Eric: [00:00:00] This is the Counseling Insights podcast, a podcast to help you develop your own insights and practices as you listen to Vicki Enns connect with counselors and helping professionals about their work, their passions, and their strategies. Vicki: Welcome to the Counseling Insights podcast. I'm so glad you're here, and I'm very excited for today's conversation. My name is Vicki Enns, and I am your host. And today, I'm having an important conversation on the topic of AI ethics and mental health. My guest is Candice Alder. Candice is a clinical counselor, psychotherapist, and an AI ethics and policy specialist. In this work, she is focused on the governance of artificial intelligence in high-risk human impact domains, including systems affecting children and other vul- vulnerable populations. She is the co-founder and principal consultant at Synthetica, advising on responsible AI governance, risk management, and [00:01:00] ethical integration in high-stakes environments. Candice brings practice-grounded insight to AI policy design and implementation, and she authored Canada's first professional guidelines on ethical AI use in clinical counseling and psychotherapy, and her work emphasizes ethics, risk, and duty of care in clinical practice. In this conversation with Candice, we explore how people are using AI within the counseling realm, whether for administration, supporting their counseling work, or for one's own mental health support. Candice describes both what concerns her about how AI is being used and also what excites her about its potential. And we start to get into specific strategies that every counselor can learn to start to do to ensure we are approaching this ethically and effectively. Today's conversation is part one, and we'll have a follow-up coming in the future months. Let's dive in. [00:02:00] Welcome, Candice. Candice: Thank you so much. Thanks for having me here today. Vicki: Really looking forward to this conversation. Um, as I already noted, and I think my anticipation has been building a little bit because I- we've met and chatted a little bit. Um, but in the meantime, I've also been listening to some other recordings you've done, some other teaching, and that's prompted me to read some other blogs and websites. And, um, so I feel like it's, it's generated a lot of different feelings and a whole lot more questions, maybe even a little confusion. So, um, and so I'm really looking forward to talking to you today, and I'm, I'm excited to offer this to our listeners too, 'cause I think it's just so important. Um, but before I dive into my questions, I always like to give my listeners a chance to just hear from you a bit, and I like to learn a little more from the guest. And so do you wanna talk just a bit about how you found yourself in such deep waters of exploring AI and- Yeah ... and all things ethics and mental [00:03:00] health? Candice: I, absolutely. Yeah. No, and, and it's a common question that I get because, I mean, I've been working with children, youth, and families for over 20 years now. Um, I- my, the therapeutic, uh, context that I have been working in have spanned from, um, public education settings to governmental settings and nonprofit, and of course I have, um, a private practice as well. Um, so how does somebody who has been working in the therapeutic space end up up to their eyeballs- Yeah in, uh, in, in, in AI? Well, you know, it's actually kind of... well, I think it's an interesting story anyways. Mm-hmm. Um, so this was back in 2020, 2021. Mm-hmm. Um, my partner actually was invited to be a beta tester- Oh ... for ChatGPT before it was ChatGPT. Vicki: Oh, very cool. Candice: Um, it was called The Playground at that time. Mm-hmm. Um, and, you know, I, I was a very busy person then, as I am now, and, um, you know, I knew that he was doing some beta [00:04:00] testing. Uh, but I didn't really have a, a good sense or to be perfectly honest with you, I didn't really care very much as to what exactly what it was all about. Vicki: Right. Candice: I had my own stuff going on. But there was this one particular evening where I wandered past him on the computer. He was on the computer. I wandered past him and I was like, "Okay." what is going on here? Like, what are you actually doing? And he showed me. Ah. And the version, I mean, like I say, it wasn't ChatGPT then. It was called The Playground. Yeah. Um, but, but i- i- for all intents and purposes, it was what was to become ChatGPT. And the ChatGPT then is not anywhere near the, the, the kind of power and capability that it has today, but it was amazing still even then. And my mind was completely blown. Um, so, you know, basically from there it was off to the races. Yeah. Because one of the things that really struck me was I was thinking about other technological advancements and how those have [00:05:00] impacted the field of, uh, therapy, of, of clinical counseling- Yeah ... psychotherapy, whatever it is the term that you're, that you're using to describe it. And one of the things that I thought about, because of course at that time we were still in the throes of the COVID pandemic- Right. Yeah ... um, was, was virtual counseling. Um, you know, prior to virtual, or par- pardon me, prior to, um, uh, c- uh, the COVID pandemic, we had had capabilities since about 2006, 2007- Right ... I think is when Skype came out. Right. We had had the capabilities to be able to do virtual counseling. Right. Yeah. Um, and even the stuff that you and I are doing right now, I mean, y- you and I are physically not in the same place- Right but we're able to have this conversation. Um- But, but counselors, particularly in Canada, I wouldn't generalize this across all, uh, communities, but certainly in Canada, uh, counselors, therapists, uh, psychotherapists, uh, you know, psychologists, we're a very [00:06:00] cautious people. Right. Mm-hmm. We're very ethically minded and cautious. And, um, a lot of the efforts to bring in virtual counseling were met with- Mm ... you know, "No-" Mm-hmm ... we don't understand- Yeah ... this." Yeah. You know, "Stop, stop, stop," right? Yeah. Well, Vicki: I remember hearing about it- But- ... as like, like just thinking of it as telehealth. Candice: Yeah. Vicki: And it's something nurses did sometimes. Yes. And yeah- Yes ... I didn't really wanna touch it. Candice: And, and, and, and, and your, and your feeling about it was consistent across most- Yeah ... uh, most, most folks, right? What happened, though, when we were all forced inside our homes- Right. Mm-hmm ... and out of our workplaces in, uh, during the COVID pandemic, was that counselors in the mental health industry, as a general statement, were forced to adopt a technology that they were resisting up until then. Vicki: Right. Yeah. Candice: So what does this mean for AI, and what does, what did this mean for me? This was back in, I think, 2021 at this point, um, when, when I was exposed to, uh, you know, OpenAI's, uh, [00:07:00] you know, what would become ChatGPT. Well, it immediately struck me that if the mental health industry did not become involved in the deployment and the development- Mm of AI- Vicki: Right ... Candice: that it was not going to be something that we could jump on the bandwagon with later. This horse was gonna get out the gate. Yeah. And it was gonna turn- Yeah ... into something- Yeah ... that would not be consistent with what we in the mental health field really want for it. Right. So i- it was, like I say, it was off to the races from there. Wow. And that's, that's how I got into- Wow ... AI. Vicki: That's how you got interested in it. Okay. And, and obviously we'll hear more about what you've been up to since that spark of interest. Yeah. Um, but even just to build on, you're already starting to talk a little bit about how you first... I mean, you're describing very vividly, it sounds like your response to AI was one of fascination, fascination and curiosity- Mm-hmm and kind of excitement it sounds like. Absolutely. So I, I will disclose that I think my f- first response, [00:08:00] as we already alluded to, was something a bit more cautious, certainly fearful. Yeah. And just, and probably just outright avoidant. Um, however, but with some fascination as well. 'Cause, you know, I hear about it from, say, students that I was supervising or teaching and- Mm-hmm um, or just my c- uh, clients, and m- you know, we'll maybe talk about that at some point, too. And, um, and so, uh, so certainly some fascination in it, and, um, intrigued by the capability overall. Um, however, it's, I think that's such a big part of this whole stage we're all in, is we're all figuring out our own- Figuring out what it is and how, how to use it, but also whether to use it, our own relationship with it as well. Yeah. Candice: Absolutely. And I mean, this is... E- even though, you know, for all intents and purposes, the, the... First of all, let me, let me just say that artificial intelligence and the, the technology that sort of surrounds it, um, has been in development for decades. Vicki: Mm-hmm. Right. Candice: [00:09:00] Okay? This is not new technology, and the truth of the matter is, is that you and everybody else have been essentially, um, interacting with, with artificial intelligence long before- Right ChatGPT even entered the vernacular, to be- Right ... to be completely honest with you. If you've ever spoken with, uh, Siri, if you have an Apple phone, or Alexa- Right ... or if you have Spotify and you've received, um, you know, recommendations from Spotify or from Netflix- Yeah ... or if you've called into, um, if you've called into a, a customer service center and an automated voice has asked you to say a little bit about what you're calling about- Oh, I know. That's so annoying ... to get you to... It's, it's all AI. That's true. This is the whole thing. That's true, yeah. So it's actually been kicking around for a little bit. Yeah. However, that being said, generative AI really burst onto the scene in November of- Mm ... of, uh, 2022 when ChatGPT was released. And, um, and, and I think that, that Vicky, your, um, your response of concern and, uh, [00:10:00] "Oh my gosh, that makes me feel worried," um, that's not an uncommon response. Right. This is a, this is an emerging technology even three years later. Yeah. Um, the, the capabilities change- every few months. Right. Um, and we still don't know exactly how this is going to, um, how this is going to play out for humanity- Right ... because this is a humanity-impacting technology. Absolutely. Yeah. This is like the printing press. Mm-hmm. This is like the steam engine. Mm-hmm. And we are writing the story as we are living it. Yeah. And, uh, you know, there's, there's absolutely fear around that. Yeah. There's fear as it relates to, um, you know, the, uh, issues that relate to the job market. Um, there's fear as it relates to, you know, some of the popular culture depictions- Right of AI. Yeah. I'm thinking like iRobot and Terminator- ... and all of those kinds of those pieces that have been put in our head by, by Hollywood and what have you over the years. Vicki: Yeah. Candice: Um, and just not [00:11:00] understanding something that everybody is really excited about and hyped up about- Right ... um, that can bring on anxiety in and of itself. Vicki: Right. Yeah. But it... And, and I can feel as I'm listening to you, which is part of why I was really excited to talk to you- Mm ... I can also feel that excitement in me. Like, as you describe it- Mm-hmm ... you know, alongside the printing press and the ste- Like, it's like, yeah, like, that's pretty cool too, right? That we're, we're living it as it's unfolding. So, um, so wanting to invite people to kind of let's, let's turn towards it and- Yeah ... learn about it, right, rather- Candice: Absolutely ... rather than just avoid it. It's not going anywhere. Yeah. Mm-hmm. It's not going anywhere, and the truth of the matter is, is that even as a mental health clinician, however you frame yourself- Yeah the fact of the matter is, is that you're interacting and your life is going to be impacted- Right ... if it hasn't already- Exactly ... by AI. Yeah. Because AI adoption is happening across all sectors. Yeah. So- Mm-hmm ... you may choose to slowly learn about it or jump into [00:12:00] adopting it within your own clinical practice, but at the end of the day, you can't escape it. Vicki: Yeah, absolutely. Okay, so let's, let's go more down that path and, um- Mm-hmm ... uh, I mean, as you said, AI is in every part of our lives, uh, right now, but today- Yeah ... let's focus... We're gonna focus in on, you know, go through the doorway of just thinking about it within our mental health world and, and counseling field. Um, and so, um, at, like, as you were just saying, we're all encountering it. Some people are diving in- Mm ... and using it or learning about it. And even if not, I mean, I, almost every day I'm seeing a pop-up trying to encourage me to s- to use AI assistance for writing my session notes, for- Mm-hmm ... transcribing things, and, and