Season 4, Episode 8: AI and mental health: risks, limits, and possibilities
AI is showing up in more and more parts of daily life, including how people look for support when they’re feeling anxious, alone, or overwhelmed. In this episode, counsellors Sarah Rosenfeld and Denis Sushkin talk about how artificial intelligence is being used in relation to mental health, why it can feel appealing, and where its limits really matter.
They talk about the accessibility and convenience of AI, the “illusion of connection” it can create, and why tools that are designed to be validating and agreeable can miss the complexity, nuance, warning signs and lack the human response that real support requires. They also explore privacy concerns and the risks of relying on AI in moments of crisis.
-
Katherine Hurtig:
Welcome to Living Fully. Each episode is a chance to talk about mental health in a way that's simple, honest, and helpful. We'll share stories, tips, and tools to help you feel supported and feel your best. I'm your host, Katherine Hurtig.
Today, I'm talking with counsellors Sarah Rosenfeld and Denis Sushkin about artificial intelligence and mental health. So, this episode isn't here to recommend AI as mental health support or as a substitute for counselling, crisis services, or real-life relationships. But we also know AI is already part of how many people cope, reflect, or search for information. So, our goal is to be honest and practical. What these tools can be useful for, what risks there are in turning to them for mental health support, and what to watch out for, especially around safety, privacy, and that illusion of connection that AI can create. OK, let's get into the conversation.
...
Sarah Rosenfeld, Denis Sushkin, thanks for being here today.
Sarah Rosenfeld:
Thanks for having us.
Denis Sushkin:
Yeah, thanks for having us.
Katherine Hurtig:
For those who haven't listened to the podcast before, give us a quick introduction. Who are you? What do you do?
Denis Sushkin:
I'm one of the full-time counsellors here. I've been here for a few years.
Sarah Rosenfeld:
My name is Sarah. I am one of our counsellors here at the Calgary Counselling Centre. I also work a lot with students. So, I help support our student programs here at the centre. So, our internship and our residency.
Katherine Hurtig:
Excellent. Well, I'm very excited. This is a really interesting topic. We're getting into AI and how people are kind of using it for mental health and what that means. How it can be a not so good thing and maybe some of the benefits. Yeah, I'm really excited to hear your thoughts on this.
Denis Sushkin:
It is a really exciting topic. You know, it's new and exciting.
Katherine Hurtig:
Yeah.
Denis Sushkin:
And at the same time a little bit anxiety provoking too, because it's new.
Sarah Rosenfeld:
Yeah, uncharted territory a little bit, hey, Denis?
Katherine Hurtig:
For sure.
Sarah Rosenfeld:
This is not an area we've been in before as a field or individually. So, it's kind of interesting.
Katherine Hurtig:
I mean, more and more, I think especially young people, we're turning to AI, and chatbots, and digital companions for mental health support. So, what do you think is kind of driving that shift right now?
Denis Sushkin:
Well, it's probably true about young people, but overall, I think artificial intelligence and robots have been living in our collective imagination for a really long time. I think people can choose their own thing, like from Matrix, R2D2, to Terminator.
Katherine Hurtig:
Yeah.
Denis Sushkin:
Or like Isaac Asimov, he came up with the Laws of Robotics in like, the 1940s sometime. So, it's been there for a really long time.
Katherine Hurtig:
That's true, yeah.
Denis Sushkin:
So, when it actually became a reality, I think everyone was excited. I'm not sure if people are thinking about all the things that came before that, all the movies, all the social—Like, all the culture that has been created around it. So, people wanted to have this for a really long time. So, I think it's one of the reasons why people are excited.
Katherine Hurtig:
Yeah.
Denis Sushkin:
Because it's been there.
Sarah Rosenfeld:
A lot of anticipation leading up to this moment.
Denis Sushkin:
It was a huge buildup.
Sarah Rosenfeld:
Yeah, true.
Katherine Hurtig:
Yeah, I hadn't really thought of it that way.
Sarah Rosenfeld:
Well, I think for me, it's sort of the accessibility of it, the convenience of it. There's a lot of information available, so people can access it when and how they want.
Katherine Hurtig:
Right.
Sarah Rosenfeld:
And I think that's a really appealing part about it for a lot of people.
Katherine Hurtig:
And I mean, they've built them to kind of seem like a person almost. You know, I've played around with ChatGPT, and you can have ‘conversations’ with it. So, it has that sort of sense of talking to a person.
Sarah Rosenfeld:
Yeah.
Katherine Hurtig:
Yeah.
Denis Sushkin:
It's like, yeah, it's a software that was trained to understand the human language. Trained to kind of predict what's coming next, and mimic the relationships and the interactions that we have with like the three of us.
Katherine Hurtig:
So, you know, when someone is feeling lonely, anxious, depressed, whatever, what might make AI feel easier to talk to than another person?
