Episode 348: Claire O’Connor: Exploring AI as a Support for AAC Users

This week, we share Chris’s interview with Claire O’Connor! Claire O’Connor is a speech-language pathologist and first-year PhD student at the University of Maryland. Her research focuses on the intersection of artificial intelligence (AI) and AAC, exploring how tools like ChatGPT can support AAC users better.

Before the interview, Melissa, surprises him with news that his Assistive Technology team, comprising nine members, won the "Most Inclusive Team" award from their county's Special Education Advisory Committee Council! Chris himself also won the "Most Inclusive Administrator" award!

Key Ideas This Week:

Exploring AI as a Support for AAC Users: Claire is researching how large language models (like ChatGPT) can support—not replace—AAC communication by reducing delays, easing the cognitive load of message construction, and expanding short inputs into fuller messages .

Co-Design with AAC Users: The project uses a participatory design approach, meaning AAC users are directly involved in shaping tools and prototypes. This ensures that their voices guide development rather than having solutions imposed on them .

Focus on Empowerment and Trust: The research emphasizes building trust in AI and empowering AAC users. Instead of only testing tools, Claire plans to create a support group to hear AAC users’ experiences and perspectives, fostering comfort and inclusion before running compensated co-design workshops.

Transcript of the Episode

Please Note: This transcript was generated using speech recognition & AI tools; it may contain some grammatical and/or spelling errors.

00:00:00 Chris Bugaj
Foreign. Welcome to Talking With Tech. My name is Chris Bugaj and unfortunately, I'm not here with Rachel Madel. It's just going to be me solo doing this.

00:00:16 Melissa Bugaj
Chris.

00:00:17 Chris Bugaj
Hey. Hi. Hi, Melissa.

00:00:20 Melissa Bugaj
Hi. Can I join you? I'm sorry, you were recording.

00:00:23 Chris Bugaj
I was because Rachel can't be here today, so I was. Oh.

00:00:26 Melissa Bugaj
Can I fill in?

00:00:26 Chris Bugaj
Sure, sure. What do you. What do you want?

00:00:28 Melissa Bugaj
What are we talking about?

00:00:29 Chris Bugaj
What do you want to talk about?

00:00:30 Melissa Bugaj
Well, I have something to talk.

00:00:32 Chris Bugaj
What is it?

00:00:32 Melissa Bugaj
Well, I just wanted to share because you won't share with everybody, but I'm going to. That your assistive technology team was nominated and won an award for our special education committee. And they are a group of parents and advocates and educators and directors. And they work so hard for our students that have disabilities in the county to make sure it's equitable. And every year they have one of the best ceremonies of the year. Parents, community members, business partners, other teachers nominate these wonderful people who do this rich work every day. I just wanted to point out that your assistive technology team. There's nine of you, right?

00:01:30 Chris Bugaj
Nine plus me, nine plus you.

00:01:33 Melissa Bugaj
It's amazing how many people and children you have made a difference for. And it came out very clearly this year because your team won for the most inclusive team in our county.

00:01:48 Chris Bugaj
Yeah. It's kind of a big deal. I'm not going to lie. I teared up a little bit when I got the news that the team, the whole team won the Special Ed Advisory Committee Council Inclusive Team award. In the past, there have been individual team members that had won. And then actually there's two individual team members that won as well. And you won in the past.

00:02:09 Melissa Bugaj
Right.

00:02:09 Chris Bugaj
You were the most inclusive administrator once upon a time.

00:02:12 Melissa Bugaj
Well, and that leads me into. You also won, Chris, individually. I know some of your team members won as well, and they are absolutely incredible. I've been lucky enough to work with all of your team, but you won for most inclusive Administrator. Congratulations.

00:02:31 Chris Bugaj
Thank you. You were saying that I appreciate it.

00:02:33 Melissa Bugaj
I'm very proud of you.

