OTS 196: AI and Clinical Reasoning in School-Based Occupational Therapy
- 9 hours ago
- 37 min read
Click on your preferred podcast player link to listen wherever you enjoy podcasts.
Welcome to the show notes for Episode 196 of the OT Schoolhouse Podcast.
Artificial intelligence is quickly entering healthcare and education, and occupational therapists are asking an important question: How can we use AI responsibly without losing the clinical reasoning that defines our profession? In this episode, Jayson Davies sits down with OT educator and researcher Tara Mansour to explore practical, ethical ways school-based OT practitioners can begin using AI tools.
Tara shares how she teaches future occupational therapists to use AI as a “first draft partner” while still prioritizing evidence-based practice, professional judgment, and student-centered care. They also discuss privacy considerations, prompt strategies, treatment planning ideas, and how AI can support data collection, documentation, and intervention development.
If you're curious about how AI might fit into your school-based OT workflow—or concerned about how it could impact clinical reasoning—this episode provides a thoughtful and practical perspective. Tune in to learn how AI can support occupational therapy while keeping the human clinician firmly in the loop.
Learning Objectives
— Learners will describe appropriate ways AI tools can support school-based OT practice, including treatment planning, documentation drafting, and intervention idea generation.
— Learners will explain ethical and privacy considerations when using AI, including FERPA and HIPAA concerns and strategies for de-identifying student information.
— Learners will identify the "human-in-the-loop" approach to AI-assisted practice, distinguishing between AI-generated versus AI-assisted work and the role of critical clinical reasoning in evaluating AI outputs.
Guest Bio
Dr. Tara Mansour, OTD, MS-HPEd, OTR/L, is an Assistant Professor and Academic Fieldwork Coordinator at the MGH Institute of Health Professions with over 23 years of clinical experience. As a PhD candidate in Health Professions Education, she investigates innovative approaches to prepare students for professional practice, with a focus on integrating artificial intelligence into fieldwork education to strengthen learning and critical thinking. Dr. Mansour co-chairs the Generative AI Task Force at MGH IHP, leading initiatives to explore AI-driven innovations in health professions education.
She has authored scholarly articles on AI integration in clinical education, including publications in Frontiers of Medicine, Medical Education, and the Journal of Allied Health. Nationally recognized for her contributions, she has presented at AOTA's Education Summit and Academic Leadership Conference and serves on ASAHP's Clinical Education Committee, shaping best practices in clinical training across disciplines.
Quotes
"AI is really… a first draft partner or an option generator. It is not a decision maker.”
— Dr. Tara Mansour
"The risk isn't the tool—it's turning off your critical thinking. When used well, AI can actually strengthen reasoning."
— Dr. Tara Mansour
“Unreflective use weakens reasoning. Over-reliance without thinking weakens reasoning.”
— Dr. Tara Mansour
“AI probably won't replace us soon, but we need to use AI. to basically enhance OT and to keep it at the forefront of health sciences.”
—Jayson Davies
Resources
Episode Transcript
Expand to view episode transcript
Jayson Davies
Welcome back to the OTs schoolhouse podcast, your source for all things school based occupational therapy. I'm Jayson Davies, and today we're diving into a topic that is reshaping our profession, artificial intelligence and occupational therapy. Joining me today is Dr Tara Mansour, assistant professor and academic fieldwork coordinator at the MGH Institute of Health Professions with over 23 years of clinical experience, including extensive work in school based occupational therapy, Tara has become a national leader in integrating AI into health professions education. In this episode, we are tackling some of the questions that you have been asking me, like, Can AI help with IEP goals. How do we ethically use AI in our practice? Will it weaken clinical reasoning for current and or maybe even future occupational therapy practitioners? And most importantly, how can AI give us back the time that matters most so we can work directly with our students, whether you're AI curious or AI cautious, this conversation will give you practical, ethical strategies to make AI work for you, not replace you. Let's get started.
Amazing Narrator
Hello and welcome to the OT school house podcast, your source for school based occupational therapy, tips, interviews and professional development. Now to get the conversation started, here is your host, Jayson Davies, CLASI is officially in session.
Jayson Davies
Tara, welcome to the OT school house podcast. It is a pleasure having you on as you know, I love AI, and it sounds weird to say I love AI, but it is true, and I love it, how it can support us. And I know you have a lot of experience with AI, so I'm excited to dive into this topic with you. How are you doing today?
Tara Mansour
I'm well, thank you for having me here. I'm excited Absolutely.
Jayson Davies
And, you know, as we get started, I just kind of want to give you an opportunity just to share a little bit about where you are in the world of occupational therapy, and you can even start to dive into, kind of how you came around into the world of AI,
Tara Mansour
sure I'd love to. So I've been an occupational therapist for almost 25 years, and I spent a large portion of my clinical career in school based practice. I started my career in inpatient rehab with a specialty in neuro rehabilitation, and then transitioned to school based practice working in an elementary school that housed a medically complex sub separate classroom. Over time in that same public school district, I worked across different elementary schools, the middle school and the high school, really working with a variety of students with different needs. In that setting. Also, I served as a level two fieldwork educator for many students, which led to my transition to academia, where I now serve as the academic fieldwork coordinator and assistant professor at the MGH or Massachusetts General Hospital's Institute of Health Professions. There, my role focuses on preparing future occupational therapists for clinical practice, especially through field work education.