so we're all encountering it in some kind of way. Um- Mm-hmm. Yeah ... where, where would you like to start in that? Like, where are you seeing the main entry points of, of people in the mental health world, uh, kinda using AI? Candice: Well, it's-- [00:13:00] I, I would say that it's definitely two-pronged. Okay. So, so folks are either finding themselves being drawn into learning about AI, um, because they are finding out that their clients are using AI- Right. Vicki: Yep ... Candice: or they're attracted to the idea of reducing that administrative, uh- Right ... burden that comes along with clinical notes and, uh, billing and, and all of the other pieces that, that come along with, um, you know, running a private practice or- Right ... a group practice- Yeah ... or, or, or a community service agency or something to that effect. Yeah. So I would say that those are the two primary, um, entryways that, that folks are finding, um, their way into use of AI in mental health, particularly as it relates to individual practitioners. Um, I'll start with the piece around clients, um, because this is actually one of the more, um, emerging stories that I'm- Yeah starting to hear when I- Hmm ... have conversations with clinicians. Um, because they're saying, "You know what? [00:14:00] Like, listen, I'm not really using AI, Candice. I, I'm trying to learn a little bit about it, but I'm a busy person, and I just haven't gotten there yet. But one of the things that I'm noticing is that when clients are coming into, into session, they're talking about- Right conversations that they've had with ChatGPT, and I'm really concerned about this." Yeah. "How do I, how do I, um, approach this?" Um, and, and they talk about how, you know, uh, this has actually motivated them to learn a little bit more about, uh- Right ... about generative AI. So, uh, first, I think it's an important distinction to make that when we say AI, we're actually talking about a really broad, um, a, a really broad array of, um, technologies. The one that people are int- are, are, um, uh, interacting with the most is generative AI. So that's gonna be your ChatGPTs, your Claudes, your Geminis, and the like, okay? Right. Um, and, [00:15:00] and, and that's the primary one, but it's really important to remember that it-- generative AI is just a branch on a tree. Right. The tree of AI. Right. Um, you know, i- AI systems are used for prediction, they're used for classification- Mm-hmm ... for biometric identification, recommendations- Right ... decision supports, even, even automation. Vicki: Right. Candice: Um, but, but we'll focus on generative AI because that, that really is the, the, the primary one that folks are using. Now, the first thing that I wanna say is, is that I always caution everyone that I speak to- Mm-hmm ... whether it's a client or whether it is a clinician, against using generative AI for ther- general purpose generative AI for therapeutic purposes. Vicki: Right. Candice: Okay? So the ChatGPTs, the Clads, the Geminis of the world, these are not AI systems that are developed for, for therapeutic and mental health applications. Right. They are prone to [00:16:00] hallucinations, which is to say they're prone to making up information. They're prone to having, um, inappropriate outputs, and they are prone to... A- and, and as we've, we've heard in, in, you know, multiple tragic, uh, news stories- Right ... they're prone to actually engaging in a way with users, particularly in the chat function, uh, when it's a back-and-forth conversation, in a way that is deeply therapeutic, uh, in a- deeply inappropriate in a therapeutic context- Mm and has actually been proven harmful. Vicki: Right. Candice: Now- ... I want you to hold that in one hand- Right ... because on the other hand, we also know that our clients are gonna do what they want. Vicki: Right. That's right. Candice: So- Yeah, yeah ... it, it, as much as it's not advisable to be- Right ... using ChatGPT- Yeah ... for, for therapeutic support, we know that that's what folks are doing. Right, Vicki: right. Candice: So what we don't wanna do- Yeah ... is we don't want to shame folks into- Right ... um, [00:17:00] in- into not using or feeling like they shouldn't be using it. Right. Because what that ultimately does anytime we use shame- Yeah ... and any therapist that is listening to this knows exactly where I'm going with this. Right. When we introduce shame or- Yeah ... we introduce a, a punitive component, anything that even feels punitive into the conversation, what happens? People don't stop what they're doing, but they then start doing it- They might stop seeing us ... Vicki: and not Candice: telling Vicki: you Candice: about it. Yeah. Hmm. Vicki: Or they pull Candice: back. That's ex- Yeah, Vicki: yeah Candice: exactly. Yeah. Exactly. And, and really at the end of the day, what we want is we want folks to talk to us about it. Right, yeah. Because when they come to us and they say- Yeah ... "Well, ChatGPT told me this- Okay ... about my presentation," it gives us the opportunity to be like Well, that, that's a little bit true, but here are the pieces that are missing. Vicki: Right. Candice: Or that's actually not true at all. Mm-hmm. And this is actually something- Right ... that they call an AI hallucination. Vicki: Ah, yeah, yeah. So- So this is an important, um, I'm hearing a really important takeaway already [00:18:00] from what you're talking about, like... And, and pardon me, Candice, we, this is maybe a whole nother section we wanna talk about, but an, I have the, the phrase that I've heard you use when I've listened to you at other times of- Mm-hmm in my mind, which is AI literacy. And- Yes ... and I, and I wanna ask you more about that. But this is sounding like, like this is maybe one step in that, like for us as- Mm-hmm ... counselors, as therapists, is to even just- Mm-hmm ... have that ability to engage in a conversation and, and maybe encourage that conversation of, of, like you said, not shaming and saying like, "Oh, that..." Like, like adding fear into the use of it, that that's dangerous or something like that, but inviting a conversation about it, and even that we know enough to, to, uh, invite that conversation with folks. Yeah. Candice: Absolutely. Absolutely. And, and, and I wanna be super clear. I am not, I am not against general purpose generative AI. I think it's amazing. Mm. And it has wonderful, wonderful, exciting applications, not just for clients- Vicki: Right ... Candice: [00:19:00] but for- Yeah ... for, for