Dr. Jeremy Sharp (01:19)
Hey folks, welcome back to the podcast. Today’s conversation centers around AI. We’re having another conversation on AI. This time I have a guest with me to chat about AI. It’s Dr. Adam Lockwood. So Adam is an associate professor of school psychology at Kent State University and founder of Lockwood Educational and Psychological Consulting. His work focuses on the integration of artificial intelligence into psychological practice.
Including AI-assisted report writing, ethical and governance frameworks, and AI-supported research methods. He serves as a clinical subject matter expert for PAR, where he contributes to the development of AI supported psychological assessment tools. Adam also chairs the National Register of Health Services, Health Service, Psychologists, AI, and Emerging Technologies Committee, and has served on AI task forces for the National Association of School Psychologists and the Ohio.
School Health Services Association. In 2025, he was selected for the Open AI Academy, Professors Teaching with AI, and his article examining the capabilities of GPT-4 to write an APA-style school psychology paper was named the 2025 Article of the Year by Contemporary School Psychology. So suffice it to say, Adam has the credits to have this conversation. he is one of the experts in this area and
As you’ll hear on the podcast, I am just super grateful to have folks like him who are actually doing research in this area and putting data to frankly a lot of opinions. you know, we all have thoughts and feelings around AI. We may read things in the headlines, but to be able to talk with someone who’s truly on the front lines and digging into the research as it pertains to assessment is super valuable. So we talk about lots of different things. That’s the
Benefit of talking with someone like Adam, who’s deep in the in the research in this area. So we generally talk about the content of his existing papers up to this point. So, you know, how ChatGPT did at writing an APA style paper a few years ago. We talk about his more recent paper on writing psychological assessment reports and comparing human writing to AI writing.
We talk about more, I would say philosophical questions around, you know, risks of AI like de skilling and how it is potentially making us rustier at the work that we do, if we’re not deliberate. We talk about bias in AI, what Adam is seeing in that regard. and we also talk about a number of strategies to utilize AI effectively and wrap it up with a discussion of
Concerns and reasons that people are not adopting AI and you know ways to think about AI and our practices. So there’s plenty to take away from this discussion. And if you’re interested in AI, if you’re well versed in AI, I think there’s a little something for everyone. So please enjoy my conversation with Dr. Adam Lockwood.
Dr. Jeremy Sharp (04:33)
Adam Hay, welcome to the podcast.
Dr. Adam Lockwood (04:36)
Thanks for having me.
Dr. Jeremy Sharp (04:38)
Glad to have you. Yeah, yeah. This is one of those kind of happenstance meetings that turned into a podcast interview, which which I appreciate. We got to connect at NASP, I think, a few months ago and ⁓ yeah, chat a little bit. So yeah, I’m I’m thrilled to be talking with you just because you are I think one of the folks who are doing great research into AI and practice and school psych, and there’s not a whole lot of research out there at this point.
Dr. Adam Lockwood (04:49)
Yeah.
Dr. Jeremy Sharp (05:05)
Good research anyway. So yeah, grateful to have you here.
Dr. Adam Lockwood (05:08)
Thank you.
Dr. Jeremy Sharp (05:10)
So I’ll start with a question that I always start with, which is why is this important to you of all the things you could presumably spend your time and energy on with your work and your life? why this?
Dr. Adam Lockwood (05:21)
Well before I get started, is there if I do my conflict of interest? Yeah, yeah. ⁓ so whenever you know you listen to someone talk about AI and and they’re they’re especially if they’re very pro AI, it’s important to know kind of what interest they have in in promoting AI. And so I I do have a consulting business and I work with various companies, most mostly with par on their
Dr. Jeremy Sharp (05:24)
Yes, of course. Yeah, let’s get that out of the way.
Dr. Adam Lockwood (05:47)
AI report writer and so I receive you know compensation for that. So I just want be upfront about that. They’re I I think they’re a great company, but yeah just want to be clear about that. Why AI? So I my my original research was an assessment. And in academia after a set amount of years you go up for tenure and promotion.
Dr. Jeremy Sharp (06:03)
Mm.
Dr. Adam Lockwood (06:09)
And I was at that point and I knew that I wanted to change to something different. I still really enjoy assessment, but I just wanted to try something new. And I knew that I was pretty sure that I had enough pubs for tenure and promotion. And so I was I was ready. I was ready.
And 2022, Chat GPT came out. I started messing around with it and was pretty quickly convinced that this was going to be revolutionary, disr disruptive, whatever you want to call it. And so I just jumped right in and started doing my first paper. I think it was published in 2024 and just started doing research and just being fascinated with the with the technology. So that’s that’s really it. I I see this as just being something that’s going to change our lives, our society, and practice.
So I wanted to be part of it and helping to, you know, be a part of that conversation.
Dr. Jeremy Sharp (07:07)
Mm, yeah. Yeah. I we have a lot in common already, you know. So I’m totally on board with you. I’m an AI you know self-disclosed AI optimist, right? And I do think there’s a lot of potential for it. and caution, of course, but I am optimistic about it. But I’m a little jealous, I have to say, just because you were in a position to actually dig in and like do the research, you know.
I feel like I do a lot on the practice side and of that. But you know, to be able to have the infrastructure and time and resources to do the research and do good research is such a such a gift. And I feel like we need that in the AI conversation. So I’m really glad that you decided to make that career pivot for everybody’s sake.
Dr. Adam Lockwood (07:48)
Well, you’re always welcome to join in. It’s not a paid it’s not a paid endeavor. That’s that’s the thing. ⁓ but it it is fun and rewarding, but yeah.
Dr. Jeremy Sharp (07:51)
Ha ha ha ha.
Sure, sure.
I believe it. I believe it. Yeah. I have to chat with my wife about taking on another project that doesn’t pay me. So I’ll get back to you. I’ll get back to you. Yeah. So tell us just briefly, what was that first paper? Like how did you wade into this area?
Dr. Adam Lockwood (08:06)
Yeah.
So the first two papers were essentially, boy, this is good at writing. I wonder how good it can do. Is there a way to look at that empirically? The first one was like, I wonder if I could just have it write a school psychology paper and proper APA. And so I did, and then I wanted to look at how good of a job did it do? Was it factual in the information that it was generating? Was it generating hallucinations, which are just, you know, we’ll just call it misinformation or false.
in information and then how much of what it was saying could be backed up. This is before Notebook LM or retrieve augmented generated, you know, grounded models where you could kind of see where information came from. It just it just it was a black box, you know. It still is to some extent, but it was really was then. And so I had it generated paper, pretty easy prompt.
Dr. Jeremy Sharp (08:57)
Mm-hmm.
Dr. Adam Lockwood (09:09)
And and then I went through with a colleague of mine and we picked apart every sentence. Is this sentence true? If it is, is the citation to where it says from where it where it says it is. And what we found is first of all, it did a good job of APA report writing. And this is like using 3.5 or 4, I forget. Totally outdated model. but what we also found is that its citations were just completely made up at that point. And so
Dr. Jeremy Sharp (09:28)
Mm-hmm.