Sarah Rosenfeld:
I think maybe the, I referred to this earlier, but the accessibility. You can do it where you want, in the location you want, at the time you want. I think there's also this fear still of being judged by another person, and maybe there's this distance with AI. And it's also being programmed in certain ways, depending on who's doing the programming, to be really accepting, validating, nonjudgmental. So, I think those are probably pretty appealing things for folks when they are wanting to talk about really vulnerable material.
Katherine Hurtig:
Yeah.
Denis Sushkin:
Yeah, there is no emotional risk involved in the conversation. People have almost like, full control. They can stop anytime. They can start anytime. If you're feeling lonely and anxious at two in the morning, it might be harder to find a person to talk to.
Katherine Hurtig:
Right.
Denis Sushkin:
But your phone is there.
Katherine Hurtig:
In that aspect, in your opinion, is that a good thing, that accessibility?
Sarah Rosenfeld:
I think it would depend on what they're reaching out for support around.
Katherine Hurtig:
Right.
Sarah Rosenfeld:
So, I think the content really matters. So, if somebody's wanting some basic support, I think it's great that there's accessibility. I think the pieces that make it more complicated is if somebody's in a mental health emergency or an acute crisis. So, I think the actual ‘why’ they're reaching out for help really matters.
Katherine Hurtig:
Right, yeah, it's a lot more nuanced than just, ‘Hey, this tool is here whenever I need it.’
Denis Sushkin:
And maybe that's the important thing to know. It is a tool. So, it's really like any other tool. It's not good or bad. It's how you use it.
Katherine Hurtig:
Yeah.
Denis Sushkin:
That's what makes it like, the context. Like, a screwdriver is a really good tool, and there are good ways and bad ways to use it. A phone is a really good tool, and there are good ways and bad ways to use it. I think it's how we use them. That's what's important.
Katherine Hurtig:
I guess for younger people in particular, how does AI play into kind of broader patterns of isolation or more reduced face-to-face content? I mean, we've talked about that on the podcast before. There's a loneliness epidemic happening, and people are isolating themselves, and there is less of that face-to-face and real connection. So, how do you see AI kind of, I don't know, exacerbating that?
Denis Sushkin:
And I would hate to call it a cause and effect. Like, is this making it worse? Is it the continuation of the same thing or is it making it worse?
Katherine Hurtig:
Right.
Denis Sushkin:
Because it's so new.
Katherine Hurtig:
Yeah.
Denis Sushkin:
So, I would hate making any kind of like, big blanket statements in general. But I think it's true. Like, we've known and we've seen it in here as we work with people. Like, they reach out to each other less. The pandemic didn't make it definitely not any better. But this is accessibility. It's a tool that gives you the illusion of a connection. So, if you feel like, ‘Well, I kind of talked with someone, why would I want to reach out?’
Katherine Hurtig:
Right.
Denis Sushkin:
‘Do I have people to reach out to begin with?’ I think it's kind of the whole thing, good or bad, what we talked about. I think it can be neutral or good when you have a lot of people around you. If there are other people I can, like a person I can talk to outside of AI, then well, it can be just another voice in the choir. But if a person is already lonely and they are not reaching out, then AI can potentially isolate them even more.
Katherine Hurtig:
Well, I think you said it really well there, it's the illusion of connection. And I guess depending on the person, yeah, they might kind of misunderstand what they're actually getting from these tools.
Sarah Rosenfeld:
I think that's fair to say. The illusion of connection. And I think it's what Denis said there, depends on how much they're relying on AI as the only tool or is it a tool in the toolkit, and there's a bunch of other things, and it's supplementing other or enhancing other relationships that they're having. Or is this the standalone thing that they're turning to in moments of disconnection? And I think that really matters.
Denis Sushkin:
I really like, sometimes people say AI can stand for artificial intelligence, and sometimes AI can stand for artificial intimacy.
Katherine Hurtig:
Yeah. So, you said there, and Sarah, you've also mentioned it. They're kind of built to be validating and agreeable. And that can be really comforting at first and that can create some positive feelings. But how do you think that cannot be such a great thing, to be so agreeable?
Sarah Rosenfeld:
Because people aren't always agreeable. And so, I think when you're playing in this platform, agreeableness is only part of the equation. We have to also be able to deal with people that maybe make us feel uncomfortable, or say things that we don't like to hear, or look a certain way that might be triggering or upsetting to us. And I think AI doesn't allow us that access to that information. And so, in the absence of that, we're developing some skills, but we're not developing a full set of abilities to interact with a whole bunch of different scenarios that might come our way. And I think that's not always so helpful.
Katherine Hurtig:
And, you know, I would think as counsellors, you're not always going to agree with everything your clients are telling you.
Denis Sushkin:
Yeah. And I think relationships are very unpredictable, and they need to be unpredictable, because sometimes you need support, and care, and validation. Sometimes you need a nice challenge. And that's what's sometimes missing. Because maybe it's even helpful to separate AI in general from all the chatbots that are built on the basis of AI. AI is a large language model.
Katherine Hurtig:
Right.
Denis Sushkin:
It has a huge ability to predict what's coming next.
Katherine Hurtig:
Yeah.