clinicians as well, for, for everyone. It just needs to be used with caution. Right. And it just needs to be used appropriately. Right. Vicki: And I like how you say- And when- ... like it's not devel- Like it, it's l- I don't know. I, this is a dangerous analogy perhaps, 'cause I'm not handy necessarily with tools, but it's, it's like, uh, it's like trying to pick up a tool in a wood shop or something, and you wanna use the right tool for the right job. Mm-hmm. Um, or else you might wreck your project, right? So I don't know. Candice: Yes. Vicki: It's Candice: a bit Vicki: of a Candice: parallel. Exactly. It's like one of the things that, one of the analogies that I, that I like to use, it's like using a garden hose as a rope, right? It, it, it will do the job- ... but it doesn't do it very well. Right. Oh, that's a good one. And it's not going to- Yeah ... you know, it's not going to achieve the sturdiness that comes along with using a rope if you're trying to, like, tie something, uh, tie something off or what have you. Right. Um, so, so really just being able, being comfortable- Yeah ... asking your clients- Right ... about their use of generative AI, you know? A- a- and gett- [00:20:00] becoming curious about that, building your AI literacy so that you understand. Vicki: Mm-hmm. Candice: Um, even just the basic pieces around things like hallucinations. Right. Around things like, um, you know, the ChatGPTs, the Clads, and the Geminis being a little bit, um, agreeable by their very nature. This is something that we've seen change over the course of the last year. ChatGPT, for instance, is not nearly as agreeable, um, in its chat function, uh, with, with users as it was, um, as it was a year ago- Okay or even six months ago. Yeah. But it still does happen. Right. These are services that are designed to keep you engaged. Right. They want you to keep- Right ... you using it. So it, it does have a tendency to tell you things that you want to hear. Right. Now, that's not a big deal if you are, uh, you know, talking about something, uh, that, that really doesn't matter. But if you're a 14-year-old, for instance- Right ... [00:21:00] that is, um, that is feeling really down in the dumps, and you're being encouraged to go down, um, go down a path that is ultimately gonna be really dangerous for you, that's incredibly problematic. Yeah. And while that isn't something that we necessarily have a really good example for as it relates to ChatGPT, I can tell you, and a lot of folks will remember this, that the, um, uh, the Character AI situation from a couple of years ago where a young man tragically completed suicide- Because he was encouraged to do so by a gamified chatbot. Right. I mean, this has been weaving its way through the courts in the, in the United States, and- Right ... it sounds like, um, this, this young man's- Mm-hmm ... poor mother who's been left behind, uh, is gonna get some side of, sort of financial settlement if she hasn't already. But none of that is gonna come close to the fact that she- Right lost her son. Vicki: Yeah. Candice: Um, so when we're, when we think about generative AI and we think [00:22:00] about, you know, folks that are struggling, um, with mental health presentations, and particularly young people, we're thinking about vulnerable people- Mm-hmm ... that are engaging in a service that feels a lot like a human on the other side. Vicki: Right. Candice: And we as clinicians need to be equipped to be able to use terms like AI hallucination, uh, to be able to identify the things that are going on when we are actually correcting our clients- Right ... or pointing them in a, uh, in a better clinical direction. Vicki: Right. Yeah. Are there actually, um, better options, like, uh, of tools that are out there for- Yeah folks to use for these kinds of purposes? Candice: Absolutely. So, you know, the two that come to mind for me are Wysa and Woebot. These are, um, these are chatbots, uh, that have been designed specifically for mental health use. Um, and they have been designed, uh, by folks that have a mental health [00:23:00] background. Um, they, they s- both of them sort of work on a CBT, um, inspired, uh, basis, if you will. Vicki: Mm-hmm. Candice: Um, but they have, they have chat functions, and they have, um, they have the ability to, uh, give some light homework, if you will. Right. Mindfulness exercises, these kinds of things. Right. However, I do wanna caution that regardless of whether the chatbot is something that has been designed specifically for mental health use or not- All chatbots across the board- Right at least as the technology stands right now, are very poor- Right ... with regards to crisis response. Yeah. So thinking about folks that are thinking about harming themselves or others- Vicki: Mm-hmm ... Candice: time after time, these chatbots have been shown that they really don't do a great job of being able to address those. Yeah. So when we do think about the [00:24:00] specifically mental health-developed, um, applications, the, the Wisas and the Woebots of the world, when clinicians are thinking about recommending those to clients, because they do have, they do have good uses- Right ... they, they really do, and I don't wanna downplay that, and I'll talk about this in a second. We really want to be mindful of ensuring that it's clinically appropriate- Right ... to be recommending these to clients. Right. We don't wanna be recommending these to clients that are in crisis. We don't wanna be recommending these to clients that have a history of suicidality. Um- Yeah ... these are not going to be appropriate. Yeah. However- Mm-hmm ... thinking about clients that work irregular schedules, I'm thinking about some of the hardworking Canadians in our northern communities that, that work in camp, uh- Right ... you know, uh, in oil and gas, for instance, that, that are gone, uh, you know, for, for three weeks out of every month, shall we say. Um, you know, those folks that [00:25:00] aren't able to get into, um, you know, virtual environments for- Right ... for in-person therapy or- Right ... or actually in-person therapy, these are things that can really help them with, with- Vicki: Hmm ... Candice: um, you know, low-level depression and anxiety. Again, as a clinician, you're gonna want to, uh, really think about whether it's clinically appropriate- Right and always use your clinical judgment- Vicki: Right ... Candice: um, uh, with, with regards to making those decisions. Right. Um, but, but they do have really incredible uses, and