00:02:35 Chris Bugaj
Yeah. So there's an awards ceremony coming up and we'll get to take some pictures. Yeah. And I know people are asking for media, so I was helping like submit videos and pictures of stuff so there'd be a little slideshow behind. Behind the scenes. Quick story about the last Special Ed Advisory Committee award ceremony. The president of the. Of SEAC had. I know her presentations and she had learned to turn the captions on. Right. Something I say at all the presentations and she Was like, okay, let's turn the captions on for the awards. And as I remember going, you were there. I wasn't actually there.

00:03:13 Melissa Bugaj
It got stuck.

00:03:14 Chris Bugaj
It got stuck. The. And it got stuck with a word up on the screen that we won't repeat here on the.

00:03:20 Melissa Bugaj
It was not school appropriate. Yes.

00:03:21 Chris Bugaj
It's only three letters, though, so let your guess. It's only three letters. But still, it was not school appropriate. It got stuck up there. But what we say when the technology breaks. It's just a teaching moment. It's just a teaching moment. Doesn't mean not to use it. Sometimes the tech glitches and then you say, hey, this stuff happens. We're all mature and let's just turn the. Turn the. Turn the computer off and turn it back on again, and then it'll be fine.

00:03:42 Melissa Bugaj
But still, we were very sure the captions.

00:03:44 Chris Bugaj
Yeah, better to have the captions on. Even if there's a naughty word that pops up every once in a while.

00:03:49 Melissa Bugaj
And some people giggle, it's fine.

00:03:52 Chris Bugaj
All right. Thank you so much for saying that. Thanks for. For pointing that out. Really appreciate it. I'm really in here.

00:03:58 Melissa Bugaj
I'm sorry.

00:03:59 Chris Bugaj
I barged in the closet.

00:04:00 Melissa Bugaj
What else are you talking about?

00:04:01 Chris Bugaj
Well, I'm going to talk about. I'm going to segue right into the interview today, which is with Claire O', Connor, who reached out. She's a PhD student.

00:04:09 Melissa Bugaj
Nice.

00:04:10 Chris Bugaj
Who is working on studying the AI and AAC. And so she reached out to me like, hey, I know you're into aac and I know you're into AI. I wanted to ask you a couple things. And so I said, sure. Can we record? And so this is a conversation with Claire o', Connor, talking all about AI and aac.

00:04:40 Rachel Madel
We'd like to thank all of the wonderful Patreon supporters who make this show possible. This podcast is funded by listeners just like you who've signed up at patreon.com talkingwithtech to show their support. Because of the generosity of our amazing Patreon community, we're able to pay Luke and Mikayla, our podcast producer and audio engineer, who. Who keep the show looking and sounding great. Patreon supporters also receive bonus content such as early access to interviews, behind the scenes recordings, additional curated resources and materials, and so much more. Check it out@patreon.com TalkingWithTech now let's head back into the episode.

00:05:26 Chris Bugaj
Welcome to Talking With Tech. My name is Chris Bouguet, and today I'm joined with Claire o'. Connor. Claire, am I saying Your last name correctly? Yes, and your first name correctly. So Claire, tell us a little bit about who you are and what you do and what we're doing here today.

00:05:41 Claire O'Connor
So I'm a first year PhD student at university of Maryland, the College of Information, and I'm also a licensed speech language pathologist in Maryland. My research interests primarily lie within the intersection between AI, AI and AAC communication.

00:05:59 Chris Bugaj
Go, go, Terrapins. Yes. You reached out because you have some research that you're doing, is that correct?

00:06:08 Claire O'Connor
Yes, yes.

00:06:09 Chris Bugaj
How can I help?