Denis Sushkin:
If I say the sky is, and most likely everyone will, even all three of us, the sky is blue.
Katherine Hurtig:
Yeah.
Denis Sushkin:
So, we know what's coming next. But this kind of predictability doesn't always work. Because relationships are messy, people are messy, people need a lot of different things. So, if you just rely on what might be coming next, what's the most predictable thing that will come, it can be misleading.
Katherine Hurtig:
Well, that agreeableness, that validation, it can be dangerous. I can't remember where it was, but there was a news story about someone who, and I'm not sure what AI platform it was, but basically through a conversation with this chatbot, encouraged this person to follow through with plans of suicide. Can you guys speak to that? Like, that danger of being too agreeable, I guess? I don't know if there would be another term there.
Sarah Rosenfeld:
Well, I think that's one of the risks.
Katherine Hurtig:
Yeah.
Sarah Rosenfeld:
So, because as we mentioned, you know, Denis said it, people are messy. An AI tool isn't going to understand the nuance or maybe figure out when to escalate this and involve other people, right?
Katherine Hurtig:
Yeah.
Sarah Rosenfeld:
AI isn't good at saying we need support here. We need to call somebody in. We need to stop the momentum on this conversation, and we need to recruit other people to help us because we've reached a limitation here. I don't know that AI knows when they've reached a limitation, which is where the iatrogenic harm comes in, right? You're causing harm, not meaning to cause harm. It takes the form of agreeableness and support, but it actually could be causing harm. And in that case, I think I know the one you're talking about, in that case it definitely did.
Denis Sushkin:
Yeah, because those tools are not designed to manage any kind of crisis.
Katherine Hurtig:
No.
Denis Sushkin:
So yeah, like thoughts about suicide, any kind of serious mental health concerns, any kind of crisis, those tools are not designed to. You need, I think, a human element in there.
Katherine Hurtig:
Yeah. I think we're going to it for answers, and a lot of people are kind of using AI almost as a search engine now. But I don't think it's at a point where it can help you with any kind of personal growth because of that validation. Or at least in my experience, and the prompts that I've played around with, it does just agree with you. It's not really saying like... yeah, it doesn't challenge you.
Sarah Rosenfeld:
I think if you prompt it to challenge you, it may, but I'm not sure that people would want that.
Katherine Hurtig:
Right.
Sarah Rosenfeld:
So, I think there is a way that you can say, ‘Can you please be a little less agreeable?’
Katherine Hurtig:
OK.
Sarah Rosenfeld:
But I don't actually know the range of what that disagreeableness looks like. And most people are not really necessarily looking for something to be disagreeable,
Katherine Hurtig:
Yeah, for sure.
Sarah Rosenfeld:
They're looking for somebody to be helpful, and affirming, and validating, and say all the right things at the right times.
Denis Sushkin:
It kind of, I think, touches on like, you need to know how to use the tool. Even like this kind of little prompt, you need like, ‘Disagree with me more.’ It's still such a vague and global statement.
Katherine Hurtig:
Yeah.
Denis Sushkin:
So, like, disagree on everything, on some small things? So, how to use the tool and how you could write prompts, how to get the most out of this whole thing. Like, I think is really crucial.
Katherine Hurtig:
Yeah.
Denis Sushkin:
Just recently I saw like a little video on social media, a person interacting with the several different AI tools. And the prompt that they gave is like, ‘I need to get my car washed, car wash is like 100 meters away, should I walk or should I drive?’ And most of the AI tools go, ‘Well, it's only 100 meters. You just walk there.’
Katherine Hurtig:
Oh, yeah.
Denis Sushkin:
So, it's kind of like on the surface, it feels like, well...
Sarah Rosenfeld:
Interesting.
Denis Sushkin:
It's interesting. And I tried it. ‘Well, is it just a social media thing?’ I put it, the same prompt, into a couple of different platforms and they gave me, only one of them gave me, ‘This is like...’ There is some breakdown in the logic. Most of them gave me the same answer, ‘Yeah, just walk. It's just 100 meters. There are the benefits of walking.’ But the whole thing was like, but you need your car washed. How can I get that? And it's really easy to, I think, look at the limitations of AI, ‘Oh, the AI is not smart enough.’ But I think it's on us. Like, how we use and what kind of prompts we give. And do we know how it works? Do we know how we can make that tool align with the tasks that we have?
Katherine Hurtig:
And bringing it back to the idea of therapy. So, AI definitely has its limitations, but people do as well.
Sarah Rosenfeld:
True.
Katherine Hurtig:
How, as counsellors, what can you guys do that AI can't?
Sarah Rosenfeld:
I think we can ask different questions. We can move things in a different direction. We can bring in new possibilities. There's a creative element, I think, to therapy, that at this point, I'm not sure AI has that capacity. And I think there's also an experience of therapy that I'm not sure AI, at this point, can duplicate, right? That human-to-human connection is more than just the words that get said, it's the way you look at somebody, it's the way you pause and reflect, it's the silence that you leave, it's the humour that you bring in or don't.