Jayson Davies
And that is not an easy job. There's a lot of people that you have to coordinate with, the students, the field work educators, the people in your program
Tara Mansour
there sure are. It is satisfying. I love when our students come back from their level two field work placements and are ready to jump in and ready to make the transition for my students to my future colleagues awesome.
Jayson Davies
And before we really dive into everything, I kind of want to ask a fun question here, and that is, how is AI been the most helpful for you in this role as a field work coordinator,
Tara Mansour
from an education perspective, it helps me in developing resources for our students and our field work educators, from learning activities to evaluation tools. It is a great draft partner. You know, we'll talk a little bit more as we move forward about a human in the loop approach, but from composing emails to creating learning activities, I find many, many ways to utilize it within my current practice in education, perfect.
Jayson Davies
All right, so yeah, now let's kind of get to that point where you started to maybe the first, the first points at which you started to realize that AI could be helpful, and what brought AI onto the radar for you?
Tara Mansour
So honestly, AI first got on my radar because I have teenagers. They were the ones who introduced me to chat, the chat GPT app on their phones, and at first it was more of a curiosity thing. Coincidentally, around the same time, I was also hearing from field work educators that they wanted to see our students coming up with more creative intervention ideas. Well, I absolutely can appreciate and understand that I was a level two fieldwork educator myself, and very much appreciated a fresh set of eyes and creativity from my perspective now as an academic field work course. Of neater. My concern was different. I was less worried about whether students had creative activity ideas, and more concerned about their clinical reasoning. Were they grading activities appropriately? Could they create a just right challenge? Were their intervention choices evidence based? Could they successfully document the treatment session? I didn't want our students spending an hour and a half scrolling Pinterest or maybe these days Tiktok for treatment ideas. I'd rather they spend that time thinking through the clinical decision making behind the plan, especially because they're new and they don't yet have a deep toolbox to draw from. So that's when the question shifted for me, how can we leverage AI to generate idea options quickly while still teaching clinical judgment, ethics, safety and responsibility? And for me, that really became the entry point for my work with AI and OT education.
Jayson Davies
All right, perfect. I like that because I think there is a fear. I've done some work, some research, in AI, and some of the things that we hear is that students aren't going to have that clinical ability, that ability to process clinically because of AI. And I'm sure you probably have had this idea and wrestled with this idea as well. So I really want to kind of dive into this. It's not a resource, this program that you've started to put together a little bit and how you are trying to help the students without, you know, giving them this tool that makes it too simple for them, absolutely.
Tara Mansour
So, you know, AI is really, I frame it as a first draft partner or an option generator. It is not a decision maker. So while AI can indeed suggest things like goal language, intervention ideas, ultimately, as the OT practitioner or student occupational therapist, you are the one evaluating and editing everything clinical reasoning you know, lives in the filtering and the adapting. You know you're asking, does this actually match my client? Is it the right challenge level? Is it appropriate in this environment? Is it evidence based? So while AI might be able to handle some of the repetitive scaffolding pieces, the structure, the wording. Some idea lists the therapist always have has to apply professional judgment. It widens the option set, but the OT still chooses the path.
Jayson Davies
So I will mostly, you know, frame this all around school based OT, because that's exactly what 99% of the people that are listening to this are so I guess my question is, then, when you're supporting your students or either even other clinicians, how are you suggesting that they prompt? Is it more of because you talked about right finding ideas versus the actual clinical reasoning? Are you suggesting that they prompt in a way that's like, I have this student experiencing this strengths? Are this? What are some ideas? Or do you kind of flip that approach and say, Hey, I have this idea for painting something. How would this apply to my student? Or is it a mix of both, or more than those two?
Tara Mansour
Yeah, I would say it's really the first but that's just the first part. So I teach them how to enter in case study information while not providing any client identifying data. That's the first part. We do talk about prompt engineering, what makes a good part a prompt and what doesn't make a good prompt. And they get the opportunity to kind of play in that playground a little bit. Ultimately, though, what is generated by AI that is just really the first part of the assessment, because then they need to take what they have found, those intervention ideas. And maybe they scaffold that, or they they reprompt Based on what the output is right to narrow things down. But once they choose three intervention ideas, which is what we do as part of the assignment, then they have to go to the literature, they have to go to their textbooks, they have to go to peer reviewed journal articles, and they need to justify, for me why that intervention is indeed an evidence based choice. They have to go back to what I provide them as the patient's chart, looking at their precautions, and they have to determine why, ot what their final choice of interventions is based on that review. So we're not just teaching them to utilize generative AI for intervention planning ideas, but we're incorporating evidence based practice so they can effectively evaluate what AI is putting out there.
Jayson Davies
Yeah, that's smart. I think we all. So the term hallucination should have been like the word of the year a year and a half ago or so. It's still it's still happening. I like to think that it's happening less when I'm using it, I'm seeing less completely made up research articles being linked, but it still happens, and obviously the output definitely have to judge it and find out if it's right, just like you have to do honestly with any article, though, like any article we read, we have to go through that process to make sure it's worth reading and worth really taking information from So, yeah, absolutely okay. One thing that you brought up was making sure that your students aren't using identifiable language, right? We have to be HIPAA, compliant, fair, but compliant within the school based realm. What is important to not put into something like chat, GPT? Can we use a first name? Or can we not use a first name? Do we need to use pseudonyms? Should we just erase all information period? What's safe, what isn't?