they are really fantastic. Uh, but the technology... Not but. And the technology- Hmm ... is a little bit like fire, right? Right. Fire is, is an amazing human invention, it, it, at least in terms of being able to control it, right? Right. You know, we... It keeps us warm. It cooks our food, so on and so forth. However- Vicki: Hmm ... Candice: if we don't pay attention to it- Yeah ... it can burn our house down. Yeah, Vicki: yeah. Candice: And it's a little bit the same way with AI. Vicki: Okay, okay. And- [00:26:00] Um, and it occurs to me, you know, that as you say, like, f- suggest for us as counselors to be able to determine if this is clinically appropriate or not, um, perhaps an important step, 'cause I haven't heard of these, uh, that you're mentioning. And so, uh, perhaps an important step might be to go check it out ourselves, try it out ourselves, get familiar with it, kind of engage with it in that way. Yeah? Candice: Absolutely. In the same way that you as a clinician would not go and, you know, read a short paragraph about CBT and go- Yeah ... "You know what? I think that fits for this client. I'm gonna s- go send them off to, to, to, to engage in this CBT workbook," for instance. Right, right. As much as that wouldn't be clinically appropriate and that wouldn't be ethical or responsible- Right ... it's the same sort of thing. Yeah. You wanna be sure that you know what you are recommending, um, when you're, when you're sending- Yeah your clients off to, to check these things out. Um, that being said, though, there's, there's also [00:27:00] AI-powered apps that are broadly wellness-focused but are not therapeutic per se- Right ... that, um, that can be really helpful for clients. So I'm thinking about, and I don't... You know, there, there's so many examples out there. You, you really could- Yeah ... you know, Google search or just scroll through- Yeah ... the App Store to be able to find them. Um, but, you know, thinking about, thinking about, uh, journaling prompts, AI, uh, generated journaling prompts that you can- Mm ... point your clients in the direction of. Um, you know, guided meditation. Mm-hmm. AI-powered guided meditation. The, these kinds of things can be, can be really helpful. Right. And again, you want to, um, we really wanna caution folks around the use of free apps, okay? Mm. Not that you shouldn't use free apps, but that ultimately when something is free, you are paying for it in a different way, okay? Mm, okay. So this is another piece that I'm gonna weave into- Okay ... using, uh, general [00:28:00] purpose AI and, and why a lot of people tend to use general purpose AI, um, over these, you know, uh- mental health specific apps is simply because the accessibility. Right. They are inexpensive if they are not free completely, um, which also brings up pretty significant privacy and data sharing concerns. Yeah. And sometimes when folks are sitting there using the free or low cost version of ChatGPT or Claude or Gemini, and they're talking about, you know, their, their deepest, darkest secrets- Mm-hmm ... or the things that have been, that have been really bothering them, what they're not realizing is that their information is being shared. Their information is being used in ways that they didn't understand. Now, this- Right ... isn't inherently wrong. I wanna be super clear about that. Where it becomes problematic is where when it's happening and you don't know that it's happening. Vicki: Right. [00:29:00] Yeah, yeah. Okay. There's a lot to unpack there potentially. Yes. Candice: Yes. Um- Yes, there Vicki: is ... uh, yes, there is. And, and, um, so but I also wanna come back to, I don't wanna lose track, you had mentioned- Mm-hmm ... kind of the two main ways, one being as we're describing- Mm-hmm ... and discussing how our clients might be engaging with this, but I also wanna make sure we loop in how clinicians might be using it. And, and whoops, part of it is exactly what you were just talking about, like, um, uh, clinicians also may be using AI- Mm-hmm ... to generate a mindfulness script to use with their client, that sort of thing. Um- Absolutely ... as well as the administrative piece that you were mentioning, which is maybe a whole other- Absolutely. Yeah. Candice: Mm-hmm, mm-hmm. So, so when we're thinking about clinicians using ChatGPT, one of the big things that I would always say to clinicians is do not, in big bold letters- Mm ... do not enter client information- Right ... into ChatGPT. [00:30:00] Yeah. Yeah. Uh, it doesn't matter whether it's the free version, the $20 version- whatever. Vicki: Yeah. Candice: Anonymized or not, and this is where folks sometimes- Mm ... get hung up because they go, "Ah, but I'm anonymizing the information. It's not that big of a deal." Vicki: Ah. Candice: Mm, it is a big deal, and I'll tell you why. Okay. Because that information isn't yours. Right. You may be the steward of that information because your client has shared it with you in a clinically, uh, appropriate setting, but that doesn't give you license to then go and disseminate that information into AI systems without your client's knowledge and consent- Right even if it's anonymized. Vicki: So when you say that- So- Or maybe you wanted to... Like, when you say anonymi- like, so this, does that mean, like, I'm not gonna use my client's name, I'm not gonna put in their address or any per- or even their age maybe, but maybe I describe their situation in a general way. Mm-hmm. Is that... And that's still their, that's, I know, so that might be an example of me thinking I'm anonymizing. As an Candice: AI ethicist, it would be my recommendation, my strong recommendation- [00:31:00] Yeah ... that that is not something that clinicians do. Right. Vicki: Yeah. Candice: Yeah. Yeah. Yeah. That is not something that clinicians do, not with general purpose AI. Yes. Yeah. That makes good sense. Now, if you have an application, an AI application that you're using that is designed specifically for mental health purposes- Right ... and you have s- received meaningful informed consent from your client, then we're in a completely different- Right ... ballpark. Yeah. Yeah. But if you're using, if you're using ChatGPT to support, for instance, components of your clinical practice- If it includes any client information, I'm going to strongly caution you- Yeah against doing that. Yeah. Um, again, you don't have consent. Um, and, and there are s- pretty significant privacy and, and data handling concerns that go along with that. Right. However, ChatGPT can