Mm-hmm.
Dr. Adam Lockwood (09:38)
You did good job, everything seemed correct or was correct, but it just wasn’t pointing to where it came from. It just was totally it just it was horrible at sighting at that that time. That’s changed since then. You know, it’s gotten better, it still makes mistakes. But for example, if you’re using Claude Cowork, do you use Claude Cowork? Claude Cowork is a great writer, and if I point it to
Dr. Jeremy Sharp (09:51)
Mm-hmm.
I do.
Dr. Adam Lockwood (10:04)
a folder that has articles and I say here is a school psychology paper that I want you to use as a template, here are the folders I want you to pull from, use proper APA. It’ll write a paper from start to finish that’s pretty much perfect. Or just needs a not perfect but needs a little bit of it’s better than a graduate student would do for sure. And it’s it just needs a little bit of tweaking.
Dr. Jeremy Sharp (10:20)
Yeah.
Dr. Adam Lockwood (10:28)
I’m sure people are doing it, and it’s gonna become more and more common just because there’s information that is beneficial for people and a lot of the time involved is just writing up literature reviews and things like that that take a lot of time that AI can do well. So that first paper was can AI write a report? write a paper? Yes, it can.
Dr. Jeremy Sharp (10:48)
Mm-hmm.
Dr. Adam Lockwood (10:49)
Could it cite no, but that is no longer the case. And I haven’t examined it empirically since then, but anecdotally, I can make it write papers if I want to. ⁓ and then the second paper was can AI write psychol psychological reports? And that was using ChatGPT-4, so it wasn’t any specific, you know, fine-tuned model or anything. It was just regular frontier commercial AI that we all have access to.
Dr. Jeremy Sharp (10:58)
Yeah.
Dr. Adam Lockwood (11:19)
And yeah, I can write reports that time too. And so the design was was essentially four conditions. So had mock data, made up student data, child data actually, and fed that to licensed psychologists, gave it to licensed psychologists and fed it to ChatGPT, and to sat them write reports.
Dr. Jeremy Sharp (11:24)
Can write reports. Yeah.
Dr. Adam Lockwood (11:42)
So it wasn’t school psychology specific, it was clinical psychology. And the only reason why we went towards clinical and not school is we had to have people rate those reports. And you know, you’ve worked in clinical settings as well as school settings, right?
Dr. Jeremy Sharp (11:58)
Actually, no, I’m I don’t have any school background, like formal school background. I’ve only been in private practice and I’m a clinical, you know, a clinical practitioner. Yeah.
Dr. Adam Lockwood (12:05)
Okay, okay, I’m sorry.
So you’ve read school psychology reports before though, right? How long are they? Yeah. And so when when writing your clinical reports, are they 30 pages long typically? So so for that reason we we went the clinical route because there’s no way we could have people
Dr. Jeremy Sharp (12:10)
Yes, many, many.
They’re interminable. yeah, they’re super long and super dense. Yeah.
Not anymore. No, no.
Dr. Adam Lockwood (12:31)
in a study read a 30 page paper. It just wasn’t gonna happen. And so we went we went we went for a clinical psychologist, licensed psychologist, not clinical necessarily, two were clinical, two were for school, but licensed psychologists doing clinical reports. We had about 250 licensed psychologists take a survey where they read the reports that were generated by either humans or AI, and everyone got a pair. So they got
Dr. Jeremy Sharp (12:53)
Mm-hmm.
Mm-hmm. Mm-hmm.
Dr. Adam Lockwood (13:01)
A human and an AI pair of reports along four conditions, either intellectual disability, ADHD, depression, or generalized anxiety. And they just rated quality of those those two reports. And what we found essentially is that there were some statistical differences, but practically, you know, the the clinical st the clinical.
You know, you have you have statistical significance and you have clinical significance or practical significance, there was very little clinical or practical significance, with a few exceptions. So though the effect sizes were small. So essentially, yes, people did prefer human reports, but not much. And in one specific area, which was recommendations, they significantly preferred AI-generated recommendations.
Dr. Jeremy Sharp (13:36)
Mm. Mm.
Yeah.
Mm-hmm. Mm-hmm.
Do you have any idea why?
Dr. Adam Lockwood (14:00)
Chris AI does a better job of generating recommendations than people do.
Dr. Jeremy Sharp (14:04)
That was my guess. Did do you were you able to get any more nuance with that in terms of like were the recommendations more evidence based or more nuanced or personalized or I don’t know if you went that deep with the you know, with the study design.
Dr. Adam Lockwood (14:19)
We didn’t ask, but when you have I’ll speak for myself because I because maybe I’m just the one bad psychologist, but I always pulled from a bank of recommendations and they tended to be pretty boilerplate and not very individualized.
Dr. Jeremy Sharp (14:33)
Yes.
Yeah. You’re not the only bad psychologist if that’s what a bad psychologist is. Yeah. We do the same, you know. Yeah. No, no, I mean we do the same. I think a lot of folks have recommendations banks. Of course. I mean, we have to do that to some degree. So go ahead though.
Dr. Adam Lockwood (14:36)
Okay. However sorry.
But if you use artificial intelligence, GNAI, large language models, it will pull from the students or child’s interests, areas of needed support, where they’re at, you know, and and generate recommendations that are very individualized, very bespoke. So it’ll be like, this, you know, this child likes music and is good at this, is struggling in this area.
Here’s the resources in their in in the town that they live. So it the the the recommendations are just better. If you look at IEP goal creation, which we see a lot more in special ed and school psychology, which is individualized education programs, the the research is the same where it’s like AI is as good or better than
Dr. Jeremy Sharp (15:25)
Pretty well more.
Mm-hmm.
Dr. Adam Lockwood (15:41)
special education teachers at generating these individualized programs and it does it in reduced amount of time.
Dr. Jeremy Sharp (15:47)
Pretty remarkable. Yeah. I haven’t looked at the schools like literature as closely, but yeah, that tracks at least on the clinical side.
Dr. Adam Lockwood (15:49)
Crazy.
Dr. Jeremy Sharp (15:57)
So were there I mean, there’s a lot of just given our conversation so far, I mean, a lot to be excited about with AI. I mean, in this paper, you know, we’ll stick with the current paper, what were the downsides that emerged? Like did you you know, when you say AI was maybe worse at some of the report writing tasks, what did that look like?
Dr. Adam Lockwood (16:16)
Well, I wouldn’t say worse necessarily. It’s just some things were rated more highly. But even in the areas where it was rated less favorable, it would we had them rate on a scale that was average, above average, and very good, I believe. And what we found is that even
Dr. Jeremy Sharp (16:29)
Mm-hmm.
Mm-hmm.
Dr. Adam Lockwood (16:41)
Areas where they preferred humans, they still said these are average or above in in the majority of the resp the the ratings from the participants. So I’m trying to pull up, see if I can find the actual. I was prepared, but I think I opened up the wrong document. That’s hilarious. let me see.
Dr. Jeremy Sharp (16:52)
That’s the same.
No worries.