Tara Mansour
So I think, you know, some non negotiables, no matter what setting you're in is no identifiable student information ever goes into a public AI tool. So I would not even consider using a first name and not a last name. I would say, you know, make up a name, no dates of birth, no school names, no facility names. You know, I think sometimes even within our communication as practitioners, we think that using initials is non identifying. When, in reality, if somebody was, you know, being, you know, if public records were being utilized, they're going to search for those initials as well. So no initials, no unique facts, and no uploaded reports or IEP documents.
Jayson Davies
Yeah, yeah. It's funny when I was doing sorry really quickly, when I was doing my initial, like HIPAA review, you know, for myself again, just to kind of go back and look, I forgot that you know if that first name can be attributed to you and where you work, and you only have you know one or two of people with that name on it, well, that is officially identifiable information, and anyone can go on to your LinkedIn page and figure out where you work. And so that's pretty easy.
Tara Mansour
Yeah, absolutely. And I think that you know, what I teach students and clinicians is to really make sure they're using generalized de identified descriptions. They can they can mention functional skill areas, maybe interest themes, goal categories that should be really more than enough to generate useful ideas without risking privacy. The other thing from my experience is that in the school system, they often have contracts with sites that do indeed allow you to provide identifiable client information. And I've heard some things around Google and public school access that has said it was FERPA compliant. But unless you have that from your district's IT department, I would always default to the side of caution and depersonalizing any added information.
Jayson Davies
Yeah, absolutely. While you're speaking, I was trying to look up the chat GPT free educator version, because I know they provide chat GPT to free for educators and students, I believe, but I don't know if that free version is HIPAA compliant. I want to say maybe, but definitely do your own research. And it could change, right? They can change anytime.
Tara Mansour
And I wouldn't go by the the external program telling you that super compliant. I would go to your IT department and make sure that you're following your own institution roles.
Jayson Davies
Great point, yeah, third party, or even the people that on your side, make sure you're covered from that and not don't always. I mean, we don't trust the AI for the content that they generate all the time, so we should definitely be mindful of what they are saying as well. Good point. Okay, we're gonna go ahead and take our first break, but when I come back, I want to talk to Tara about where she feels that AI can best support school based ot practitioners like us. So stay tuned. Okay, so, so there's a lot of OT practitioners in the research that we've done, they were worried about new students almost being having an over reliance on AI and not having the clinical reasoning skills because they've had chat GPT since, like your teenagers have. So the question is, is, how do we prevent that? How do we ensure that these students that are coming out of OT school aren't just relying on chat GPT, and that if chat GPT were to, you know, die overnight, they would still have that clinical reasoning.
Tara Mansour
So I would say this is a large concern, and I take that concern very seriously, especially as, you know, faculty in an OT education program, I actually don't think that the tool, it's. Self weakens reasoning. Unreflective use weakens reasoning. Over reliance without thinking weakens reasoning, and that risk exists with many tools that we already use. True from a school based perspective, let's think about the example of weighted vests. Okay, if a clinician reads an article in a peer reviewed journal about weighted vests, and then starts using weighted vests with their students, without collecting data, because the article deemed it effective, yet they proceed without any monitoring, without monitoring any outcomes. The issue is not the vest. The issue is the reasoning process around its use. So did they look at the evidence base broadly? Did they individualize? Are they measuring the students responses? I really think that AI is similar. The risk isn't the tool. It is turning off your critical thinking. I do believe that when used well, AI can actually strengthen reasoning, because it gives you multiple options to compare and critique, and that comparison process in itself is a clinical reasoning exercise. The key is how we teach and we model use AI should really expand their option set, not make their decisions. And clinical reasoning lives in the selection, the adaption, the justification and the outcome tracking, and that part remains fully human.
Jayson Davies
I want to, I want to press on that. How does, how does Chachi BT expand options? I mean, you mentioned the vestibular or the weighted vest or pressure vest, and I totally understand that, right? We need to make sure that on the back end, we are tracking data and whatnot. But you mentioned that idea of chat GPT opening up the options. What do you mean by that?