be absolutely amazing for developing things like templates- Right ... and, uh, worksheets and developing, um, [00:32:00] uh, you know, uh, de- developing, uh, you know, a, a, a form, for instance. Right. A general purpose form that you're going to use- Right ... for your, uh, for your clinical practice if you're- Right ... maybe rejigging your, um, your, uh, client, uh, agreement and consent form. Right. You know, something that, that really is quite generic- Mm-hmm ... um, ChatGPT is amazing for. Right. Now, again, I'm gonna come back to this. I've said this before, I think, in this conversation, but I'm certainly-- I say it every single time, clinicians are responsible for every single AI output that they go on to use in any way, shape, or form. Whether they rely upon it for a decision or they actually use the output in its entirety, you are 100% responsible for the accuracy and the appropriateness of that output. So blindly, [00:33:00] blindly thinking that AI is giving you the most accurate information- Vicki: Right ... Candice: is, is just simply not correct, and it's not something that is going to be acceptable. Right. So that's one way. That's one way. Yeah. The other way is that folks are using, folks are using AI for workflow support and documentation support. Right. Mm-hmm. Again, everybody should be using this. Um, everybody should be using this with meaningful informed consent from their clients. No client should ever find out that, um, that... Or be surprised, more to the point, that AI has been involved in any part of their service provision. But there are really, um, there are really great, um, AI-powered applications out there that may or may not be compliant with, um, with the, the data handling, um, expectations and legislations- Right in your particular province or community, so you're gonna wanna look into that. But they are really great applications. I'm thinking about Mentalyc, for instance. [00:34:00] Mentalyc is an amazing AI platform for therapists in particular that actually supports, um, it supports, uh, um, like generating clinical notes and summaries, for instance. Vicki: Mm. Candice: Um, Upheal is another one. Upheal, uh, really does promote its services to include things like AI notes. Scheduling is another one. Billing, customizable templates like SOAP, for instance. Right. Um, you know, these are, these are specifically designed for mental health applications. Um, and you're gonna want to, as a clinician, make sure that these are compliant with, um, with your data handling legislation- Right depending on what, where you are. Vicki: Right. Okay. So- Let, I just wanna slow down for a second and ask something because, um, 'cause I think it's so easy to... You're saying so many important things. Um, and I have about 50 questions about each thing. But I, [00:35:00] but I think what you just said is it kinda, like, it, it landed, right, where you said where each clinician is 100% responsible for their use- Mm-hmm of any AI-generated output- Mm-hmm ... that they rely on- Mm-hmm ... in their work. And so, okay. Yeah. So if I, I, if I pick that apart a little bit, it's, it makes me think about, okay, so if I, if I, let's say I go to ChatGPT or something and, and just, um, come up with, ask for a worksheet on, like, CBT, give me some questions to use in my next session, that sort of thing. Mm-hmm. Um, now how... N- now, there's a part of my brain that thinks, "Okay, now how is that different from we've all been googling things for a long time?" Um- Mm-hmm ... but is that the kind of thing that you're saying? Like, we, we should be more, uh, upfront and transparent, even if we're using it with, to, like, with our clients, get their informed consent if we're using these tools to help us [00:36:00] generate, um, suggestions, and ideas, and resources in our therapy, therapy work. Candice: I love that you asked that question because that's a really, uh, that's a really important clarification point. Okay. So when I say we need meaningful, informed consent from clients to use AI in their service provision, I'm talking about when AI is touching their personal information. Vicki: Okay. Candice: Yeah. If you as a clinician are using AI to support you in, uh, developing resources and materials that are, that are not client-specific, right? So we're not, again, we're not using any client information- Okay ... anonymized or not. It's just something- Yeah ... you know what? I need a new, I need a new worksheet, uh- Right ... to give out to clients that reminds them of interventions that they can use when they're having an anxiety attack, for instance. Right. Yeah. And you go onto ChatGPT, and you use ChatGPT to develop that. Right. That's not necessarily something that- Okay ... you [00:37:00] need to get meaningful, informed consent for. Right. Um, that is definitely something that you're not just going to take off of ChatGPT, not look at, and then start handing out to folks. Right. Yeah. Because ultimately, you need to look at that, and you need to be able to say, "Is this correct?" Vicki: Right. " Candice: Do I stand behind this?" Right. "And do I believe this? And where do I need to edit this, if possible?" Vicki: Right. Yeah. Candice: So, so that's a really... I'm glad you asked that question- Yeah ... 'cause that's actually a really important distinction. Vicki: Yeah. Yeah. And I- Thank you for that clarification. And, and also because I'm-- You know, as I'm listening, I think there's so many of these, um, steps, to put it that way, that I think we all just kind of- Mm-hmm ... we, we can blow past really quickly, right? And not even think about it. Mm-hmm. And so, um, I think that's a big lesson in all of this, is how do we-- There's a, there's a lot of speed and urgency inherent- Yes in this technology . Um, and so how do we slow ourselves down and make sure we're using it thoughtfully? Candice: Well, y- And you're actually touching on one of the things that I [00:38:00] am most concerned about. Okay. So if you would've spoken to me back in, um, back in tw- you know, the beginning of 2023 and said, you know, "Candice, what are you worried about?" I would've said, "I am worried that, um, the mental health field and clinicians are gonna be left behind because they are not, um... They're, they're too cautious around, um, around, you know, learning about and integrating technology into, uh, you know, into their practice and into their, into their lives." That was in 2023. If you asked me that question today, my answer is AI literacy. I was going, "Come