Dr. Adam Lockwood (17:02)
I would give you the specific percentages, but they were they were very high. So yeah, there were there were differences, but there there was nothing that really stood out as like, this is super problematic. There was th they were they were good across, they were average or better across humans and AI.
Dr. Jeremy Sharp (17:14)
Pretty minimal.
Yeah.
Did you notice any differences across conditions like, you know, intellectual disability, ADHD, anxiety and depression, I think you said? Were there any differences in the type of report?
Dr. Adam Lockwood (17:34)
We didn’t get that much into detail because we were already doing a number of comparisons across readability, jargon, overall recommendations. And the more you start to break things up and examine, the more chance you have of of introducing error because you have family-wise error, you know? So if it’s a five percent chance that you’re going to
Dr. Jeremy Sharp (17:39)
Sure, sure.
Yeah, of course.
Dr. Adam Lockwood (18:01)
Get a finding due to chance. Every time you multiply that, that chance goes up. So we we definitely we was pre-registered that we were going to do it in in a certain way. it’s always a it’s always a warning, kind of warning bells go off for me if I see a study where there’s a billion comparisons, because you’re like, you’re you know, you’re you’re you’re kind of going on a fishing expedition, which we we didn’t want to do. We wanted to be calc.
you know, pretty calculated. So we we stuck to the questions we had and we didn’t look across conditions. But it’d be interesting to know if it was better at ID versus ADHD.
Dr. Jeremy Sharp (18:40)
Can you can you tell me about some of the other projects? Anything? I know you have a lot in the hopper right now as far as AI research. Can you like give me any, you know, overview of some of the other projects?
Dr. Adam Lockwood (18:51)
I gotta look at my Vita real quick. Under review right now is a survey of health service psychologists. So y are you familiar with the National Registry of Health Service Psychologists?
Dr. Jeremy Sharp (19:06)
yeah, yeah. I mean superficially at least. Yeah.
Dr. Adam Lockwood (19:09)
We have about 10,000 members and I think 5,000 student members. we did a survey of AI use by them. What we found is they’re more hesitant than school psychologists, adoptions lower, something like the 56% range, which almost exactly what the APA were finding. ⁓ this these are data from November of 2025. so that
Dr. Jeremy Sharp (19:29)
Mm.
Dr. Adam Lockwood (19:36)
That there’s a preprint of that out. just looking at ethical concerns and adoption rates by health service psychologists. another one looking at AI practices across allied health professionals. So OTs, PTs, SLPs, school psychologists, counselors, have something under review about bias in school risk prediction, which is pretty interesting.
Dr. Jeremy Sharp (19:59)
Yeah.
Dr. Adam Lockwood (20:00)
with I these are all things I’m doing with other people so I don’t mean to what else?
Dr. Jeremy Sharp (20:04)
of course.
Dr. Adam Lockwood (20:05)
I’ve looked at AI for use in qualitative analysis, which is interesting with some friends.
Dr. Jeremy Sharp (20:12)
Nice, nice. Can I go back and ask a couple of questions about some of those projects? And if it’s if the that the data’s not there or not ready, that’s totally okay too. But ⁓ okay. Okay. Okay, great, great. I mean, so the first one that you mentioned is a lower adoption rate among health service folks and you know, the APA group. I I mean my mind immediately goes to
Dr. Adam Lockwood (20:22)
Those are all out there on the preprints or or published, so yeah.
Dr. Jeremy Sharp (20:39)
Well, school psychologists are like completely maxed out on time and energy. So of course they’re gonna have a higher adoption rate of a tool that could save them time and mental cognitive load. I does that at all line up with what y’all are thinking or finding?
Dr. Adam Lockwood (20:55)
Well, I think one one thing is that school psychologists spend seven and a half hours a week on average writing reports. And that’s one use case. That’s a pretty obvious one. And so if if they are able to use AI and in the the the surveys we’ve done, they’re saying they’re saving something like six
Dr. Jeremy Sharp (21:02)
Mm-hmm.
Mm-hmm.
Dr. Adam Lockwood (21:18)
I think it’s 6.2 hours, I forget, six hours a week on report writing. it’s so part of it’s the nature of the job, I think. Clinical psychologists, as you know, there are some that do a lot of assessments. I was one of them in my postdoc, but a lot a lot are doing more counseling and things that they might be using AI for to assist with billing and and transcription and things like that, but they’re not just sitting down.
For large chunks of times writing AI reports. And so I think it’s a partly a function of their job. I think partly it’s a function of the might be that the laws that govern them. HIPAA is very strict. And you know, I’m not a lawyer, so I’m not I’m not trying to speak out of school, but
Dr. Jeremy Sharp (21:52)
Mm.
Mm.
Dr. Adam Lockwood (22:05)
You know, when I first took my HIPAA trainings, they scared the you know what out of me by like saying, like, if there’s a HIPAA violation, you could be on the hook, lif you could be liable for, you know, a million dollars or something. and so I think that that is a concern for our practitioners where school psychologists are governed by FERPA, which you know, my understanding is is you know, you still you know, you still have to maintain
Dr. Jeremy Sharp (22:05)
Course.
Uh-huh.
Dr. Adam Lockwood (22:32)
student data, but it’s the the penalties aren’t as harsh.
Dr. Jeremy Sharp (22:34)
Sure.
Yes. Yes. That makes sense. And what about the Allied health professions and their rate of adoption of AI? Like our OTs and PTs and speech therapists utilizing AI as much as we are?
Dr. Adam Lockwood (22:49)
That I have in front of me. So school psychologists are the biggest adopter. And I’ll just qualify this by saying this is just in the state of Ohio, but there was like 1300 respondents. so we had, you know, over a hundred school psychologists, over 200 OTs, over 200 PTs. But we had almost 80% adoption by by school psychologists as of a year ago, almost a year to the date.
Dr. Jeremy Sharp (22:57)
okay.
Dr. Adam Lockwood (23:17)
counselors are right up there too in the mid 70s. SLPs just over 70 percent. OTs around 55-ish. nurses over 40, and then PTs like 40 percent. And you know, some of that makes sense. If I don’t know if you’ve been to a PT in a while, if you’ve been to a PT in a while.
Dr. Jeremy Sharp (23:31)
Mm-hmm.
⁓
My god. Yeah, I’m a big runner, so I feel like I’m in PT like every week basically, you know. Yeah.
Dr. Adam Lockwood (23:45)
So the the the amount of documentation my physical therapist and PTA does is is less, right? They’re spending a lot more time man manual manipulation, having me pull bands, like doing all sorts of stuff. And they do notes, but they’re not sitting down in in I don’t think doing seven and a half hour reports a week, you know. Yeah.
Dr. Jeremy Sharp (23:55)
Mm-hmm.
I hope not.
Great, great.
This is exciting. so given I mean, you have a I think unique perspective, you and your colleagues of course and co authors on everything that’s happening here. I mean, from from where you’re sitting, what are you seeing at this point as maybe the most exciting parts of AI, in psychology, you know, in assessment? what’s looking good and promising?