Tara Mansour
So it may give you a variety of interventions, ideas that you may or may not have considered, but again, that human in the loop approach. We're not using chat, GBT as our Google and then just taking what it says. We are saying, okay, these are some good options. Now I need to critically reflect on what it is suggesting and think about is it evidence based? Is this beneficial for my client, is it going to be of interest to my client? So so it expands that as long as you are not you know, what I say to my students is, if you're running late for a field work and you haven't done your treatment planning for your 8am session, you cannot use chat GBT in the five minutes that you're getting off the bus to say, I am working with a patient who had a stroke. What should I do? And then jump in there five minutes later and just translate the output of chat GBT into your treatment session? There is no reflective use in that
Jayson Davies
good point. Yeah, it's true, and it's interesting, because I think there is kind of this, these tools, I don't think are readily available yet, but eventually, I think that, good or bad, you will be able to do something to that degree. I think because eventually, and it's already happening, AI is being built into our EMRs, and therefore it knows what I did with Johnny yesterday. Knows what I did with Johnny last week, the week before, the week before that. It it has the evaluation in a form that is secure and HIPAA compliant, and it does know a lot about Johnny. And so to what degree should I still or do I still need to reflect upon something if it has all that background data on Johnny tough
Tara Mansour
question, well, we're not there. We're not there yet. I think you're right. I think that what initially was spoken about when we thought about generative AI were these standalone, large language models, chat, GBT, copilot, Gemini, you name it, but what we're really seeing now in clinical practice, just like you said, is that there are systems, there are AI systems being built in to the EMR. My concern is, less How much does it know about Johnny, and more about how much does it know about Johnny's diagnosis or appropriate interventions? What it's not going to know is Johnny's interests, that Johnny likes cars, but he hates baseball, right? And so you're not always going to be to be able to rely on the suggestions outputted, right? It also depends on what is your system. Where is your system pulling the data from? Is it pulling I would always continue to be extra cautious if data is being pulled from the web, or are you behind a firewall and it's only using the information you've provided it with? So I do think we are a long way away from simply throwing a prompt out there and being able to act on it again. From a school based perspective, it might have wonderful ideas. When students are presenting with certain disabilities or diagnoses, but it still doesn't know their interests and that personalization piece that makes us OTs, yeah. I mean, never say never. But I that wouldn't be a way in the near future I would envision teaching anybody to use the tool,
Jayson Davies
yeah, yeah. I also know, I mean, how many times have all of us gone into a session with an idea that we thought was great, whether it came from Instagram, Tiktok, our own brain, or chat, GPT, and then we had to go through three different interventions during that session, because students, you know, have days. We have days and and sometimes plan A and B don't work, and so you got to move to Plan C and D. So you've got to have that clinical reasoning skills to do it in the moment, not five minutes or five days before the session actually starts,
Tara Mansour
absolutely and I think that goes into just old fashioned treatment planning, right? Like, you know, even with with generative AI aside, students have to be able to, you know, be iterative or change their treatment plans. And to be fair, that is a more challenging experience for them, because they don't have the toolbox. They don't have the automaticity yet that those of us with more years of experience have. And so you know that goes back to a whole nother activity that has nothing to do with chat GBT per se, which is making sure our students are completing treatment plans when possible that has a variety of interventions so they're not panicking and freezing in that moment when things ultimately have to Change. Yeah, that's a best practice in field work, education, I would say, versus, you know, generative AI,
Jayson Davies
yeah, yeah. And that's fair, kind of touching, you know, going specific into school based OT. I often like to break school based ot into a few categories of things that we do. One is that evaluation piece. Then we kind of have the after the advice, after the evaluation. We have to do treatment planning. We actually have to do the treatment itself. And then there's some, you know, administrative side to it. There's some MTSS, do you think that AI can really support all of those areas equally? Do you find it more beneficial for particular areas? Obviously, we've talked about treatment planning a little bit, so little bit so far today, but what areas encompassing school based OT, do you feel like AI can really be supportive?
Tara Mansour
So, you know, I think, I don't think it's equally across all of the things that you just mentioned. I think that first draft note, writing, goal wording, again, with full human review, can be accelerated with the use of generative AI. I think that accelerated time gives students and clinicians, you know, more time to think deeply about what actually happened in the session or what to do next. I think it's important to realize, especially within the context of school based practice, this kind of technology isn't completely new. In ot we've seen earlier ways of digital assessment technology, right? Many of our evaluations moved from manual scoring to digital platforms years ago. You know, tools like Pearson's digital scoring lets you enter subset scores and generates for you structured summaries and interpretive language. If your district subscribes to that platform, the sensory profile went digital. No more hand scoring grids, right? That was an early automation supporting clinician efficiency, and I think I see AI, you know, as a part of that same evolution with a little bit more flexibility and more interactive. I think it can help with things like drafts of reports, you know, drafts of progress notes, drafts of IEP goals, drafts of present levels of performance. But ultimately, the interpretation and clinical judgment still belongs to the OT. I think the goal is, you know, time saved on mechanics, and it's kind of reinvested, reinvested in, in reasoning, if you will, gotcha.
Jayson Davies
And, yeah, I completely agree with the time saving thing. I've experienced this before, and I want to hear if you've experienced this, or, or, if you've, you know, worked with some of your students or other clinicians that have experienced this, where you think that AI is actually going to save you time, but then you get playing with it, and you go down a rabbit hole. I mean, we always talk about how many tabs we have open in our, you know, our Google Chrome, but oh my goodness, the chats that I have on the left hand side of my chat, GPT, like it's infinite. I don't know how many are there. So have you experienced that yourself and or from others,
Tara Mansour
I think, as a new clinician, or being exposed to a new tool that you are not that you are not used to using. So it's always going to be time heavy. In the beginning, the your clinical reasoning is going to be more time heavy. Learning best prompts is going to be time heavy. I don't think that it's one of these things that you introduce and all of a sudden it is, you know, a magic fix. When I was working in school based practice, I did have a student who was struggling with report writing, and a friend of mine, a peer of hers, had this company where you just entered in some of the student information from the evaluation, and it actually generated a whole report for you. Now I didn't love the format of that report, and initially, when practicing to utilize that system myself. It took me longer to write a report than if I just sat there to write the report myself, but over time, it did decrease the amount of time I needed to write the report, and I came to really appreciate that as a resource.