00:06:11 Claire O'Connor
Yeah, so right now, originally we had the idea of doing a co design workshop, having our AAC users design prototypes to use for large language models such as ChatGPT. And for those of you who don't know what ChatGPT is. ChatGPT. ChatGPT is a language model where you input prompts such as questions or copying and pasting text, and then it provides you with a response. So for example, if you ask it about something that happened in regards to current events, it will generate a response for you within seconds. So what we want to do is we want to try to take what ChatGPT does and see if it could be replicated to use as a support for AAC communication. So the reason, the motivation behind this study is to kind of target some communication barriers that AAC users experience. So for example, previous work has kind of shown that AAC users experience certain barriers, such as delayed timing when trying to form a response. Also the cognitive burden of, you know, going through all those icons, if they're keyboard users, going through all those letters to try to form those words and phrases. So what we want to leverage ChatGPT for is kind of just expanding on what they're trying to say. So for example, if they're only typing like maybe one to two words or you know, one to two letters per every like 20 or 30 seconds, maybe ChatGPT can kind of expand the phrase that they're trying to say. So that is the main goal of what we're trying to do. As we're kind of getting into the recruitment phase of looking for AAC users who, who would be interested in this, we're kind of just taking a step back and we're looking at it from a more personal lens. Because the goal of this project is not just to use this tool as a support, but also to make sure that AAC users voices are heard. So in order to do that, we want to make sure that they're comfortable with sharing their ideas when thinking about some design considerations with this. So with that being said, prior to the co design workshop, we're hoping to start an AAC support group where AAC users of gather together, share their, you know, benefits and challenges when using your AAC devices in their daily lives and also kind of sharing their familiarity with AI tools such as ChatGPT or Gemini or Claude, so that we kind of have an idea of where they are in terms of, you know, how often they use them, what tasks do they use them for, and you know, how comfortable they are with using them like regularly. With that being said, right now we're trying to target specifically adult users. We're starting off with that and eventually once we see success with adult AAC users, we're hoping to expand it to child populations as well. So that's kind of the overall gist of our plan for what we're doing. I've been working closely with my advisor, Dr. Stephanie Valencia, who does a lot of work with AAC use and merging AI technologies such as large language models for their communications. So we've been working together to try to form this group and to try to get AAC users more involved with the design process.

00:09:39 Chris Bugaj
Okay, awesome. Claire. So let me ask some clarifying questions. You've been using the pronoun we. Are you a team of people doing this or is just you and Dr. Valencia?

00:09:49 Claire O'Connor
I've been working with lab mates, part of my lab, who have assisted me in thinking about design activities and also, you know, ideas to try to engage the AAC community more. So I would say this is more of a team effort. I probably wouldn't be able to do this without them. So yeah.

00:10:09 Chris Bugaj
When you use the term co design workshop, can you describe what you mean by that?

00:10:13 Claire O'Connor
A little bit, yeah. So co design workshop is part of this research methodology called participatory design. And that's where you design prototypes with the research participants. So for those of you who have read previous literature, a lot of the research out there, specifically in the ASHA journal, it's mostly, you know, user centered, where, you know, you have them test out products or you ask some questions, maybe do like a focus group interview, maybe do a survey that's more user centered. So they're not actually a part of the design process. They're just participating in the design that you already created for them. With the co design process, you do set up certain activities for them to do, but they are the ones who design that product. So that I would say that's the difference between co design and other research out there.

00:11:13 Chris Bugaj
Okay, interesting. And I guess another form of Research that you're not doing that doesn't sound like this would be something like, we have this hypothesis and so we're going to test it by, by using very, you know, eliminating variables and try and, and isolate things. That's not it. Right. You're trying to actually create something. Right. But you're using the term research. Is that fair?

00:11:38 Claire O'Connor
There is a hypothesis. I think that's the, you know, basic groundwork for any research project. But in terms of a hypothesis, it's pretty common to have like those research questions that you want to know. So for example, if you're trying to see if something is feasible, if you're trying to see if there's a certain gap in research that hasn't been addressed yet by other people, that's kind of the basis of your research question. Can my methodology, like co design, address these barriers that AAC users experience? That is a hypothesis. What I'm talking about with co design is that's more of the methodology that we're following. So there's a difference between the methodology and then the hypothesis that you're testing, because that's kind of like, again, the framework of what your research is.