Dr. Adam Lockwood (24:33)
Well
Whenever I say promising, I mean, I think what you said at the beginning is I don’t see to AI as being necessarily good or bad. I think it’s it’s all in the implementation. But I am excited that right now I think that people can really improve the services they provide and save time in in a number of ways using especially language.
Dr. Jeremy Sharp (24:43)
Mm-hmm.
Mm, mm-hmm.
Mm-hmm.
Dr. Adam Lockwood (25:01)
Based models, large language models, to help them with basically all that administrative burden that is a burden that nobody enjoys. So that’s exciting to me. It’s just like, hey, how can I have help writing reports? How can I have help generating soap notes? How can I have help generating emails? and how can I use that time to either reduce burnout for myself or provide
More of the services that I would like to. So for school psychologists, that means maybe a little bit less report writing, a little bit more FaceTime. I also think we can improve the services we provide. I already mentioned recommendations, but you know, the if if you use AI, you can generate images that might look like more.
Dr. Jeremy Sharp (25:33)
Mm-hmm.
Dr. Adam Lockwood (25:47)
might be more represent representative of someone you work with, right? I used to do presentations and I would struggle to find images that that weren’t just white children and ⁓ that were that were free too. And now I can I can generate an image that will look like someone I’m working with.
Dr. Jeremy Sharp (25:58)
Absolutely.
Dr. Adam Lockwood (26:06)
and we know from you know social cognitive theory that the more similar a model is perceived to be, the more likely someone is to identify with them and we see a behavioral change. So, you know, that’s that’s something that excites that excites me as well is just making things more individualized, more meaningful.
Dr. Jeremy Sharp (26:28)
Mm-hmm. Yeah, I love that. I’ve heard a lot of stories of folks who are using AI to generate almost like personal stories for kids in that you know, in that like therapeutic assessment sort of framework, like personal stories and illustrations and metaphors and you know, fables and things like that that can like bring the results to life for kiddos. I was talking to a woman the other day who’s
generating almost like full-on, you know, mini books for each kid she’s doing feedback with that has, you know, 10 or 12 pages to describe them. And that’s all AI generated, you know, a lot of image creation and text. That’s I mean it’s really exciting. I love the personalized aspect. If we can use it to personalize care for people. That’s very cool.
Dr. Adam Lockwood (27:10)
Yeah.
Jim and I has
a storybook creator that is very easy to use that that does that very well.
Dr. Jeremy Sharp (27:22)
Yeah, yeah. Yeah, that’s funny. I was just looking at that this morning, you know, in the options. I use Gemini a lot just because I’m in Google workspace quite a bit. So yes. Are you seeing other helpful implementations of AI in the clinical world that we haven’t already talked about?
Dr. Jeremy Sharp (28:57)
Hey, everyone. I’m really excited that NovoPsych Psychometrics is sponsoring the show. NovoPsych is a platform for psychologists who care deeply about assessment and testing and want their self-report measures to be the very best. NovoPsych has an extensive library of 150 standardized instruments with strong coverage across the presentations many of us assess every day, like disability, functional impact, autism, ADHD, and a wide range of symptom measures.
You can also use it for broad personality assessments like the Big Five or go deeper when you’re looking to understand personality pathology. What makes NovoPsych different isn’t just the range of scales, it is the quality of the experience. So I really appreciate the depth of psychometric info that it provides and the clear graphs and visualizations that make results easier to interpret and communicate. If you want to try NovoPsych psychometrics, you can access a 15 day free trial via the link in the show notes, which is
novopsych.com slash testing psychologist. That’s N-O-V-O-P-S-Y-C-H dot com slash testing psychologist.
Dr. Adam Lockwood (30:05)
I mean
Yeah, so so much. I’m just trying to think. So I’ve done trainings for different districts and different organizations and companies and and I and I see amazing stuff all the time. I’m trying to think of some examples, but what I’m seeing is people coding, for example. You know, they they’ll they’ll create a tracker for a behavior of interest for for for students. That’s the kind of stuff that you know w would not have seen in
Dr. Jeremy Sharp (30:08)
Mm-hmm.
Dr. Adam Lockwood (30:33)
or even imagined in 2022. But yeah, that that’s the thing that I think is coolest is is this kind of use of images, creating storybooks, things like you mentioned, and then people generating easy apps to come up with ways to collect data on behaviors of interest. Which is really easy to do. I don’t think most people realize how easy it is to talk to codecs.
Dr. Jeremy Sharp (30:54)
Yeah, yeah.
Dr. Adam Lockwood (30:59)
talk to Claude Code or even to Gemini and have it make something that runs locally on your computer so it’s you don’t have to worry about your data going out and collect collect data.
Dr. Jeremy Sharp (31:13)
Yeah, yeah. Maybe we talk about that just a little bit because I I’m guessing, you know, people like heard what you just said and they’re like, I don’t know what that means or where I would even start or you know what I would do with it. Can we break that down just a little bit? So when you say like talk to Claude Code, let’s just take that use case, what would someone do to build an app?
Dr. Adam Lockwood (31:38)
Yeah. Well I guess the first thing I should say is I I just realized I’ve said talk to Cloud Code. I don’t type anymore, do you?
Dr. Jeremy Sharp (31:47)
You know, this is tough. I do, but I have been using whisper flow or you know, dictation. That’s my like dictation of choice, dictation software of choice. And it has made a pretty big difference. So I’m still I have I’ve always been like dictation resistant. but I’m doing it more. Sounds like you are like all in on dictation or talking. Yeah.
Dr. Adam Lockwood (32:08)
Yeah, I don’t I don’t type anymore. And it actually came
about because I I had an injury and was basically unable to type for a little for a little bit. And so I started using dictation, whether it be through Word or these these large language models. But the the idea in the AI space is that if you have kind of a conversation, you actually will get
Dr. Jeremy Sharp (32:12)
Mm-hmm.
Okay.
Dr. Adam Lockwood (32:34)
A better feel for what’s going on than kind of going through the filter of typing. When you go through the filter of typing, you lose some of the information. ⁓ you end up kind of self-censoring in a way. I think that’s probably not the best way to put it, but you if you and I are just talking, I might say, like, no, no, that’s not a good idea. And and it’s a it’s a back and forth. So the the first thing I’ll say is I don’t type. If I use ChatGPT or Claude or Gemini.
Dr. Jeremy Sharp (32:38)
Mm-hmm.
Mm-hmm.
Dr. Adam Lockwood (33:02)
I’m either on my phone or on the desktop app and I’m just talking. And so that’s that’s that’s the first thing. And then if I want an app, for example, I just open up one of those and say, Hey, I’ve got a child who who is I have some on my computer I could find that I created, but you know, here’s the behavior of interest. I really want to track it. They really like.
Dr. Jeremy Sharp (33:07)
Mm-hmm.
Dr. Adam Lockwood (33:29)
Cats, whatever it is, you know, what do you have any ideas on creating an app that will run you know that I can send to them and they can run on their their machine? and it’ll be like, yeah, I would give me, you know, the thing I always do is like ask me any questions you have before you start and it’ll it’ll clarify and then it’ll just go.