Jayson Davies
Yeah, yeah, and I agree. I mean, it takes a it takes some time getting used to chat GPT and or if you're going to use Claude, or whichever going to use, it takes some time to get used to it, and again, some time to understand prompting. You talked earlier about how you help with prompt generation for the students that you're working with and helping them understand AI, do you find that there are specific ways, or specific, I guess, templates for prompts that work out well for school based occupational therapy, or just ot in general? Or can you just write in anything?
Tara Mansour
So I don't, I very intentionally don't create templates as it's different for every student, every client. What you're asking, you know, generative AI to do. What I will say is, I do educate them on a process in generating good, you know, strong prompts. And I I just pulled up my slideshow that I use with them. And what I'm looking at is in generating an effective prompt. This is the outline that I suggest they go through, which is, step one, identify the client population and condition. Step two, specify the occupational performance challenge. Step three, define the intervention focus. So for example, as I bring them through this learning activity, as we are reviewing the case study, I am asking them to determine the frame of reference to their approach. Am I looking for a compensatory, rehabilitative frame approach? Am I looking for a more remediation or biomechanical frame of approach? Then I'm asking them to consider the treatment setting and available resources. Is this inpatient outpatient school based? Where is this session happening? And then they can request a specific response format, give me some intervention ideas. Or they might say, give me some ideas for a 30 minute session. Or they might say, just give me ideas that fit within the biomechanical, you know, frame of reference. And then step six, which is most important, is refine the prompt. So when you get that first round of generated text, what do you like, what don't you like? Do you need to refine your initial prompt? Or can you just dive in deeper with some additional clarifying questions? I like to think of it. This whole approach was built for those people that have experience in doing literature reviews or searching databases for literature, you might put in a few words and get 12,000 responses, and then you're going back and you're saying and this, and that narrows it down, and this, and that narrows it down. So my whole instruction of utilizing generative AI in prompting and follow up prompts really is based on that whole idea of a literature review.
Jayson Davies
And you bring up that idea of a literature review, and I love that, because one of the biggest problems that I've run into with AI, and we talked a little bit about hallucination making up links and whatnot, is that I don't know what Article, database, chat, GPT or cloud have access to, and what they don't have access to. And so if I'm asking it for evidence or best practices, I am sometimes shocked, because it sometimes has access, complete access to an article, and other times it's only using just the abstract, and I never know what it's actually using, unless I kind of go into or click on the link and try and figure that out. So I guess my question is is, how do we and maybe the answer is, we just have to do it, you know, on our own. How do we know if something is actually evidence based? Raised if it's citing these articles, but we don't know if it has access to it. Does that make sense?
Tara Mansour
Yeah, absolutely. So I think that you know, remember, some journals are open access and some journals are not open access, right? So, so it can indeed only have access to an abstract, which sometimes will give you enough information that you need, and then other times it doesn't have access to full journal articles. Regardless, you should never trust gender, you know, generative AI's output into quoting the evidence, even if you put into your prompt, find only for me. You know, peer reviewed published articles. I still would not trust the system. I think a great first step is to say, give me the DOI number and follow the link and see if it's actually a DOI, right. And again, you're still never just trusting the information that it told you about the article. But now you have the DOI. Now you have access to the article. Read the article right, or download the article, put it back into the generative AI system and ask for a summary. Even that isn't always right, but now at least you're looking at very specific information that you fit it.
Jayson Davies
Yeah, interesting that you say, say that I have definitely gotten a quote from chat GPT, clicked on the DOI command, F or whatever search feature, type that quote in, and then you're always like, wait, you told me this was a quote. Why can I not find this three set of words in a row in this article that you said was directly from this article? So yeah, you got to be careful. You mentioned pulling the article and putting it actually into the source or the AI, I am, I have gone back and forth on this all the time, and I don't know where we fall as a profession on this one, and where is the line drawn, because i don't know i i Just troubled with it a little bit. I have done it. I definitely have downloaded an article, put it into chat. GPT, hey, help me summarize this, or what are the key points, or whatnot. But I'm torn on this, and I don't know where the profession stands on this. Have you heard anything?
Tara Mansour
I haven't. I feel like I have come to my own personal judgments for how I am choosing to utilize it. What I'll say is, if an article is open access, anybody has, yeah, access to it, right? So it's not that you are uploading copyrighted protected information. I will tell you that I have, I have been able to access an article. When I say access, I mean, understand an article more that if I was just reading maybe some of the statistics, language was going over my head, and I was having trouble really engaging in that article. But rather than just pushing it aside, being able to, then again, interact with generative AI to ask it specific questions, right? Not just give me a summary, but the article talks about Cohen box, you know, statistic. What does that mean? What is that measuring that, in my opinion, is a best practice of how we're utilizing it's almost your own personal tutor. You know to go back and just say you're going to rely on something for a generated summary. I agree with you. I have fed it content different across different platforms, actually, and and the different platforms vary about how accurate the quotes are that it gives it back. But even at that, find me a quote. Find the quote right? Like you should never be in a position to just use ever what generative AI outputs and then assuming it, you know, is correct. That goes back to what I said in the beginning about that reflective reasoning. There is always a human in the loop approach. It is never copy and paste.