00:12:28 Chris Bugaj
Okay, okay. The way I understand hypothesis is not a question, but a statement like, I believe this. And then we have these questions to verify that. And regardless, regardless, it sounds like you have some sort of design in, in place and you're looking for partners to help you co design it. Is that fair? And if that's the case, so it sounds like generally you're asking about ChatGPT or other large language models. Do you have some designs already in mind of how it can be used? For instance? For instance, some apps have said, well, let's see, how can we do this? We could say we need a fringe board for the time we go to the zoo. And previously clinicians or parents or users themselves had to create a board on their device for things they might need in a specific location. And now they can go, hey, prompt, create me 25 icons. That fills out this board for when I go to the zoo. And they hit a button and it fills it out. Is are you that far along in design or it's still in the conceptual phase? Where are we with that?

00:13:41 Claire O'Connor
So right now I would say it's more in the conceptual phase because we haven't actually run the co design workshop. I just want to make sure that, you know, our ideas are preserved and that we are looking into maybe like building off of like what previous research has done in terms of already using large language models out there to expand AAC messages. But in terms of the exact activities we have planned right now, it's kind of just like, you know, thinking of ideas, trying to conceptualize them in a way where it's feasible to do it in an in person workshop. So I would say that's where we're at right now.

00:14:18 Chris Bugaj
Gotcha. So you, what you're looking for is maybe a group of AAC users that come together in person or virtual or both.

00:14:26 Claire O'Connor
We're looking to have AAC users come in person.

00:14:28 Chris Bugaj
Okay. So that means they're probably going to need to be in the Northern Virginia, southern Pennsylvania, New Jersey and then of course Maryland area. Right? That part of the, that's, you know, drivable to get there.

00:14:44 Claire O'Connor
Yeah, pretty much.

00:14:46 Chris Bugaj
And I'm going to guess that there's no compensation for this. It's a sort of volunteer.

00:14:52 Claire O'Connor
Oh no, there is compensation.

00:14:54 Chris Bugaj
Okay, awesome. Tell us about that.

00:14:56 Claire O'Connor
I also want to differentiate to be the support group is completely voluntary that we plan on doing prior, but once we launch the co design workshop, that is 100% compensated for. So any AAC user who we recruit who will be a good fit for the workshop, they would definitely receive a compensation. And then any caregiver that assists them in the workshop will also receive compensation as well.

00:15:25 Chris Bugaj
Awesome. And are you picturing these as like full day events? Like you come and work together for a full day and then go home or spend the night sort of thing.

00:15:33 Claire O'Connor
So I originally I did have planned like maybe a two hour workshop and maybe doing it in the span of two weeks. But I think with everybody's schedule because I know AAC users, you know, they do a lot like they have a lot of therapy that they have to attend to. They have, you know, daily activities they need to attend to. I'm, I've been kind of floating around the idea of maybe doing like a one day thing and making it four to five hours, having frequent breaks in between, of course, and kind of just using this as like a full day where you know, AAC users trial and error, different prompts that they could try with. Check.

00:16:10 Chris Bugaj
Gotcha. Okay. Okay. It sounds like maybe you have activities already planned for. For. This is like you said, try different prompts. What are, what are some of the things you're sort of thinking, what are some of the, the questions you're like more specifically, what are some of the questions you're. You're looking to answer?

00:16:26 Claire O'Connor
I won't go into detail with activity because I don't want to share that yet. I we haven't run the co design workshop and I want to kind of keep that like under wraps. But, but with the question that we're hoping to answer with this, we're hoping to see if ChatGPT can be used as a communication tool. The reason why I say I emphasize the word support and not replacement, a lot of. I actually did hear my advisor talking about this yesterday, but a lot of concerns with using large language models is bias. So when you're using large language models, they're learning not only your own communication style, but they're also learning about the global knowledge that's out there. So for example, like what's in like news articles, what's on like basic, like you know, blog posts. Like they're learning all that information and they're combining it with your communication styles. So for example, they may assume something that may not entirely be correct and sometimes that can pose as a barrier when using large language models because even though like yes, it can provide instantaneous communication and also responses, it doesn't really like, it's not always accurate. And you know, I've learned this in, in school right now. I've learned this, you know, using it for like writing like emails and also like you know, analyzing text. And I always feel the need to like go through and check it because it's not always correct. So in terms of using it as like using it for something that's second nature to us, which is communication, I think it's important to consider that it shouldn't be a replacement, it should be a support. And that's what we're hoping to gain from this co design workshop.