Dr. Jeremy Sharp (33:46)
I love that. Mm-hmm.
Mm-hmm. Mm-hmm. Yeah. It’s pretty remarkable. It’s pretty remarkable. It is crazy. It is crazy. My my favorite use case to date has been like nothing to do with clinical stuff, but I was so proud of myself last year when I built this like playoff chances simulator for my son’s soccer team, you know, and like got to feed in all this data and you know, it simulated their playoff chances and then it turned out to be right. So, and let me buy yeah. Well
Dr. Adam Lockwood (33:54)
It’s crazy.
Whoop, what were their chances?
Dr. Jeremy Sharp (34:18)
I mean they it changed throughout the season, you know, it’s kind of dynamic based on the results of the games and I just had to like input the results and it like whatever. Made predictions and that kind of thing. so they ended up having a hundred percent chance, I guess, ’cause they made it. But ⁓ yeah. And so it allowed but you know, it la it allowed me to track it throughout the season and I bought plane tickets like way earlier than I otherwise would have because I wasn’t sure if they were gonna make it and whatever. Anyway, it’s a lot of fun.
Dr. Adam Lockwood (34:33)
Well, that’s great.
Dr. Jeremy Sharp (34:46)
It’s lot of fun to experiment. But it’s one of those very unique personal use cases and it didn’t take very long at all.
Dr. Adam Lockwood (34:46)
There’s lots.
Dr. Jeremy Sharp (34:54)
So from where you’re sitting right now, what are you seeing as the emergent or primary risks that we’re working with when it comes to AI and clinical clinical work?
Dr. Adam Lockwood (35:09)
There’s a lot of risks. Like you know, I like I said, it’s it it’s neither good nor bad nor neutral. It’s it really depends on how you use it. If you use it responsibly, then I think you can avoid the risks. If you don’t, then you you you might be in some trouble ethically. But some of the risks are things like de-skilling. Are you familiar with that term?
Dr. Jeremy Sharp (35:10)
Yeah, yeah.
Yes.
a hundred percent. I think about those a lot. Yeah, because we have interns and postdocs in our practice and I think about it for myself a lot. I’d love to let’s talk about that a little bit. Yeah. What
Dr. Adam Lockwood (35:40)
Yeah.
So de-skilling to me is just a fancy AI or tech term for like losing or or becoming rusty in skills. You know, they talk about upskilling, which is learning something, deskilling, which is forgetting or losing a skill. And so you you put it in the you you framed it as trainees. That is one thing I really worry about is if you’re not being if you’re not building those skills in the first place, you have an AI
Dr. Jeremy Sharp (35:50)
Mm-hmm.
Mm-hmm.
Dr. Adam Lockwood (36:07)
offload them, you’re never going to get the skills that you need in the first place. And so that’s a problem. There’s there’s you know pretty incipient research, but it’s from companies like Claude who you know have no reason to to say that AI wouldn’t actually hurt learning, but they are saying that. ⁓ That is if you just have AI do things for you.
Dr. Jeremy Sharp (36:08)
Mm-hmm.
Mm-hmm.
Yeah.
Dr. Adam Lockwood (36:29)
you’re not likely to to to learn. You’re not likely to be invested cognitively, which makes sense, you know. You just if you outsource it, you’re not gonna be you’re not gonna have as much
You’re not going to be as knowledgeable. So the one thing I worry about is that trainees won’t get the skills that they need. And so they’ll just dep depend on AI. And AI will generate a recommendation or any sort of output, and they won’t know enough to know that what it’s outputting is false. And then on the D-scaling side, we’ve already seen this trend with.
Dr. Jeremy Sharp (37:02)
Mm-hmm. I’m with you.
Dr. Adam Lockwood (37:07)
physicians using elec electronic medical records. And when the when the when those were introduced, they lost skills that they used to have that were being done by the electronic medical records. And so it’s something that we’ve already seen occur and and I worry about it for psychologists, that they will become rusty and then will therefore be not as able to to do things that are important for them.
Dr. Jeremy Sharp (37:33)
Yeah. Yeah.
Dr. Adam Lockwood (37:34)
The other thing I worry
about from a practice standpoint is automation bias, which is what I can s what I conceptualize as just blindly following AI output.
Dr. Jeremy Sharp (37:38)
Yeah. Tell me about that.
Mm.
Dr. Adam Lockwood (37:46)
And I’ll I’ll be honest, I’ve noticed myself doing this more and more. Luckily it’s never happened with clinical or s or student data. But you know, I have pasted an email to someone that said, you know, would you like this to be in bullet points as well? You know, like the lastly copied and pasted it. And like I said, it’s never been, you know.
Dr. Jeremy Sharp (37:51)
Mm-hmm.
gosh. Yeah.
Dr. Adam Lockwood (38:11)
confidential data, but I’m just ripping through trying to go fast. And so that is a concern. We’ve already seen it in the law field, for example, that people are, you know, being disbarred because they’re taking output from AI and they’re putting it out there is, you know, this is case law or whatever. I’m not a lawyer, but and then it’s people are finding out this was made up and they’re they’re losing their ability to practice as a lawyer.
Dr. Jeremy Sharp (38:22)
Yep.
Mm-hmm. Mm-hmm.
Dr. Adam Lockwood (38:38)
Another example
would be Deloitte. I know if you remember that story where they I think they did a they did a study for the Australian government and they used AI and there was hallucinations, misinformation in there. And yeah, pretty embarrassing. So I worry about that for practitioners.
Dr. Jeremy Sharp (38:54)
Sure.
Yeah, yeah. I was it’s interesting. I was talking with some folks in my membership community on Tuesday, a couple days ago, about this. And it’s it’s almost like the worst case scenario where it’s like this rogue version of intermittent reinforcement, where like it’s right most of the time, like enough to get a get us to trust it, and then but it’s wrong enough that it can be really embarrassing when
Then we don’t catch it, you know? So it’s like this weird, like intermittent. And so you end up either like you trust it a hundred percent and just let it run, and that’s kind of dangerous, or you end up like reviewing all the output, which also takes time and maybe cancels out any time savings that you may have had by using it. So it’s a tricky situation.
Let me can I go back to de-skilling and present a little bit of a I don’t know if it’s a counterpoint or maybe just a discussion point, but you know, I think about de-skilling, and you know, you mentioned the example with doctors and EMR automation and things like that.
The question that I ask is what skill are we losing and i or, you know, becoming worse at? And is that skill actually important? Or is it just a thing that we had to do because there was no other way to do it? Do you ever think about it that way?
Dr. Adam Lockwood (40:13)
I wonder that out every day. I I just talked to my students on Monday. I’m like, no, Tuesday, I said, so what skills do we need to have?
Dr. Jeremy Sharp (40:23)
Yeah.
Dr. Adam Lockwood (40:24)
Do we need to be able to read and write?
Dr. Jeremy Sharp (40:26)
It’s a great question. And it sounds completely insane to even ask that, but
Dr. Adam Lockwood (40:28)
You know, people are gonna
think I’m crazy, but I just told you I don’t type anymore. And I can talk to AI and we can have a back and forth. And it’s faster to talk than it is to type.