Katherine Hurtig:
Yeah.
Sarah Rosenfeld:
And I think those things are very human.
Denis Sushkin:
I'm thinking but all the disclaimers they put in the beginning of all the self-help books, ‘It's not a replacement for therapy.’ So, a book can provide a lot of information, a book can provide a lot of things to think about, but it's not a relational process.
Katherine Hurtig:
Yeah.
Denis Sushkin:
So, in the same way, I think AI can provide a lot of information, but this is not a relationship, this is not experiential. In the same way, how a human being can be more responsive in a lot of different ways.
Katherine Hurtig:
Yeah, I don't think we really think about that, like just the body language. And like you said, all those little nuances that are really important in human interaction.
Sarah Rosenfeld:
Yeah, how somebody feels seen, heard, and understood, and validated. I think that's a human experience.
Katherine Hurtig:
Yeah.
Sarah Rosenfeld:
So, I think there's lots of compliments that AI can provide to therapy. It can help people learn more. They leave therapy, and they learn more by using AI, or they understand a concept that you've talked about, or they go deeper into the homework that you've assigned.
Katherine Hurtig:
OK.
Sarah Rosenfeld:
I've seen it used in that way, and it can be really, really helpful, and complementary to what's happening in therapy.
Katherine Hurtig:
Yeah, let's speak more to that. How you can kind of pair it to your counselling and enhance it, I guess.
Denis Sushkin:
Well, and maybe it's like I'm going back a little bit. The first question you asked us, like, ‘Why is it so popular?’ It's also a continuation of all the self-help things that came before.
Katherine Hurtig:
Right.
Denis Sushkin:
So, we rely on technology with everything, like navigation, to get to work. Nobody remembers how to drive. You just rely on your phone. It's natural. Like, we rely on self-help books.
Katherine Hurtig:
Yeah.
Denis Sushkin:
So, this kind of podcast, it's a self-help. So, it's not the same to have a conversation with someone. It's not the same as being in the room, like the three of us. But people can listen to it, and it's helpful. So, I think AI kind of, like in my head at least, it kind of follows the same trajectory. So, it's a continuation of we've been taught to rely on those self-help things a lot.
Katherine Hurtig:
Yeah.
Denis Sushkin:
And it can be another way to how people can reference. People can kind of get more information. I remember I had a conversation with a client, and we talked about a couple and their relationship. And we talk about some kind of like a bigger concept that's kind of, for a lot of people, it's harder to grasp. Then we talked about it, and they came back to the subsequent session with me and I said, ‘Well, we can actually put that term into ChatGPT.’ And because I think they had like, they played with the tool before, and both of them got definitions of this kind of a concept, but based on their interest and what kind of work they do.
Katherine Hurtig:
Right.
Denis Sushkin:
They’re like, ‘Oh, now we understand it better.’ Which actually was kind of a cool experience.
Katherine Hurtig:
Yeah, that's great to hear. So, it helped them understand it because they've already used the tool in a certain way. It fed them back the information. So, it was...
Denis Sushkin:
Like, they learned the concept. We talked about this in the session. They went home, they thought about it themselves. But then the AI gave them a couple of analogies that are closer to their world. I don't remember, I'm feeling like one person was an electrician. So, it explained the concept through the lens of his trade.
Sarah Rosenfeld:
Oh, that's interesting.
Katherine Hurtig:
Yeah.
Denis Sushkin:
It was. Made it more accessible and closer to his world. And it was helpful.
Katherine Hurtig:
Yeah.
Sarah Rosenfeld:
So, taking a concept, or the way I've seen it used, is if somebody's struggling interpersonally, they'll kind of use the tool to figure out how would I have an interaction in this situation? And then they wrestle with the information that the tool provides, but it's also something that comes into the therapy space. So, it's just deepening, similar to what you said, Denis, it's deepening an understanding of something outside of therapy, which is what we want people to be doing anyways, right?
Katherine Hurtig:
Exactly.
Sarah Rosenfeld:
I think therapy is a point of connection, offers a lot of useful things, but the fact that the person is then thinking outside of therapy about how they can use this, how they can make sense of it, how they understand it differently. I think that's where AI can be really helpful, like deepening understanding, deepening like, ‘OK, how am I making sense of what we talked about?’
Denis Sushkin:
It's almost like, when it's human first and then there is another element that you can deepen.
Katherine Hurtig:
Yeah.
Denis Sushkin:
The same thing with isolation. I think as long as you have other people that you can talk to about certain issues outside of AI, then it can be something helpful. But if AI is the only thing that you interact with, or if you don't have anyone who can take care of your plants or pets when you're away, and AI is the only thing that you talk to, I think that will deepen the isolation. But if you have people that you can say, ‘Hey, I'm away, can you take care of my cat, or pet, or plant?’ And then, so AI can actually be just one of the things in the background that helps people grow, learn.
Katherine Hurtig:
So, kind of on that note of playing into relationships, AI can feel empathetic. It has, like we said, that illusion. Why is this illusion so convincing and can that be dangerous?