Jayson Davies
Yeah, absolutely. All right, we're going to take a quick break, but when we come back, I'm going to ask Tara how she thinks that ot practitioners like us can use AI in our evaluation process. All right, I want to go through a few common tasks that school based ot practitioners would use or would complete throughout the day, and kind of just get some ideas from you about how AI might be used, and you've already mentioned a few, but kind of just task specific here. So for any evaluation, you know, most people already have their kind of set evaluation. They do certain things in order, but how might you begin to incorporate AI into the evaluation process?
Tara Mansour
So, you know, I'm not sure that I would, from the get go, be incorporating AI into the evaluation process. If I was given a referral and the students challenges, I would still initially, independently identify. By which tests I was going to complete, right? I would not utilize AI per like within the testing, because it is so much more. Testing is so much more than just following the protocol. It is observing their posture. It is absorbing it is observing their pencil grasp. It's observing their behaviors in the environment. I see the possibility of entering some of the data into the system after to help it draft with that being said, Because AI was introduced as I was outside of the clinical career, outside of my clinical practice, I'm not really sure how successful that would be or necessary. I always say that we're paid the big bucks that we are because of our clinical reasoning. So I still do think it is our responsibility to, you know, evaluate that ourselves. With that being said we said earlier. There's systems like Pearson that you're entering that information, and it's generating for you some summaries. Again, you're deciding what you're including and what you're not including. But I'm not sure that I would say in the evaluation part is where I would be utilizing AI. Now, if that was moving forward to the creation of an IEP, maybe entering my de identified evaluation and asking for some goal recommendations in the goal areas that I identify, that might be a nice step. We're always looking to kind of expand our goal bank, if you will, and then utilizing that system before you've even started the treatment to help you develop, you know, data tracking sheets so that once you start your interventions, you're able to track that data to determine if they're, you know, making effective progress. Yeah.
Jayson Davies
And I think a piece that school based ot practitioners can really use is during their like synthesis process, because I think we often question ourselves if we're thinking the right, you know, recommendations. And I think if we use it like a thinking partner and and talk to it, hey, you know, based upon what this assessment said and my clinical observations, I'm thinking this, you know, tell me where I'm right. Tell me where I'm wrong, tell me where what I'm not even thinking about, and get some ideas that way.
Tara Mansour
So maybe another way I might think about utilizing it is I've seen OTs in the past write handwriting goals that a student would likely never be able to do because they don't have the reading skills to match the expectations for the written output, right? So we don't necessarily, we're not necessarily masters of our Inter you know, professional colleagues tests that they're doing, but we might say, you know, again, de identified, how do their results on these tests impact their success with handwriting, right? What does develop, you know, what is a reading level of? I'm going to make something up, you know, for mean, oh, that means that they are, you know, at the syllable level, okay, well, then I don't want to write a handwriting goal around writing full words or sentences, because, you know, they're not necessarily able to, you know, if they have such significant challenges with encoding and decoding, yeah, we have to work within those boundaries.
Jayson Davies
Interesting, I hadn't really thought about that that much, because we often don't understand all the assessment tools that the speech therapist use, or we don't understand the psych assessment tools, or the teacher assessment tools. And that actually came up with me a few about a week ago, my son came home with assessment. He's in preschool right now, and I never heard of it. I don't even remember the name of it, but I was really quickly able just to take a few snapshots of the assessment results. And it's like, oh, this is what this actually means. I was like, Thank you, because there were no keys to what these graphs meant and whatnot on this test. So that can definitely be helpful for us to better understand the student in a way that we couldn't before, because we didn't know what the assessment results that the teacher or psych or SLP were giving us good point exactly.
Tara Mansour
And it's important that we have that whole picture right when we are creating our goals.
Jayson Davies
I guess, moving on from the evaluation piece, you've talked about an IEP and how it could be helpful with generating, you know, drafts or ideas for goals, ideas for maybe what the present levels of performance might look like, maybe even accommodations, rather than asking you the specifics of how you might do this, I want to go a different way with this question and ask, have you created, or have you advised others at all to create custom GPT prompts or custom gpts or that kind of help make this a little bit quicker, because. You're doing it so often.
Tara Mansour
So I will say right now I have two Capstone students working on developing custom gbts, one to support our field work educators, and one, believe it or not, to support their fellow peers through studying for, you know, their own courses I have, you know, I will say that within the conferences I have presented with we started at the level of playing in the sandbox, of utilizing it for things such as, I have a 30 minute session. You know, it is a six year old with this diagnosis, this diagnosis, with these goal areas. I mean, in diagnosis, in school based practice matters less, right, but they it is a six year old with these goal areas. They love trains. This is the format of my typical treatment session. Give me treatment ideas and a timeline, and that is both thrilling and exciting for my audience of practitioners, I think that what you're describing would be like next level, intermediate, advanced, right? I think people have to be comfortable before they can even wrap their head around a custom GBT. I've been playing around with it myself lately. I think there's amazing potential for it, absolutely, but I don't think that is your entry point for practitioners to utilize.