00:18:14 Chris Bugaj
Yeah, the phrases that I've been using in talking about generative AI for any purpose, like I work for a public school. Right. But just in general is as a partner. So it's not doing it for you, you're doing it, it's doing it with you. You're the final say as the human of what your, you know, producing. Recognize that there might be these biases in it, but that I also don't want to that often is a thing that people talk about with large language models, but it's also just the world in general. Like if you go into the and search to do anything on the Internet, it's going to be riddled with bias. Right. So I think pointing it out makes sense, but also pointing out that we live in that I think it's really important when we talk about the large language models. And the bias is that there is often ableism built into It. But that's not often talked about as the bias that is pointed out in many conversations. I'll give you a quick example, Claire. I was doing a presentation on universal design, and I was trying to create an image of a library that didn't have stairs, so multiple levels, but didn't have stairs, like just ramps that would roll. It could not. I mean, maybe it could. I used multiple tools. I prompted it, like prompt engineered for over an hour, trying. Okay, not that. Let me try this. I could not get it to generate an image of a multi tiered library without stairs. You know what I mean? When it creates stairs, be like, okay, now can you change the stairs to a ramp? Okay, but still stairs, you know, like, it couldn't. I couldn't get rid of it. So that bias is. There's certain language, I think that that comes out when you're using the large language models that way. So I think it's really wise. What you're talking about is to be thinking about it as a thought partner, recognizing that it's not perfect, you know, that it's not going to replace anything, that it's a. It's a do with not for.

00:20:10 Claire O'Connor
Yeah, no, I completely agree. I actually, yesterday I attended a workshop for this human computer interaction conference that I'm looking to attend eventually. In the workshop, we were talking about trust in AI. So what. How does somebody automatically trust AI? Like what. What are the certain factors that kind of entice us to use a. So I think identifying that trust first, kind of just identifying what initiates that trust. How do large language models kind of bring us into using it more in order to kind of teach us that it can't be something that we could rely on. It has to be something that can be used as a support. And I think right now with our generation, it's scary. It's used for everything. And I've noticed even my professors that have used. Used AI like this, they, you know, some of them are okay with it, some of them are not okay with it. So, like, you know, it's still like a gray area that we're trying to explore, but just identifying what makes somebody trust AI. What are certain factors that, you know, we trust about it in order to kind of understand, like how we could kind of see it more as a support rather than like a replacement.

00:21:29 Chris Bugaj
For sure, for sure. I mean, like I said, I work in public schools. It sounds. I mean, I know everywhere across the country at this point and for the immediate future, I think is going to be wrestling with how do we reconcile our current practices around education or just life with AI and integrated into our practices? And you said, I hear things all over the map where it's ban it, we don't want it in our schools or we don't want to use it to. Well, but then how are we preparing people to actually use it when they leave? I hear, I say this all the time. It's very reminiscent of the conversations I was having with people, many people were having about 10 years ago around mobile technologies. It's going to revolutionize everything. We're going to have these computers in our pockets. Students can just look up information and find it and synthesize it. And now it's largely banned in schools and thought of as a negative thing. And so I could see that as an, as a potential outcome. But trying not to get too far away from what you're trying to do with aac because I could go on about, about generative AI in just in general practice. Let me ask you, have you had any feedback from AE users yet or is it still too early? You're just in the collection phase of trying to find people to help the, in the co design process.