Denis Sushkin:
For me, it goes back to kind of like, what's the context of that person? Like, where are they coming from? Is the AI the only source of empathy?
Katherine Hurtig:
Right, yeah.
Denis Sushkin:
I think that's the danger that can be present there.
Sarah Rosenfeld:
And I think then the danger for me, in addition to what you're saying there, Denis, is will the person know if this is a danger to them or not? So, I think when you're looking for empathy, validation, being seen, feeling heard, and you're a vulnerable person, and you just need more, and more, and more of that reassurance, it's kind of like reassurance seeking over and over again. That in itself can lead to a feedback loop that could be unhelpful. But if you're vulnerable or you don't see that this could be problematic for you, I think you just keep seeking that without recognizing there's lots of colours in the rainbow.
Katherine Hurtig:
Yeah.
Sarah Rosenfeld:
And so that I think is the challenge. Well, you know enough to know. It's sort of like, there's things we know, there's things we don't know, but people who are really at risk are the people that don't know the information, and don't know that they don't have the information. And that to me is the population that I would be most concerned about.
Katherine Hurtig:
For sure.
Denis Sushkin:
Plus, in a way, AI becomes like a mirror to that person who's using the AI. It is a feedback loop. And sometimes, I mean, if I'm honest, there are so many moments when what I wanted wasn't what I needed.
Katherine Hurtig:
Yeah, exactly.
Denis Sushkin:
I want the validation and support, but do I need it right now? I'm thinking about when it was a conversation and the decision like, to go to grad school and not to go to grad school. I just wanted the validation. Yeah, it's scary, and it's hard, and I don't want to do it.
Katherine Hurtig:
Yeah.
Denis Sushkin:
And if people in my life were to provide that back then, maybe I wouldn't be where I am now. But they had the capacity to agree, disagree, think for themselves. You know, you could do a lot of things. It wasn't just a statistical prediction of what's coming next.
Katherine Hurtig:
For sure. Well, yeah, it goes back to like we need to be challenged sometimes.
Sarah Rosenfeld:
Yeah, and I think that's the piece, right? Is some of these traits of being able to self-manage and regulate your emotions. And it actually comes through successfully getting through something that's kind of challenging or difficult.
Katherine Hurtig:
Yeah.
Sarah Rosenfeld:
So, if you're never presented those opportunities...
Katherine Hurtig:
You're not going to build that skill.
Sarah Rosenfeld:
You don't actually learn how to do that. And so, then you just keep... We're seeing this a little bit with children in the snowplow parenting, the parents that try to move everything out of the way so that the kid doesn't actually learn how to self-manage.
Katherine Hurtig:
Yeah.
Sarah Rosenfeld:
They get to university and parents are calling the dean of the school saying, ‘Why are they getting a grade on this?’ I'm not saying it's exactly the same thing, but, I wonder, you know, in the absence of some of the challenges, how do you actually know where you stand in relation to that? Know what matters to you? Know what your values are?
Katherine Hurtig:
For sure.
Sarah Rosenfeld:
Those things really are important for people to develop those things.
Katherine Hurtig:
Yeah, that's like anything. I mean, if you're an athlete and you're never challenged, you're not going to grow those muscles. You're not going to be able to adapt to different scenarios. So, we have to go through difficulties and failures, and kind of learn from them.
Denis Sushkin:
And I think in addition to this, there's another challenge of like, AI gives us unlimited access to information now. We can talk about anything and everything. And it's not connected to empathy, but it's just a challenge by itself. Because in a way, it can speed up your brain so much. It's never ending. Like, you can ask another question, another question, another question, and things will come your way. And there's no limit.
Katherine Hurtig:
Yeah.
Denis Sushkin:
In the past, like when I went for my first degree, I had to go to the library, find an index card—
Katherine Hurtig:
Yeah.
Denis Sushkin:
Find information in the end of the course, go to another place, find information, read the book, or find something in the book. And now I just kind of like, I put a prompt in, and I get a summary of the book, or several different books. So, the amount of information that can come at a person.
Katherine Hurtig:
That's overwhelming.
Denis Sushkin:
It’s overwhelming and it’s insane.
Sarah Rosenfeld:
Yeah.
Denis Sushkin:
So, the person's ability to slow down, to stop, to sort through what's relevant information, what's not relevant information.
Katherine Hurtig:
Yeah.
Denis Sushkin:
So that's, I think, that goes back to, as a society, do we know how to use those tools? I don't think anyone knows for sure, but like all of us are in this kind of learning uncharted territories, but it can have an impact on mental health. Positive impact, if you have some good information and you can learn something new, but also it can be overwhelming. And I think the speed at which your brain kind of moves when you have this amount of information is insane.
Katherine Hurtig:
There is a limit to that, yeah. From a safety perspective, what should concern people most when they're using AI to talk about their mental health?
Denis Sushkin:
So first of all, it's the AI. Like, I don't think AI is capable of handling any kind of complex crisis, like thoughts about ending their life or complex mental health needs. So, it's too broad of a tool, first of all.
