephanie Hepburn 13:20

You know, one thing that I've been thinking about is with providers, have there been any shifts in that area where maybe at the beginning when you did these trainings, it really wasn't as much of an issue? Something that the therapist should ask, but wasn't yet as pervasive. But as it's become more pervasive, have those questions changed? So for example, instead of it just being, are you using an AI chatbot to talk about personal issues? How often are you also turning to friends? Are you also turning to family? And then if you start noticing that that percentage is increasingly towards the chatbot, what can the therapist do? And do those from your side, from the training perspective, has that shifted over time? It has.

Rachel Wood 14:07

And it's shifting pretty rapidly. You know, every room that I'm in with therapists, I ask the same question, which is how many of you have clients who are bringing up AI in sessions? You're not talking about it, they're bringing it up. And 90 plus percent of the hands go up in the room consistently. I kind of do this informal poll to get a, you know, a pulse, a sentiment pulse. So it's being talked about, and there's a lot of nuance here with how that's happening. Because, like you said, just the question of are you using it or how often is not nearly sufficient enough because when you dive into different ways that people are using AI and particularly kind of different personas that the AI can take on. Maybe it's because somebody's using like a character AI, and so there's a specific avatar persona, or that somebody is using kind of a general purpose chat bot, but they have trained it in a certain way to react and respond to them. So really I'm saying you need to dive into what's the dynamic between the person and the chat bot. How does a chat bot act, you know, and what does it say and what is it reinforcing? So you have to go pretty deep into the layers of the dynamic itself. And, you know, another thing we see when I sit with therapists is I was just talking with a couple's therapist recently who she is working with a couple where it's two humans, but one of the partners has a chat bot that is a romantic partner. And so it's absolutely happening. And we're holding questions like is that infidelity? Is this some sort of polyamory? Is this a brand new category? What is this? And of course, there is no black and white answer there because every couple has to figure that out for themselves. However, it's pretty interesting to see the way that it is coming into the therapy space. And let me just say one more thing that we're seeing in this space, which is people who use AI somewhat heavily are becoming very accustomed to direct, short, advice-driven answers, because that's what AI provides. And so it's like a quick answer, it's a right-of-way. And so that's changing the therapeutic space because if you or anybody who's listening has ever been to therapy, you understand that's not the point of therapy. It's not like quick, short, direct advice, you know. But as people are getting used to that, they kind of come into the therapy office and they're expecting that. Like, well, you should have, you know, an answer or advice or direction for me when that is not at all what the purpose is. So that kind of very subtle shift is really going to be changing a lot down the road.

Stephanie Hepburn 17:09

That's so interesting because that's shifting expectation. It's that chatbots, I mean, therapy has been around for a long time, and that chatbots, while novel, are shaping what people's expectations are in a session.

Rachel Wood 17:24

Yes, it is. It's shaping expectations and sessions, but it's also shaping our expectations of other people. And as a little kind of interesting tidbit here, the first chatbot, which was built in 1966 by Joseph Weisenbaum, it was called Eliza. It was built off of Carl Rogers' work. And at the time, Carl Rogers was a very prominent therapist. So the first chatbot ever was really built around a therapeutic approach of what's called person-centered therapy. So that's kind of an interesting thing to see how far we've come. So, what did Eliza do exactly? So, Eliza, obviously very rudimentary in 1966, yet very alluring for people. It would respond back to whatever you said and just attune to you. So let's say you typed in, I'm having a really hard day. Eliza might say, You're having a hard day. Tell me more about that. Well, you know, this and this happened. Okay, how did you feel when this and this happened? Really acting as that mirror of attunement that sees you, hears you, validates you, which is not a bad thing. It's like we all need to be seen and heard and understood and validated. So it's like these things are fundamental core facets of humanity and they should not be squashed. It's important for us to have this. But are we doing it in a reciprocal way? When we turn to a machine, we don't have to give, we just receive. And the best way to have these relational needs met in the workplace or personally is to have it be reciprocal, where we have to practice giving as well as receiving.

Stephanie Hepburn 19:13

And also the idea of mirroring without challenging. It's one thing to validate somebody's feelings or to say, I hear where you're coming from, or be like Eliza and say it back to the person. But then without challenging, it just says what you're feeling and what you're experiencing is the only perspective in this situation.

Rachel Wood 19:37

Right. And that's not what relationships are. That's definitely not what therapy is. You know, the point is growth, and growth usually comes from having to look at some pretty hard truths.

Stephanie Hepburn 19:48

And so then, Rachel, what happens next? So you talk to the therapist, ask them how many of the people that they're seeing are using a chatbot, how often it's brought into conversation. And like you said, you need to go deeper to see how it's being used, whether it's being used as a resource, whether people are turning away from connection and turning to the chat bot instead. But then what what's the discussion with the therapists at that point?

Rachel Wood 20:12

Like any other thing that um a client might be turning to to distract themselves or assuage maybe some parts of their life or anything else that they might be turning to. And so what you do is you really you look at why the client is turning to the chatbot and what void it's filling. So the reason I do say that the chatbot is kind of singular in this is that most other vices, if you will, don't talk back. Right. You know, like they don't, like if there's maybe a problem with gambling, let's say someone has a problem with gambling. Yeah, there's kind of a reinforcement loop that happens if they, you know, gamble and they win some money and all this, and there's a cycle. But it's not the same as the pseudo-relationship that an AI offers. But you can still use all these tested, tried, and true modalities that therapists learn of let's first start with where the pain points are, why you're using this to maybe numb or distract or avoid other things in your life? You know, we can come down to really common core questions that you would really use for anything, which is this is doing something for you. What is it doing and why? And then once we figure that out, how do we then address maybe the wounds, the hurt parts that are needing to find an escape? And then as we work to heal those, really, a lot of that happens when you kind of fold back into community, you fold back into relationships. You practice the skills for connection. And it's hard work, you know. I mean, I'm not saying any of this is easy, but it is simple in terms of let's find the pain points, address those, and then you resource the client, not only with community, but their own internal resources for things like self-regulation. That's part of the approach.