00:22:45 Claire O'Connor
Like you said, we are in the collection phase. But I actually did a talk with the speech and hearing discussion group in the DMV area. They have virtual meetings once a month. And I did a talk back in February about, you know, this project and what I had planned for the CO design workshop and, and actually use that as an opportunity to gather feedback from SLPs around the area on their thoughts about using AI with AAC users. And it was really surprising as how common it is like to not just with like integrating it with AAC use, but also just like SLP, like as a practice, like so many of them have mentioned, like using it for like creating like lesson plans or you know, grammar checking emails. And those are kind of just like some of the tasks that they had mentioned using it for. And they also offered like, you know, potential benefits of using AI. So a lot of them kind of said the same thing that I said in the beginning of this interview, like increasing timely responses and also, you know, relieving that cognitive burden of forming those messages. I liked how like we were kind of on like we had the same mindset in terms of how AI can be used to enhance communication. You know, I think I did learn a little bit more about how AI was used used in, you know, the SLP practice as well. So it really was surprising.

00:24:13 Chris Bugaj
I mean, to me it makes a lot of sense for the, just the two purposes you mentioned, like enhancing this, the rate of communication. Well, that's really what it's good at in general is that it helps you be again, if you're using it wisely and you've sort of recognized what it can do for you, it's a time saver for everybody. Right. Like I will never submit a proposal for a conference again without using generative AI to help me write that. Right. I mean I use it every single day to make myself more productive. And I feel like, you know, there's this phrase that we use when we talk about it from, from an employment standpoint is that AI is not going to steal your job. It's the people who know AI are going to steal your job, right? Because you're, they're going to be much more productive. So it makes a lot of sense like especially from a productivity standpoint when it comes to speech and language therapy or education in general. Why wouldn't I use it to help me generate more respectful goals, maybe more poignant goals written from a strength based perspective. Analyze my data. As long as I'm doing it in a way that is not, you know, that is not sharing PII or personal identifying information, right. If I strip out that information and just say hey, can you analyze my data for me and send like these are all things. How many people as speech therapists, I think speech, I'm sure you'll, you'll resonate with this is that when you were in college you did a language sample and in practice you never do a language sample again. Or maybe once or twice when you first started and said whoa, this is way time consuming, I'm not doing this, I'll just use. But now you could like hit a record, have it transcribe it, ask the mlu, right? Like ask for errors and it can give you all that information. So there's lots of ways you can use it to increase your, your productivity. And I can AAC users too. You know, when I first, about, about two years ago now when I went to a conference and ChatGPT had just come out, I sat with two AAC users. One who was a heavy word prediction user. And I just said to him, hey dude, have you seen chatgpt yet? And he's like, no. I'm like, it's. I, I don't know how to describe it. Maybe since you're a word prediction user, tying it to your background knowledge is that it's sort of souped up word prediction on, on steroids to, to, to quote you know, to overuse that, that phrase. And he's like, show me. And so it's like, well, let me show you. And he's like, oh my goodness, this can totally enhance my rate and it's going to help me generate thoughts. And then, then again, like, I'm the final person that gets to choose, but if I can edit what's already created, oh my gosh, it's going to save me so much time. Then I showed it to another person who is a direct selector who is super wicked fast at using his AAC to the point where he doesn't really need to look at the device in the same way that I might not look at my keyboard while I'm typing. He doesn't look at his AAC device when he's generating sentences because he just knows where the words are. Super fast. He was like, no, I mean, this is going to slow me down. I'm already faster than that. Why would I do that? So again, I think it's, that's why it's so important that you're getting user feedback is because I think you're going to see that my hypothesis would be you're going to see lots of different viewpoints about how it can be used.

00:27:44 Claire O'Connor
Yeah, no, definitely. I totally agree. And you know, I think again, it's important that we kind of see it as like something that can be useful and not something that can be harmful because, you know, again, like, we're slowly kind of learning about AI as it's kind of like a gray area right now. Like, you know, there, there are just certain, like things that we don't know about AI that we may know within like the next couple months or maybe next year. But in terms of its potential with AAC communication, I definitely do think like a, it can serve AAC users in respect to increasing that response time and also, you know, relieving that cognitive burden. You know, I think like this co design workshop we're going to have plan will almost like empower AAC users to kind of share their input. Because when AAC users are maybe given a device for the first time or maybe, you know, they're given a new device that they have to try out out, I feel like sometimes it's hard to adjust because there's a learning curve like you're trying to learn how to adjust to this new set of communication. But if you're the one who's designing it, if you're the one who's offering those ideas to try to make that communication for yourself, then you know, it kind of mitigates that learning curve and you know, kind of offers you like that accessible communication that you've been looking for. So we're hoping to kind of use the idea for the co design workshop and we're hoping to make sure that AAC seek users like are empowered and can speak their mind in terms of the ideas that they could offer. So awesome.