Dr. Jeremy Sharp (40:34)
Right.
Yeah.
Mm. Mm-hmm. Right. Right. And I’m guessing you’re not doing a lot of writing by hand these days.
Dr. Adam Lockwood (40:48)
Yeah, I mean I I’m just it’s it’s talking.
Dr. Jeremy Sharp (40:53)
Yeah. Yeah. It’s a good example. It’s a good example. And you know, I even I think about it in the context of our reports, of course, and especially with trainees, that’s a that’s a big lens that I look through. But it’s kind of like, Okay, what aspects of writing this report are we actually trying to develop? You know, like is it important to know how to write a background when you’re kind of like regurgitating information you’ve already gathered? I don’t know.
I mean I guess it is to some degree, but I don’t know that we need like a prosaic background section that is easy to read. I don’t know. that’s just one example. So yeah, this comes up a lot for me. I really think about what skills are we do we need in this conversation.
Dr. Adam Lockwood (41:35)
Yeah.
And part of it is to what extent is the writing process helping us to case conceptualize?
Dr. Jeremy Sharp (41:44)
Yes, yes. Right. Cause I think we know that, right? I mean writing does help process. Yeah. There’s a layer of thinking.
Dr. Adam Lockwood (41:50)
Mm-hmm. So if we’re
if we’re offloading that, are we getting to know patients, students, clients that we’re we’re working with as well as we would have if we would kind of wrestled with stuff in the writing phase? I don’t know. My guess is that this t in some cases probably not.
Dr. Jeremy Sharp (42:07)
Mm-hmm. It’s a break.
Mm-hmm. Mm-hmm. I think you’re right. I don’t I don’t know what to do with this exactly. And obviously it’s a it’s a data point of one, but you know, for a while there, I was just I moved to recording my intakes and just having them transcribed. And, you know, so I stopped like writing or typing notes during my intakes. And I found that I I was missing something, like some depth or you know, my brain wasn’t
going, you know, generating more questions or deeper question. And so I’ve gone back to now, I mean, I still record and transcribe them, but I’m also typing notes, which is super redundant, but it has helped me, I think, you know, generate more questions and think more critically about the client. I don’t know if that makes any sense or what’s going on there.
Dr. Adam Lockwood (42:58)
No, it completely
does. And I still take notes by hand, sparse, and then I go back and have transcripts and and summaries generated for me by AI for important meetings. And but but it does free me up to be more present and to not be taking trying to take furious notes and just to do a couple of things that I really want to stick and then letting the rest of it I can come back and revisit stuff in those transcripts.
Dr. Jeremy Sharp (43:02)
Mm.
Mm-hmm, mm-hmm.
Mm-hmm.
Yeah, yeah. Maybe that’s the way, you know, combo is the way to do it. It seems to be working for me at least at this point. Yeah. So with the I would love to hear more about this discussion around skills that you were having with your students or just de-skilling in general and how you’re thinking about, you know, combating that when AI is becoming so rampant.
Dr. Adam Lockwood (43:46)
So that’s the that’s the tricky thing. You know, one thing is you know
Dr. Jeremy Sharp (43:48)
Yeah.
Dr. Adam Lockwood (43:51)
You said like what things do we need to get rid of? I was thinking about transforming scores from raw scores to standard scores. And there was a point where that was considered like a c a critical clinical skill that was necessary. And to me, I don’t I don’t think it is. Like in training programs, they still make you do it, you know, like you have to line up a piece of paper and and there’s there’s some reason that they give you, but it never made sense to me.
Dr. Jeremy Sharp (43:57)
Yeah.
Mm-hmm, mm-hmm. Mm-hmm. Mm-hmm.
Right.
Dr. Adam Lockwood (44:20)
So I’m never sure what’s gonna go away, but I I I I I do worry. I do worry and I think if my students are gonna use AI, which I’m okay with them using it, I want them to write a report first themselves and then use AI and then compare the two and engage in a little bit more of a of a
Dr. Jeremy Sharp (44:36)
Mm-hmm.
Dr. Adam Lockwood (44:45)
interactive or engaged instead of just offloading. And to me that’s a that’s a good way to to actually get more exposure to the data and more exposure to a client is to like, I’m gonna write this up. I’m gonna have AI and and when you’re comparing the two, you’re probably gonna get a better outcome, a better report because you’re gonna take the best of both. And you’re also gonna might get exposed to some things you hadn’t thought of, like, yeah, that’s that that is right. That’s that’s that’s a
Dr. Jeremy Sharp (44:54)
Hmm.
Dr. Adam Lockwood (45:12)
You know, that’s something I should have thought about.
Dr. Jeremy Sharp (45:15)
Mm-hmm. Mm-hmm. Yeah. I’m a big fan of I don’t know if you call it like the thinking partner approach, but you know, something where, yeah, you take a crack at it, maybe feed it to the AI and say, you know, I’m still wrestling with X. ask me questions to help me figure this out on my own without giving me the answer. You know, it’s like some version of that prompt. And then it, you know, of course, like engages with you and forces you hopefully to do the critical thinking and
You know, you it’s more of a process versus just a done for you solution. So no, no, go for it.
Dr. Adam Lockwood (45:48)
Yeah, AI is a sorry.
AI as a thought partner is one of the best things that you can do, I think. And especially if you say, I want you to be a critic. ⁓ there the there’s very much problems if you if you have nothing but yes people around you. So you want to be like, you know, I want you to be critical. And then also
Dr. Jeremy Sharp (46:00)
Yes.
Dr. Adam Lockwood (46:12)
Take the perspective of the person I’m going to be having this conversation with or whatever and think of what what they might think might actually can help with perspective taking, I think, and give you like, yeah, that might, you know. So I I really like that process.
Dr. Jeremy Sharp (46:23)
Mm-hmm.
Mm-hmm. Mm-hmm. Mm-hmm. Yeah,
I’m right with you. Yeah. It’s a little bit of an offshoot or detour, but you know, I’ll talk with parents sometimes, like if we are or even adults who’ve been diagnosed with autism, you know, or kids who’ve been diagnosed with autism and talk about that as a use case for AI, you know, if they’re like struggling with a social situation, you know, kind of.
put in the details and then help me see this other person’s perspective or help me see where I might have missed something here, you know, that kind of thing. Just to open it up a little bit for yeah, more to consider. Yeah. Yeah. So let’s see. What else we didn’t talk about bias at all yet. And I’m wondering, yeah, what you’re seeing as far as bias in AI at this point.
Dr. Adam Lockwood (46:58)
Yeah.
Dr. Jeremy Sharp (47:12)
I know is a big concern. Initially, is it still a big concern? Where are we at?
Dr. Adam Lockwood (47:18)
Yeah, so definitely a big concern of mine. there was a study in 2025 looking at bias in psychology, and what they did is they took vignettes and they fed them into CLAD, ChatGPT, and Gemini, which are the three biggest models, and they had 10 patient cases across five diagnoses.