Jayson Davies
Ai, yeah, we've been playing around with one. I created one specifically to help students or not students. Sorry, I created one specifically to help ot practitioners take their de identified, present levels of performance, their goals for a student, or sorry, their present levels of performance, maybe some accommodations that they have in place, and any other information about the student, of course, de identified, and the tool will then, basically automatically, without Any additional prompting, create smart IEP goals. It will create some treatment ideas based upon those IEP goals. It will create both pull pull out and push in ideas. It also kind of will generate, if you tell it what grade the student is, it'll say, hey, this goal is related to these standards for that grade level, which I think is a piece we sometimes miss a school based ot practitioner. So I think it can be very helpful, but it's, it takes, like you said, a little bit more advanced, and it takes some time to set up in order to actually use it. So, yeah, yeah, all right, we talked a little bit about evaluation piece. IEP, we really started this, this process with talking about treatment planning per se. Where do you feel like the best opportunity for AI use exists for school based ot practitioners? Do you feel like it is that treatment planning piece, something we've already talked about or something that we haven't discussed yet?
Tara Mansour
I do think it's the treatment planning piece, and I think it's the data collection piece. I will say that you know, learning to take data that is not just a conversation you have with a teacher in the hallway is something that is our speech and language counterparts are so good at, and I don't think that OTs are always great at collecting data. So I think saying these are my goals. This is my timeline. How can I track this is an amazing use to show that you know our role is justified. I just had a student finish level two field work, and she speaks to how amazing it was to personalize sessions for her children, for her students, and how much more motivated they were to participate in the session when she came to the table with pictures, she quickly printed of resources, you know, of their favorite Disney characters or their favorite something. And so I think that the hard, the clinical reasoning part of you know, what level do we want to perform this at? Do we want to be asking them to copy letters? Do we want to be asking them to, you know, write it from memory? Do we want to be asking them to write words or full sentences? But then we can kind of say this person is a reading level of four. Again, they love trains. Give me some words I can ask them to write that have to do with trains that are able to be, you know, encoded by somebody with a reading level of four. And now, in an instant, you potentially can get so much more buy in and ultimately it, you know, the impact of that is not just efficiency, but it really is about, you know, increasing students participation in performance, and that motivation might help your your students, like, become even more engaged. Watched in their sessions.
Jayson Davies
Yeah, yeah. I've heard the same thing from from therapists as well, really being able to kind of bring the interest of the child into therapy much more easily. I mean, if you wanted to bring bluey into your session five years ago, you had to go back and research Bluey, watch some bluey episodes. Now you can just go on and say, hey, look, this is my treatment plan. Let's, let's implement the bluey version of this treatment plan. And, you know, put a few details, like their reading level or their writing ability and and you could get a lot of pretty ready to go stuff. So yeah,
Tara Mansour
and look, I'm a theater kid at heart. You want to have a conversation with me about Broadway shows, I know a lot, but if you want to have a conversation with me about how the Patriots did in yesterday's game over my head, but I can do a quick search in a system that looks at the web that if I get it wrong because it's not validated, that's fine, but like, I could easily bond with somebody in the first 30 seconds of building rapport, you know. Think about how many times, you know, our middle school students come into an evaluation session with the hood over their face, and they don't want to shut they don't want to say anything, you know. But if you can do a little bit of work in advance and find out what they're interested in, all of a sudden, you can come to the table with some facts that maybe wow them, or, like, at least engage them, and then all of a sudden, you know, what used to take, you know, a lot of time. You can shorten that rapport, building time. Good point.
Jayson Davies
Very good point. You mentioned a little bit about the research that you got going on, but I do want to give you an opportunity to share a little bit more about that, because I personally think that, and I think you kind of share this as well, that OTs need to be addressing AI in multiple fronts, otherwise there might be some negative consequences. And so I want to understand what you're working on in terms of AI. You mentioned a few of your your students doing some custom gpts? But what else you got going on?
Tara Mansour
Sure, so you know again, what was already published was from something that we did in 2024 about this education process. I told you about this learning activity, and we measured for the students, what did they learn? What did they like about use learning to use generative AI? What were the ethical challenges, safety challenges? And so that's something that we did and was already published. I just recently published something else that we need to study a little bit further. It was kind of an opinion piece on how can we use things like voice GBT to train our students with professional behaviors or difficult conversations, or how to explain the role of OT to a patient, to a family member, how to have a difficult conversation with a field work educator and be able to practice in a psychologically safe environment, right? And so again, depends on the prompting, because you're asking it for feedback, but that is one pretty cool way I'm currently having my students utilize it, and something that we need to get a formal study around. One of the bigger projects I'm working on most recently is this idea of using AI for qualitative data analysis of field work educators narrative comments in the field work performance evaluations, where there is so much valuable written feedback that clinical educators or field work educators provide about students strengths, their challenges, their professional behaviors during level two field work, but traditionally, the qualitative analysis of narrative data is incredibly time intensive. You're hand coding comments, you're building theme structures, you're checking reliability, and you know, it limits how large a data set most ot education faculty can realistically analyze. And so what we're studying right now is how AI can assist with this early stage qualitative coding and theme clustering, again, with a human researcher absolutely still in the loop validating and refining the findings. The goal is not to replace your typical qualitative methods, but to scale them in a responsible way so that we can use that information to, you know, adjust our curricula, better support our students and better prepare them for clinical practice.