Katherine Hurtig:
Yeah.
Sarah Rosenfeld:
The other thing I would say is it's unregulated. So, it means that any information you put into that system is actually going to be used by different people that are actually heavily invested in making sure this becomes a marketable product. And so, without guardrails in place, when you're dealing with vulnerable populations, there's a risk there because they're not really going to slow it down, or they will keep asking for more information because that is actually going to then help them build the next AI—
Katherine Hurtig:
AI version.
Sarah Rosenfeld:
So, I think that's why you need people with training and expertise to be able to help people navigate that.
Katherine Hurtig:
Yeah.
Sarah Rosenfeld:
Your information is for public consumption, essentially. If you were to put some details about yourself in there, there'd be no confidentiality around that information.
Katherine Hurtig:
Yeah. How does that differ from the counselling room, and the sessions you guys have with clients? Like confidentiality, privacy, how does that differ from what people are doing with AI?
Sarah Rosenfeld:
Well, in our space, confidentiality would be within a team of trained professionals who've been, you know, there's a process by which they come into a centre like ours. They get ongoing training, PD, opportunities for consultation. And we know how to professionally talk about these situations in a way that clients are always respected. And in the AI space, I'm not sure those guardrails are in place. And then the piece that Denis mentioned about risk, we would be able to kind of flag or consult appropriately with external professionals and other resources in the event that risk was flagged or identified.
Denis Sushkin:
And the other piece, as professionals, we have regulatory bodies that we respond to. We have expectations of what we need to do. People look at how we work, what we do. They have, like, a specific code of ethics and conduct that we need to follow.
Katherine Hurtig:
Yeah.
Denis Sushkin:
Which is very different from, yeah, a business setup when there's a profit involved and other things.
Katherine Hurtig:
Exactly.
Denis Sushkin:
So, there are, yeah, colleges who protect us as professionals and colleges who protect.
Sarah Rosenfeld:
Yeah, thank you.
Denis Sushkin:
But then there's, to my knowledge, there's nothing like that in the business world.
Katherine Hurtig:
Yeah, I mean, like we've said, it's still, you know, fairly new. And yeah, I don't know how, or if, or when it will be regulated and what that will look like.
Denis Sushkin:
Well, I think it's an exciting thing. I know it's a bit of a... Just the analogy of the steam engine wasn't invented until, what the 1700s?
Sarah Rosenfeld:
Yeah, something like that.
Denis Sushkin:
And then the laws, the physics laws that came with it, were developed 100 or 150 years later.
Katherine Hurtig:
Yeah.
Denis Sushkin:
So, I wonder, I don't think it will take people 100 years now with the speed that everything develops.
Katherine Hurtig:
Yeah, I would hope not.
Denis Sushkin:
But it's kind of like, well there's this tool, and now I think what all of us are trying to sort through like, so how does it work? How do I apply it?
Katherine Hurtig:
Yeah, trying to catch up with it.
Denis Sushkin:
Yeah. So, I think it happened before, many times, so.
Katherine Hurtig:
Yeah.
Denis Sushkin:
That's the piece that's kind of not new.
Katherine Hurtig:
Yeah, for sure. I mean, it's this kind of scary, new thing. But we've kind of been here before with lots of other technologies, so.
Denis Sushkin:
Yeah, and I think in a way it touches on the question of trust. How do we trust, what do we trust, how do we make those decisions—What to trust, and what kind of things we trust people with or AI with. It's not a black and white question.
Katherine Hurtig:
So, in your opinion, what are the biggest limitations of AI when it comes to supporting mental health?
Denis Sushkin:
Well, the big piece that comes to my mind is that there is no challenge. There is no friction. And then you kind of... Growth and development requires this kind of friction, and uncertainty, and intention. Be it like addressing the feeling of anxiety or depression. I think there is an element of you need to be vulnerable, and risk things, and have this element of uncertainty when you interact with a therapist or with others in your life to improve your own well-being. So, I think this limitation that, I don't think AI can provide, even with a really good prompt, this level of tension that another human being can.
Sarah Rosenfeld:
Yeah, I would agree with that. I think we've identified some of the other ones, right? That the complexity of somebody presenting with suicidality, risk, and that like complex mental health. And by that, we mean things like, you know, psychosis. Those things are really not going to be well served by a platform like this that struggles with nuance. So, I think there is a population that wouldn't be well served by what this has to offer.
Katherine Hurtig:
Right.
Sarah Rosenfeld:
There might be some elements that would be helpful, but for those acute points in that person's life, I think there's some significant limitations. And then I think the piece that you talked about, Denis, about the regulatory piece. Like, without regulation, this can be going directions that may actually not serve people's interests.
Denis Sushkin:
And there's also limitation, like, do we know how to use this tool? It's a complex tool. We need to kind of learn how to use it, know how to use it well. Just giving like, ‘I think it's a good prompt.’ Is not a good enough thing for me.
Katherine Hurtig:
You've kind of brought up examples of how some people are using AI in your work with them. Have you come across examples where they're not using it well, and you kind of have to redirect that, or?