Stephanie Hepburn 22:23

Because this is evolving so very quickly, and again, it's not just that the chatbots themselves, that large language models are evolving very quickly. They are, but it's also that more and more people are using them not only for work or as a tool, but also, like you mentioned, for personal reasons, whether that is to discuss mental health challenges, whether that's to discuss a relationship issue. You know, you mentioned speaking with mental health professionals about what to say or what to ask in a non-judgmental way so that people can share what they're using it for. Maybe they're using it to learn some CBT, some cognitive behavioral therapy techniques. Maybe they're saying, I'm dealing with anxiety. Can you list out some CBT skills that I can learn? So you can see how a large language model can be useful in that way as a resource. Using it for research purposes is a little bit different than let's say I'm dealing with this situation with the person that I'm dating, and this is what this person said, this is what I said. And almost turning to the chat bot as I don't know if mediator is the right word, but to get an objective opinion similar to what you would do with your friends, except maybe you're providing more transparency because even with friends and with family, people who you're close to, we share things, but we may not share all the details because we're worried about their perception either of us or the person that we're having the conflict with. And so you have this new introduction of using a large language model in that sort of way, which is to me very different than just as a resource.

Rachel Wood 24:13

Absolutely. And it it is not a glorified Google. And, you know, there's studies done right now on kind of this perception gap, particularly between parents and their kids, of how parents think it's being used, like glorified Google, and how it's actually being used. And some of recent research that I've been reading is that actually, whenever AI comes up for kids, for youth, it's always about cheating. It's always talking about the conversation of don't use AI to cheat. But rarely is the conversation, don't turn to AI as a boyfriend or a girlfriend, you know, or your best friend. These kinds of conversations are actually what's going to be shifting the relational bedrock of society, is how it's weaving its way into how we relate to ourselves and how we relate to one another. So can it be used as a tool? Yes, it can. Think about it as a reflective journal or to do some guided meditation. Even think about it as a way to practice role play. If you have a difficult conversation on the horizon, maybe with someone at work or someone at home, you could put that in and try and do a role play back and forth with the chat bot to rehearse that difficult conversation so you feel more prepared for it. So there's definitely ways to be using it that can support and help us. But there's a large caveat there, which is it takes a lot of user knowledge and awareness to keep it as a tool and also to understand that it doesn't have the context to really know the full picture of your life. And even friends, even if they are mostly leaning on our side and they're for us, they still, if we're telling them a one-sided story, they still probably understand that there's more context to the story, which AI doesn't have that. So there's ways to use it that can help, but there's also a lot of pull to use it in ways that replace instead of rehearse relationship.

Stephanie Hepburn 26:29

I like the idea that you're mentioning, talking about role play, but when does that slip into some sort of dependence on the chatbot for basically telling you how to navigate your own life? What are the beginning and end of those boundaries? Would it end up being one of these things that you just become increasingly dependent on?

Rachel Wood 26:52

Well, let me give my thoughts by sharing a brief study, which was looking at university students. And they were asking about how university students are using AI. And of course, this is not all of them. So I'm not trying to paint this as the majority, but one of the participants said that she was walking to campus and she came across someone on the street who did not have a home. And she paused, turned to her chat bot, and asked the AI whether she should give this person money or not. Now, what's so fascinating about this scenario is that this is a moral, philosophical, some might even say a religious question. This is a question that doesn't really have a black or white answer. There's really no right or wrong. It is something deeply personal, whether you want to give or not. And the other thing is it's a highly contextual question, which is well, is it nighttime? Is she alone? You know, are there other people around? Like there's a huge context that the AI does not know. Now, I don't know what the chat bot said in response, but what's more important about this scenario is if she or any of us continually make a habit of offloading these types of decisions, there will be a complete void of building up self-trust, of listening to your own voice, of building up even self-confidence. And those are the things that we need to navigate the world, to kind of carve our own path in the world. You're kind of saying, where is the line here? And it is very hard to tell because it's a bit of a spectrum and not a line. You know, can you use a chatbot but then walk away from it? Or are you experiencing distress if you can't turn to the chatbot? Can you make decisions without the chatbot? Or do you need it to be your sounding board or give you direct advice? So there's really a spectrum here, and there's a lot of contextual factors of what other resources are in your life, what are you going through in your life right now? But if we lose our own self-trust and the development of our own discernment and wisdom, things in society are going to look very, very different within five, 10, 15, 20 plus years.

Stephanie Hepburn 29:30

That was Rachel Wood. She's a cyberpsychology expert and the founder of the AI Mental Health Collective, an online community to help mental health professionals navigate the intersection of AI and mental health. I'll put the link in the show notes. If you enjoyed this episode, please subscribe and leave us a review wherever you listen to the podcast. It helps others find the show. Thanks for listening. I'm your host and producer, our associate producer is Rin Koenig, Audio Engineering by Chris Mann. Music is Final Couch by Blue Dot Sessions.