Dr. Jeremy Sharp (47:29)
Mm-hmm.
Yeah.
Dr. Adam Lockwood (47:41)
Depression, anxiety, schizophrenia, eating disorders, and ADHD. They had three conditions: race neutral, so you could not tell the race of the person in the vignette, race implicit, where they they gave quasi-identifiers, essentially things that would let you the the model know the race of the person and race explicit, where they would say this patient is a black male or is a it’s a white female. And what they found is
Dr. Jeremy Sharp (47:52)
Mm-hmm.
Dr. Adam Lockwood (48:09)
That all of the large language models gave inferior treatment recommendations when race was implied or stated. And the interesting thing I forgot to add was they had one model that was a fine-tuned medical model. That was the most bias of any of them.
Dr. Jeremy Sharp (48:30)
It’s actually not super surprising just knowing a little bit about the medical literature and knowing there’s been like some pretty high profile cases of immunobias and
Dr. Adam Lockwood (48:40)
Discrimination? Yeah. Yeah. Or they’ve done they’ve done these similar studies with with humans for years where they’ll they’ll come up with a case vignette and the only thing that changes is is the the race and of the of the person in the vignette and
Dr. Jeremy Sharp (48:42)
Discrimination. Yeah, exactly. Exactly.
Dr. Adam Lockwood (48:56)
You know, a black male will tend to get a more harsh diagnosis like schizophrenia or bipolar. When if it’s changed to a white male, they might get you know, just depressions, you know, one of the depressive disorders or something. So it doesn’t surprise me because this is what we see in our society and in our field, if we’re honest with ourselves already.
Dr. Jeremy Sharp (49:01)
Mm-hmm.
Yeah.
Right, right. I mean, is there anything what can we do about this? Is there any way to proactively mitigate this? is it just careful review of the output? How do we tackle this?
Dr. Adam Lockwood (49:32)
Well, I mean, one of the things is probably having more diverse people be part of the conversation and part of the building of these systems. That’s that’s really important. But from a practitioner standpoint, we need to be, I think, super self-critical because essentially what we’re what you’re gonna find is is confirmation bias. People tend to think of these tools, these models or systems as
Dr. Jeremy Sharp (49:46)
Yes.
Mm-hmm.
Dr. Adam Lockwood (50:00)
subjec are objective and they’re not. And if it says something that you might already consciously or unconsciously believe or certain stereotypes that you already subscribe to, whether you know it or not, if it if it generates something that that reinforces that, you’re likely to go along with it. So I think being really
Dr. Jeremy Sharp (50:25)
Mm-hmm.
Dr. Adam Lockwood (50:26)
skeptical, really thinking about any output if it’s someone from a from a group that’s different from you, especially if they’re from a historically minoritized group, is is a place to start. I don’t know that it’s sufficient, but it’s it it’s something we should be doing. It’s like, yeah, am I being racist here? Am I discriminating? Am I being biased?
Dr. Jeremy Sharp (50:41)
Mm-hmm.
Mm-hmm.
Yeah, I think that’s a great point. It’s a great point.
So this might be getting in the weeds just a little bit, but I wonder when we’re thinking about bias, I mean, is there anything that we can do on the prompting side or instructional side for the AI model that we’re using to help almost like help it monitor itself for bias and at least alert us when something you know, the the content might be biased? Or is that just too iffy?
Dr. Adam Lockwood (51:14)
I mean I I’m not
aware of of any guidance in in regarding that. you know, it’s like don’t be bad. Don’t be biased. I mean it’s like, okay,
Dr. Jeremy Sharp (51:19)
Yeah.
Sure.
Yeah. Yeah. It’s a great point. It’s a great point. Yeah, this is one of the worst parts. Worst parts about it, I think, is like, yes, it’s trained these models are trained on tons of information and data, but you know, it that data was generated by humans and humans are flawed in this way.
Dr. Adam Lockwood (51:43)
Yeah, it just
perpetuates the the inequity that we see in our society already.
Dr. Jeremy Sharp (51:50)
Right, right, right. Have you found yourself like gravitating toward one model over another lately? Like do you have a favorite, so to speak?
Dr. Adam Lockwood (52:00)
like f from like a frontier model, like a commercial model?
Dr. Jeremy Sharp (52:05)
Yeah, yeah, yeah. Yeah, that’s what I was thinking. Most listeners I think are only gonna be messing with frontier models.
Dr. Adam Lockwood (52:10)
Yeah, so it depends on the use case. which which sorry, I don’t mean to use too much jargon. Depends on what I’m using it for. But ⁓ I personally think OpenAI is the best model currently. depends on which benchmarks you you you look at. But so I use it as my most go to for writing. I use claud code.
Dr. Jeremy Sharp (52:14)
Yeah. Yeah.
Mm-hmm.
Mm-hmm.
Mm-hmm.
Dr. Adam Lockwood (52:35)
I think I think Codex, which is OpenAI’s coding app, is better than than Anthropics. ⁓ and part of that comes down to tokens. If you look at Cloud sideways, it it runs out of tokens, you know. And and even with a $200 plan, you you’ll you’ll go through your tokens really quickly. ⁓ so
Dr. Jeremy Sharp (52:44)
Mm-hmm.
sure. Yeah. Yeah.
Yeah.
Dr. Adam Lockwood (53:00)
ChatGPT is just way, way more generous with their with their tokens. I’ve never maxed out ChatGPT even on a twenty dollar plan. Now I have with Codex. but Claude, yeah, you look at it wrong and it’ll be like, and I’ll come back in three hours.
Dr. Jeremy Sharp (53:17)
Sure, sure. Yeah. I’ve heard stories of that, for sure and run into it myself a couple of times. And just for anybody, you know, any audience members who might not know what we’re talking about here, it’s like basically, you know, you get a certain allotment, like a usage allotment with these models and depending on what you’re doing, you know, with the model it’ll use these tokens. You get a allotment of tokens and Claude uses them.
Excessively, it seems like. It runs out quick. Yeah. Great. Great. What let’s see, what are we looking ahead to? Do you have anything on your mind that’s coming up, either research-wise or practice-wise, that is actually exciting use cases that have a lot of potential that we haven’t touched on at this point.
Dr. Adam Lockwood (54:07)
I I have no idea where this is headed, but I believe that that this technology is going to change things radically. I think it’s probably going to be I don’t know what is my my area, like I said at the beginning, is assessment. I don’t think assessment’s gonna look the same way ten years from now than it does today. Although I’ve been saying that for quite a few years and it hasn’t changed. So so I I whenever I make predictions, it’s based on what the technology can do and not
Dr. Jeremy Sharp (54:16)
Yes.
Mm-hmm.
Mm-hmm.
Mm.
Dr. Adam Lockwood (54:34)
what adoption will be like from people because what we’re finding is people are much slower to adopt this technology and and to realize this capabilities but you know i i i think many many data points are going to be used across different modalities or whatever you want to call it so it’s you know assessment we’re still using these really subjective paper and pencil self-report
Dr. Jeremy Sharp (54:38)
Mm.
Yes.