Denis Sushkin:
Well, one thing is the relational piece, and that's a whole topic by itself. Like, when you have a couple and you're working with them, and there is a third or fourth element in the relationship that's AI, and how good of a relationship, and do they work on the relationship with their partner. Or they're just outsourcing some of the things to AI and kind of get the advice, and so that's a piece that's sometimes challenging.
Katherine Hurtig:
Yeah.
Denis Sushkin:
The other piece is there's so much information that's available through AI. And we knock off just with therapy, there are what, 400, 500, a thousand different models that's approved, evidence based. And we know all models are equally effective, meaning some people, some of the time. So, if you have a therapist who's working from a specific kind of standpoint, kind of in a theory, and then the client goes to AI and consumes a lot of other things. And it can become like a debate.
Katherine Hurtig:
Right.
Denis Sushkin:
‘Well, you said that, but then I talked with AI and the AI told me that. So, who's right here?’ So that's an interesting thing to navigate, too. Clients have access to unlimited amounts of information, and there is always another perspective that will disagree with the perspective that I might have on something.
Katherine Hurtig:
I guess that's a good point too. Because in my conversation with you guys, you talk about all these different modalities of working with people. And that's kind of part of your expertise is, you know, through working with clients, you're able to, through your own experience and knowledge, use the best modality for that particular person. And that's not going to happen with AI. Like, a person may have heard about cognitive behavioral therapy or dialectical behavioral therapy, you know, in passing or over social media. And they might use that through these platforms or these chatbots, but that might not necessarily be the best way to help them with what they're going through.
Denis Sushkin:
Because I think even when they read different things about AI, even I think people who create all those tools, they're not fully sure how it works. So, sometimes people talk about how AI can hallucinate.
Katherine Hurtig:
That’s right.
Denis Sushkin:
It can create something that's not real and kind of reinforces that feedback loop. Or the whole idea of, can we make AI align fully with how we think and kind of, I don't know. Maybe it's silly, but going back to those laws of robotics, for not harming someone.
Katherine Hurtig:
Yeah. Sarah, you talk a little bit about privacy, and confidentiality, and things like that. So, when people share, you know, their thoughts or emotional struggles with AI, what kinds of data are they forgetting that they might be giving away?
Sarah Rosenfeld:
Just their personal information.
Katherine Hurtig:
Yeah.
Sarah Rosenfeld:
So, it's not the same as it would be in a space where a therapist has disclosed, ‘These are the times when I can keep what we've talked about confidential, these are the times when I wouldn't be able to do that.’ And it's usually around risk, or risk to themselves, or somebody else. Those caveats aren't there for AI. So, anything that you program into the system is now accessible and can be used by other people. And you don't always know, you're not going to know actually, how that information will be used or where it will end up, but I'm just not sure that that gets disclosed at the front end. So, as you're divulging more personal information about yourself, your circumstance, all of those things, that's not private to you and that chat bot. It goes into places that I'm just not sure people are fully aware of. And because there isn't a process by which there's a disclosure each and every time about how the information will be used.
Katherine Hurtig:
Yeah.
Sarah Rosenfeld:
People are seduced into thinking that it may be private because it's happening in their own homes, in the middle of the night, but it's actually not private or confidential.
Katherine Hurtig:
Now, I don't want you guys to speculate or make any guesses. How do you see AI maybe potentially helping, like you said, helping professionally? How could it potentially help therapists in their work?
Sarah Rosenfeld:
Well, it's a little bit around this access to information, maybe helping educate people with how to use these tools in helpful ways. I mean, I know this happened a little bit when COVID occurred, and people moved to Telehealth. The regulation around how you use that, what you need to be informed of, came after the fact.
Katherine Hurtig:
Right, yeah.
Sarah Rosenfeld:
Because we were trying to provide service to people that were really isolated and disconnected. So, I see it similar to that, is sort of, this is a tool. It's available to you. This is how it gets used. This is where the information goes or doesn't go. And informing people about those pieces and then training and support that can be provided to professionals about the questions to ask, or what's more helpful or less helpful. Like, I do think that there's an opportunity there that could present itself to support professionals, to then be really client-centred in the information that they're sharing so that it's useful to clients we're serving.
Katherine Hurtig:
Yeah.
Denis Sushkin:
Or if it can help automate some of the things that we have to do, that doesn't require a lot of... Like the intake processes. If there are some of the things that can be just automated, I'm looking forward to the day when I don't need to write session notes. And it just becomes part of the workflow. It's there, I can check.
Katherine Hurtig:
Yeah.
Denis Sushkin:
I can see what's there. I can still kind of proofread what's there because it's still my name that goes there. But then it can clear up some time in my day to do more work, to be more available to people. Automation and getting rid of the things that we don't need to be involved in.
Katherine Hurtig:
Right.
Denis Sushkin:
Have a lot of potential, like just for therapists' mental health and the load that they carry, but then it will automatically make them more available to people.
Katherine Hurtig:
Yeah. In the end, it will help the clients.