on, come on, come on, guys, get excited about this." And now, three years later, I'm going, "Okay, everybody, I'm-- thank you for being excited." "Now let's all take a deep breath- Right ... and let's talk about what we actually need to know." Because it's very easy to all of a sudden be enamored by the sparkly aspects- Right. Yeah ... of, of AI and completely forget about all the [00:39:00] risks. Right. And the truth is, is that there are risks. Vicki: Yeah. Candice: And that doesn't mean that it should be avoided. It just means that we need to be mindful of it. Vicki: Right. And develop that, that AI literacy, as you were saying. Yeah. Candice: Absolutely. Yeah. Absolutely. Okay. So, you know, d-does AI literacy mean that you need to go out and get a data science degree? Absolutely not. Does it mean that you need to go get, uh, you know, a two-year diploma in, in AI? Absolutely not. You do not need to become a technologist in order to be able to use AI. But there are really amazing, um, educational opportunities on platforms like Coursera and LinkedIn Learning that are going to be able to help you understand some of the ways that AI can pose, um, a bit of a risk that need to be considered and lead to-- need to be mitigated against when you're thinking about using it, not just in your clinical practice, but actually for you as a person as well. Right. Vicki: Absolutely. Candice: And, and, and, you know, even, [00:40:00] even on YouTube, I-- You don't have to put out a bunch of money. Coursera and LinkedIn Learning definitely have payer-type options. They tend to be a little bit lower cost, but they also have free options as well. Um, and just, you know, building up your, building up your AI literacy is, is a big piece of that. But even then, once you've built up your AI literacy, thinking about things like starting with the use case, not with the tool itself. Okay? Okay. So when I, when I, what I mean by that is we don't wanna be finding an exciting AI application and going, "How can I bring this into my practice?" What I would encourage mental health clinicians to do is to take a step back and go, "Okay, I'm thinking about using AI. What is a problem that I could possibly solve with AI?" Hmm. "Am I looking for, um, am I looking for, uh, documentation support [00:41:00] to reduce the time that I do that to improve efficiency? Or am I looking to create psychoeducational handouts?" Okay? Right. Those are two very different applications, and they come along with two very different risk levels. And that is step two. So once you've identified what you actually want to do, what the problem is that you want to try and solve with AI, then you're gonna wanna think about what is the level of risk that comes along with that? So it's a pr- it's a fairly low level of risk when we're thinking about drafting worksheets, um, without client information to be, to be very clear. Summarizing public research is another one, um, and developing administrative text, for instance, that is not client-facing. Right. We're talking about a low level of risk. Okay. We're not talking about no risk. Right. We're talking about low risk. Mm-hmm. Right? Moderate risk, that's when we're heading into things like [00:42:00] summarizing session content- Right creating treatment plan drafts, um, using AI to develop things like client letters and what have you. Right. Um, and then of course there's high-risk applications like using AI to assess suicide risk, for instance. You know, so there's different levels of risk, and once you've identified where the level of risk is that the problem sits that you are trying to solve or you would like to solve with AI, then comes the question: Do I have the competency- Vicki: Hmm Candice: and the capacity to tolerate this level of risk? Right. Because having a high level of risk isn't a bad thing necessarily. Having a high level of risk without competency and without the comfort level in tolerating that risk level is absolutely problematic. Vicki: Right. Whenever you're ready. Mm-hmm. Candice: So s- so step one really is, like I said, it's identifying the use [00:43:00] case, not the tool. Making sure that you have identified what it is that you want to use it for, not the tool that you're then going to try and fit into your practice. From there, you're gonna go to step two, and step two is to classify that level of risk, okay? So again, that's that low risk, the moderate risk, and the higher risk applications. And you're really gonna wanna take a moment to think about after you've identified the t- the use case and the level of risk, do I have the competency or alternatively, am I willing to develop the competency- Yes ... to be able to tolerate this level of risk? And if the answer from there is yes, I can tolerate this level of risk, I do wanna start doing this, my step three would be head on over to the BCACC's, um, guidelines, uh, AI in Clinical Practice. Okay. This is a document that I wrote, um, and was released in March of 2025, and it is an [00:44:00] incredibly accessible, uh, step-by-step process of really what you need to consider when you arrive at that place of knowing the problem that you wanna solve and understanding and accepting the level of, uh, risk that is associated with AI being used to solve that problem. Vicki: Yeah. Candice: Um, you know, this includes things like considerations around fairness, accountability, transparency- Vicki: Mm ... Candice: confidentiality and privacy, and of course, competency that we've talked- Right. Yeah ... we've been talking about. Vicki: Yeah, so really diving into that ethical, the- Yes ... the meat and potatoes of the ethical stuff. And just, just so our listeners know, we will include, um, the, that, that, a link to that document you're talking about in the show notes so people can find it easily and aren't trying to scramble to write down the big, long acronym you just used. We'll have it there in the show notes for folks. Um, and, and that's really exciting, and it sounds like, um, as you just described, that that starts to lay out, um, in, in a framework that as clinicians [00:45:00] especially, we're already familiar with thinking about codes of ethics and- Mm-hmm you know, ethical practice through that, those kinds of concepts. So I really appreciate that. And, uh, yeah, and really we've already been talking about, you know, that we'll probably follow up with, with our own conversation to unpack that together. So in the meantime, folks can start to get familiar with, with what you've written