Dr. Adam Lockwood (54:58)
sort of approaches. I see this as being wearables, you know, all sorts of
Different approaches that will all go through AI because AI can analyze millions and millions of data points and and very quickly and efficiently. And so the approach, I don’t know what it’s gonna look like, but I don’t think it’s gonna be you’ve got a piece of paper and a pencil or two iPads and your your your
Dr. Jeremy Sharp (55:16)
Mm-hmm.
Dr. Adam Lockwood (55:27)
Or you’re asking questions. I that that’ll definitely be part of it. I think someone’s lived experience will always be important, but I think it’ll be more complex than that, more more detailed. I don’t know.
Dr. Jeremy Sharp (55:42)
Yeah, yeah. It’s funny. I heard somebody say, I don’t know, six months ago or something, I think related to AI and technologies like there’s always way more progress in a year than you think there’s gonna be, but way less progress over ten years than you think there’s gonna be. And I don’t know, that seems to to fit here. It was there was a huge spike, initially in the first couple of years. And I don’t know, it feels like we’ve settled a little bit, and it’s not moving forward.
Quite as fast. But I did want to ask you about the adoption. This is interesting. I am still kind of shocked, I suppose, at how many people are not utilizing AI at all or are concerned about it or whatever it may be. So do you have any insight into, especially among psychologists, what what the hurdles are to adopting AI, like reasons people choose not to adopt?
Dr. Adam Lockwood (56:30)
Yeah, sure. We
Pull that up.
Sorry, I have to pull up this preprint here.
Dr. Jeremy Sharp (56:38)
I love that you just have it. The data’s right there. That’s great.
Dr. Adam Lockwood (56:43)
And I also blog about this. I have a blog I write about all this stuff. Anytime I come up with a paper or or I read a paper that’s super interesting to me, I try to blog about it. Let me just see.
Dr. Jeremy Sharp (56:52)
Nice.
Is that a in case the readers or listeners want to check it out, is that a substack thing or is this just a personal blog? Like how do people find that?
Dr. Adam Lockwood (57:01)
This on my website,
I don’t have a sub stack, but I should just create a agent to do that for me.
Dr. Jeremy Sharp (57:08)
Hey, there you go.
Dr. Adam Lockwood (57:10)
So let’s see.
The main barriers to AI adoption among health service psychologists, so PhD level psychologists, 71.3% ethical concerns, 70% privacy, ⁓ 69% accuracy, 53% legal, 50% AI replacing human judgment, 47% bias. And then we start getting into things like
Dr. Jeremy Sharp (57:24)
Mm-hmm.
Dr. Adam Lockwood (57:37)
Not being familiar with the tools, lack of ev evidence-based research, lack of training. But yeah. Good I mean good reasons to be concerned. And and it there I I I don’t know, I don’t mean to minimize these at all.
Dr. Jeremy Sharp (57:43)
Mm-hmm. Yeah.
Right, right. Do you I mean, do you have any deeper like when people say ethical concerns, what that means exactly?
Dr. Adam Lockwood (58:02)
you know that we should dig more into that. because that could mean I to some people that’s more like moral issues. I’m I’m concerned about wealth inequality, or I’m concerned about you know electron elect electric electricity use, ⁓ which is on the rise, you know, quite a bit of the more tokens you use, the more power, the more compute you use, the more power you’re using. but yeah
Dr. Jeremy Sharp (58:10)
Mm-hmm.
Mm-hmm.
Right.
Yes.
Dr. Adam Lockwood (58:28)
If it’s broken down to privacy confidentiality, that’s that’s a little bit more that’s a little bit more clear bias. But yeah, that’s just more general ethical concerns. And sometimes when I talk to people, they say they have ethical concerns and it and you ask what are your concerns and it’s more vague. ⁓ they’re not sure. And that’s okay too. Like I I’m not, I don’t know what I don’t know is is legitimate. Like I have ethical concerns and I I’m not
Dr. Jeremy Sharp (58:46)
Mm-hmm. Mm-hmm. That’s
Dr. Adam Lockwood (58:56)
I don’t know enough about the technology to be fully to have cogent, you know, reasons, but I just I just worry, what is this thing?
Dr. Jeremy Sharp (59:07)
Right, right. Yeah, that’s fair. That’s fair. I feel like I talk with a lot of folks who have quote unquote ethical concerns, but they they feel more vague, right? And I will say this though, in private practice, ’cause I I’m primarily, you know, consulting with folks in private practice, it’s more around I don’t know that this is ethical necessarily, but something around like, well then what is the client paying for? Like how can I charge the client if I’m utilizing these tools?
And I’m not doing the work anymore. You know, like it’s like my value is decreased, you know, because because of these tools. I’m I don’t know what to do with that. I d I don’t necessarily agree, but I know that comes up quite a bit, at least in the private practice realm.
Dr. Adam Lockwood (59:49)
Well, and you know that that was the a pretty big concern. And it was a big concern of mine at the beginning too. And you have people like basically the CEOs of of the some of the main AI companies saying, Yes, is gonna put out, you know, fifty percent of knowledge workers. And we haven’t seen that happen. ⁓ and it’s because you need experts overseeing the output. And as we
Dr. Jeremy Sharp (1:00:08)
Mm-hmm.
Mm-hmm.
Dr. Adam Lockwood (1:00:16)
Get these tools, it’s not just you’re saving time and therefore you’re laying people off. It’s like they’re like, I save time and here’s all the other things I can do. I can provide better services. So what they’re paying you for is the knowledge and expertise you have, the human qualities you have, empathy. I mean, like, you know, I don’t know how much of a humanist you are, but you know, if you’re if you’re a humanist psychology, there’s psychologists, there’s no way that an AI system can ever do that because it
It can’t have a relationship. so to some extent, they are and have always been paying for a relationship, in my opinion. Someone who’s going to provide them with you know unconditional positive regard and and there’s a lot of value in that.
Dr. Jeremy Sharp (1:00:45)
Exactly.
Yeah. I agree.
Yeah, yeah, I’m with you. That may be a nice note to end on, a hopeful note that AI is not gonna replace us in relationships right.
Dr. Adam Lockwood (1:01:12)
I don’t think if you would have asked me two years ago it would have been a lot more
I would have thought that AI was going to replace more jobs than it did and and now I’m I’m not I’m not as sure of that. I think we’re we’re going to just use that time to do more with these tools that are allowing us to.
Dr. Jeremy Sharp (1:01:34)
think that’s true. I think that’s historically consistent too, right? I mean, at least from different things that I’ve looked at. When there’s an advancement in technology, we tend to fill the space. Like those original jobs maybe shift or maybe go away, but then new things come up and new tasks. And there’s always demand for these advanced skills that we have. That’s the
I really appreciate your time. we could talk about many things in this realm and and hopefully might have another conversation down the road as things evolve. But this has been a lot of fun. Thanks for going back and forth and, you know, batting around some ideas around AI and just sharing your expertise.
Dr. Adam Lockwood (1:02:13)
Well, thanks for having me.
Click here to listen to the podcast instead.