Jayson Davies
Yeah, it's it's going to be amazing. The next five to 10 years of research with AI being embedded into every facet of the world, it's going to be interesting. I mean, in OT alone, I can only imagine all the studies that are going to be if not already happening right now. I mean, we've used AI to support some of the research I've done, and I know others are doing the same. I cannot wait to get to an. Inspire in Anaheim and and wherever it's going to be next year, and just see how many, how much research is going to be coming out that is AI supported OT, basically, to a degree. I think it's going to be interesting. I think it's something that we got to do. I've said it several times that like AI will not replace us, and we're very fortunate in OT, by the way, that we have registration, like we're certified, we're licensed, we have some protections, and I think we need to lean on that, but so AI probably won't replace us soon, but we need to use AI to basically enhance OT and to keep it at the forefront of Health Sciences.
Tara Mansour
So yeah, we do. You know, I've heard some valid concerns, right? Like, if you speak to folks, especially those in outpatient and acute care settings that are dealing with significant challenges in productivity demands, I think there, I've heard the concern, and it's a valid concern. Well, if you make my documentation time even more efficient? Are they going to expect me to see even more patients, right? So that also limits people's interest in adoption. Per se, they're not going to be eliminated. AI is not going to take them to the bathroom or teach them how to shower, right? But how many people you might have to work with could change if things get, you know, potentially more efficient.
Jayson Davies
Yeah, I mean, I think it's hard, it's gonna be hard to get more efficient, because we already have these high I mean, I'm in the school based role. We don't have this nearly as much, but I hear people in, like, acute rehab. It's like 90% plus efficiency. Like 90% of the time is supposed to be with a client, and I don't know how much more you can really get out of someone, even if documentation gets easier, but I'm sure, yeah, private companies will squeeze when when given the opportunity, so hopefully it occurs in a way that is sustainable, because otherwise, yeah, won't, won't be good.
Tara Mansour
Well, I think that's where our own research and being on the forefront and being proactive about how we best utilize it in our practice is really important, rather than waiting for someone to tell us how we should be utilizing it in our practice,
Jayson Davies
I think that is possibly the most important point of this episode, because we really do have to understand it so that we can make recommendations based upon it. Otherwise, someone else is going to make those recommendations, and it's going to be someone who is at the hospital above our head, is not an OT and is worried about profits over patients, and so absolutely got to do that. I was going to ask you, I think I still will ask you this one, it was kind of on the same, same line that we're talking here. And that is, if it's not that, what are you most concerned or fearful about? Ai, is it that, or is
Tara Mansour
it something else? So what worries me most? I guess, I mean, yeah, while I do recommend the importance about teaching reflection. I think what worries me most is over trust and over automation. People treating AI output as an authority instead of just an input. If we do stop questioning, verifying and reasoning, that is indeed a problem. I'm also concerned about privacy, equity of access, and, you know, tools being used without the proper literacy or guardrails. My hope is that, you know, in OT, we begin to model this balanced path forward of innovation with ethics, efficiency, judgment, and always, always, you know, a human in the loop approach.
Jayson Davies
Yeah, that's going to be key, especially for us who are so, you know, whole human centric as occupational therapy professionals. And, yeah, we got to, we got to keep that in the loop. Tara, it's been wonderful talking to you. Where Can anyone interested in you and your work learn more about you.
Tara Mansour
So you know listeners can find my work through my presentations and publications, both in occupational therapy and Health Professions, education, I mean, especially around AI, literacy, clinical education, field work support. I regularly present at OT and health professions education conferences on practical and ethical AI use in both teaching and clinical practice. I'll be at a ot a inspire in California in April, and I'm running a pre conference Institute focused on using generative AI for content analysis, for both educational and clinical applications, it's going to be a very hands on and grounded in real ot workflows, ethics and that human in the loop practice. You can find me on LinkedIn. I've made some wonderful connections for research and collaborations. I with people across the country, so you can find me there.
Jayson Davies
Absolutely Tara thank you so much. We really appreciate it. We will link to all those different sources, your research articles, as well as your LinkedIn so anyone interested can learn more. And you're going to be in Anaheim for like five, six days. Enjoy that. Have fun. Thank you. Thank you. I'm looking forward to it as well, so we'll see you there. And yeah,
Tara Mansour
thanks again. Wonderful. Thank you so much, Jayson.
Jayson Davies
All right, that wraps up episode 196 Thank you, Dr Mansour, for sharing your insights on how AI can support and, of course, not replace our clinical reasoning as school based ot practitioners from treatment planning to data tracking, it is clear that when we use AI thoughtfully, it can help us to free up time for what matters most, meaningful student centered practice, if you're looking for more practical resources, professional development and real time support from a community of school based ot practitioners who get it and are in the thick of things just like you check out the OT schoolhouse collaborative. It's where you'll find mentorship from myself and other school based ot practitioners, access to tools like our ever growing goal bank and documentation templates and a space to connect with other practitioners navigating the same challenge as you are. Head on over to OTSchoolHouse.com slash collab to learn more. Thanks for listening, and I'll catch you in the next episode.
Amazing Narrator
Thank you for listening to the OT schoolhouse podcast for more ways to help you and your students succeed right now. Head on over to OTs schoolhouse.com Until next time class is dismissed.
Click on the file below to download the transcript to your device.
Be sure to subscribe to the OT Schoolhouse email list & get access to our free downloads of Gray-Space paper and the Occupational Profile for school-based OTs. Subscribe now!
Thanks for visiting the podcast show notes! If you enjoyed this episode, be sure to subscribe on Apple Podcasts, Google Podcast, Spotify, or wherever you listen to podcasts.