there. That sounds really great. Candice: Absolutely. Vicki: Yeah. Candice: Wonderful. Vicki: Yeah. Um, Candas, this, as, as I anticipated, um- this conversation has, again, sparked a lot of new questions. It's, there's clearly lots more to talk about, and as I already just commented on, we've alread- we're already planning to have a follow-up conversation for our listeners, so you can watch for that. Yes. And, um, m- s- but before I let you go t- today in this conversation, Candice, I mean, for you in particular, I mean, you, you, you clearly have excitement in, in, on this topic, and you're always- Mm-hmm ... learning about it, it sounds like. But I just wanted to ask you, like, um, what are you excited about, just [00:46:00] you personally in this whole realm? What are you particularly kind of fascinated by or excited about in terms of what's happening right now in the- Mm-hmm ... in this whole realm? Candice: Yeah. So I guess what I'm excited about when I think about applications- Yeah ... in mental health is actually the integration of AI technology into, uh, virtual reality and augmented reality technologies. Oh, okay. You know, I'm thinking about, I'm thinking about our folks, for instance, who suffer from really debilitating, uh, social anxiety, for instance. Folks that are just not even able to, uh, maintain employment that isn't remote or, or go out and maintain relationships. I'm thinking about how virtual reality with, uh, AI-powered technology in it is, would be able to create, uh, a learning environment where a client would be able to test drive new skills in a low- Hmm uh, certainly monitored by, by- Yeah ... a mental health professional and supported through the process, but, but, uh, is able to test drive [00:47:00] different skills and abilities and push themselves to, um, to, to walk into virtual coffee shops, for instance. Um, and AI is integrated to be able to simulate, um, you know, the, the wind blowing as the, as, as you're walking up to the door of the coffee shop- Yeah for instance. Um, and you know, all of those things that are really gonna be able to help folks to, to be able to build skills. PTSD is another, is another- Oh ... really, uh, good, um, application that I, that I'm thinking of that I, that I hope to see more of in the future. What am I excited about as a general statement in AI? I gotta tell ya, the buzz around AGI is huge right now, and for folks who don't know what I'm talking about, that's artificial general intelligence. Mm-hmm. And that is a totally different level of AI and probably a topic that we could go down a huge rabbit hole in. Uh, but that is like, it, that is something that is theorized to be next level AI. That [00:48:00] is the thing that all of the big tech companies- Mm-hmm ... have been leaning hard into since ChatGPT hit, hit the ground running in, in 2022. And what, one of the things that we know is that there was three statements that were made at, um, the end of Ap- uh, sorry, not the end of April, the end of March. Um, one was by the CEO of Nvidia Um, and his position on the Lex Fridman podcast was that AGI had been achieved. Now, he used a fairly narrow, um, uh, you know, definition of AI that I think a lot of people would disagree with, but it was still a pretty profound statement coming from some- Right. Yeah ... the CEO of Nvidia. Then there was Mark Gubrud, who is a, um, a physicist, if my memory serves me correctly, but he actually coined the term AGI, artificial general intelligence. He tweeted, and I believe it was the same day, he tweeted that AGI had been achieved. Then the very next day, [00:49:00] Sam Altman makes a bunch of announcements about OpenAI and s- announces amongst those, those announcements, one of the announcements was that his product team Was being rebranded as the AGI deployment team. Vicki: Oh. Candice: So what that says, what that at least seems like it's communicating, is that OpenAI either has AGI or it's imminent. Now, there's largely been radio silence since then, but those are three really big pieces that in the tech community, very big news. Very big news at that time, and I'm excited to see, uh, what actually comes of it. Well, and, and what would that actually mean- Though I'm not sure I'll hear anything. Vicki: Yeah. What would that actually... Like, how would we actually see that happening in the world? What would actually happen if, if that Candice: indeed is- Who knows? Oh, y- Who knows? That's part of the excitement. This is, this is, this is it, right? Oh. And, and I would suspect that part of the reason why, uh, if not the entire reason why we have heard absolutely nothing since then, is that if [00:50:00] AGI has been achieved, um, the proprietary nature of that technology, um, is beyond words. Uh, so I'm sure that there are folks that are working behind the scenes to ensure that, uh, eh, that technology stays out of the hands of nefarious folks. Right. And certainly, um, you know, remains profitable for whoever reached, uh, that little goalpost first. Right. But it's very exciting in the world of, in the world of, of technology and AI. Vicki: Okay. Well, you have successfully brought us to the edge of the cliff with you. In a virtu- in a, an absolute cliffhanger there. And, um, so that is exciting. And now w- I don't even know what you're talking about really, but I'm gonna be watching for it, and it'll be exciting to see. I'll be much more attuned if I hear that term, that's for sure. Candice: Fair enough. Vicki: Candice, I just wanna thank you so much. You are clearly just a font of information on this topic, and I'm so appreciative of you giving us some of your very precious time. Um, and, [00:51:00] and giving us such an accessible and engaging conversation to help us all, you know, come along with you a little bit and, and to understand, um, this really important topic and get us thinking about it a little bit more. And I am very much, um, appre- looking forward to a second conversation with you. Um, so thank you so much. Thank you for your kind Candice: words, and thank you for having me. And I too am looking for the second part of our conversation. Vicki: All right. Wonderful. And to you, our listeners, thank you so much for joining in. I hope you have enjoyed this conversation, and come on back for the next one in a month's time. And still, so until then, be well and stay curious Eric: Looking for free resources and training in the areas of trauma, counseling, and mental health? Be sure to check out our website at [00:52:00] www.ctrinstitute.com