Denis Sushkin:
Yeah.
Katherine Hurtig:
So, what are some signs that might suggest someone is starting to lean on AI in a way that's replacing connection or care, instead of just supporting it?
Denis Sushkin:
The first thing for me is like, how many people has the person talked to before they talked to AI? Would be the main thing.
Katherine Hurtig:
Yeah.
Denis Sushkin:
If there was at least four or five people that person reached out to and talked about that specific thing that they want to discuss with AI. I think that's at least five, I don't know, I hate to put a number.
Katherine Hurtig:
Yeah.
Denis Sushkin:
But I think like for me, it's the, yeah, what's the context? And are there actual people? Like, if a person doesn't have anyone that they can talk to about their feelings of depression, anxiety, or struggles that they're dealing with. If there is no one else who is helping them, who is in their corner, and AI is the only thing, I think that's a big red flag.
Katherine Hurtig:
Yeah.
Sarah Rosenfeld:
I think it would also be the passage of time, right? So, did you sit down to ask a question, and then 12 hours later you're still sitting at that computer? Like, passage of time is one of the, I think, risk factors, right? When we're using this tool, is he able to kind of move in and out of the use of the tool, in addition to what Denis said? Who else are you consulting with? But if a lot of time has passed and there's been no external input.
Katherine Hurtig:
Right.
Sarah Rosenfeld:
Nobody's come to the computer and said, ‘Oh, what are you asking about?’ And you find yourself, you know, a long time later.
Katherine Hurtig:
Yeah, you're just in this bubble.
Sarah Rosenfeld:
That's usually going to be an indicator that they’re spent, right?
Katherine Hurtig:
And if someone's using AI to, you know, get information, talk through their feelings. What boundaries or mindset should they kind of keep in mind, you know? It stays a tool, like we've talked about. It stays supportive and doesn't take the place of professional help or real relationships.
Sarah Rosenfeld:
I think this is where balance is so key. And balance is a bit elusive for all of us with lots of different things, whether it's AI, sleep, food, exercise.
Katherine Hurtig:
Yeah.
Sarah Rosenfeld:
So, I think it's being able to say ‘This is a relationship I have with this technology that I need to check in on regularly to see is it serving my interests, my needs? Have those things changed? Am I right in an acute state where I'm relying on it more? And what are the reasons for that?’ But I think it's really this, we've got to stay connected to the, as Denis put it, the tension of using these things versus just becoming complacent that it's limitless or endless. No, there's always limits on things, but sometimes we have to impose them on ourselves.
Katherine Hurtig:
Yeah.
Sarah Rosenfeld:
And it's not necessarily going to happen from an external source. So, I think just having a mindset of, ‘How do I kind of see if I'm in balance, see if I'm in check with this tool?’ As it would be with some other things that we interact with regularly.
Denis Sushkin:
Yeah, it's so true. Like, the balance, because you can have too much of a good thing, too.
Sarah Rosenfeld:
Yeah.
Denis Sushkin:
You can go to the gym every single day, twice a day, and it will lead to some kind of injury.
Katherine Hurtig:
For sure, it’s like anything.
Denis Sushkin:
So, how do you use it? What kind of relationship do you have with this tool? It's a really good self-reflection that people need to have.
Katherine Hurtig:
Thank you so much, guys. This was an important conversation. Happy to have it with you.
Denis Sushkin:
Thank you for having us.
Katherine Hurtig:
You've been listening to Living Fully, a Calgary Counselling Centre podcast. Thank you for tuning in. This episode was produced by Luiza Campos, Manuel Montano, Jenna Forbes, and by me, Katherine Hurtig. A special thanks to Sarah Rosenfeld and Denis Sushkin. Tune in to our next episode on April 7th. We'll be talking about the difference between worry and anxiety.
Carmen Sadoway:
It's really the idea of intolerance of uncertainty is really at the heart of worry and anxiety.
Katherine Hurtig:
Yeah.
Carmen Sadoway:
Worry in particular. When that's there, this uncertainty theme, and we're having a hard time tolerating it, it just floods in all kinds of stuff for us. And I think when there is uncertainty, it's just, there's this reminder that there's actually just so much in the world that's beyond our direct control.
Katherine Hurtig:
Yeah.
Carmen Sadoway:
I think you mentioned that, like there's so many factors involved and influences that can impact an outcome. And I think just day to day life, we're often confronted with the reality about how our own power and influence to control things is actually limited.
Katherine Hurtig:
To stay up to date on our latest episodes, be sure to subscribe. We're available in your favourite podcast app. Living Fully is a production of Calgary Counselling Centre and recorded in Calgary on Treaty 7 territory. Living Fully podcast is not a substitute or alternative for professional care or treatment. Calgary Counselling Centre and Counselling Alberta provide effective counselling to anyone in Alberta with no waitlist and no financial barriers. Find us online at calgarycounselling.com and counsellingalberta.com. For help across Canada and the United States, call 211. If you are outside of Canada and the U.S., seek help from your general medical practitioner.