00:29:26 Chris Bugaj
Awesome. Claire, do you mind if I plant something in your brain, maybe get your feedback on it? Because there's something in this conversation that is rarely brought up and I'd like to bring it up now to put it in your brain. The conversation we're having right now, I've had many times with many people over the last two years because the immediate thought is just like you're saying, wait, oh, let's talk about how generative AI can help people with producing language learning, language rate enhancement, right. They're really focused on the AAC user. There are companies right now, like I said, working on their apps and thinking, how can generative AI help us with our apps? Maybe it's in helping understand symbols better or like I said, fringe boards. Right. But there is a conversation that nobody is having that I'd like to put in your brain. So for years and years and years, being someone who's been supporting AAC for 20 years of my 20 plus years of my career, something that I recognized has been, all right, we picked the thing, here's the tool, I'm going to deliver it to the staff and to the kid user and I'm going to say, here's what you need you to do. I need you to model on this communication device and I need you to give them wait time. And I need you when I by modeling I mean I need you to use a communication partner. And, and over and over and over again, I'd come back and I think anyone listening right now, and you might resonate with this too. How come they're not doing it? And I told them to model on the device and they're not doing. I told them to give them more wait time. Like I explicitly told them what to do and they're not doing it. It's still on the shelf. It's not implemented. They're not teaching, they're not doing any of the things I said. There's the people who win, meaning the people who can say whatever they want to say, say however they want to say it. With AAC coming out of schools is still few. Like the numbers are still small. And so recognizing that was an issue I did some like, okay, well, what can I change? Right? And part of it was helping the communication partner become better at being a good communication partner, and that is through a coaching model. So what. So what we do in our neck of the woods, Claire, and this is people listening to the podcast will have heard this before, but the way it works in our neck of the woods now is, okay, teacher and or parent. We could do better with the parent side, but just say communication partner. Here's the AAC tool. Here's some explicit instruction around it. You need to model on the device. Here's some mnemonics you can use to help learn how to be a good communication partner who models well. But that is a skill. And like any other skill, like throwing a football or, you know, driving a car, it takes practice. And the best way you can practice is if you had a coach. So we're going to give you at a minimum six sessions where we're going to coach you through where you're going to try and do this skill, you're going to try and model on this device, and we're going to give you feedback. That feedback is rarely. Is it directly going to tell you you should wait longer? No, it's not that. It's, hey, remember how we talked about waiting longer? How do you think that went? Do you think, like, you waited long enough? Hey, remember how we talked about you were going to add one more word onto what the. The person said? So that's the, the AC user said one word and we're the, the skill has been to use. Use that word and add one more to it so that they're getting the same repetition with variety around the language that they're exposing that you're exposing them to. Holod is a skill. So here's the coach, you know, how do you think you did with the expansion? And I don't hear anyone talking about using generative AI to help the communication partner get better at the skills of being a good communication partner. Imagine the generative AI being, let me record the session or let me analyze the buttons you pressed on your communication devices, the communication bar partner and give you feedback about what you could have done differently or just ask you questions. And the generative AI being your thought partner because we don't have enough coaches to go around to all the people that, that are, that are communication partners. Just, I know I've been talking a lot, but just let me give you one like scenario here, okay? Imagine a little kid learning how to use a communication device. Their parents sitting on the couch next to them and they're reading a book together. And the parents are meant to try and do, do comment, ask, respond, which is a strategy to, hey, I'm a parent. I don't know how to really read to my kid. Like, I picked up this picture book. I know I'm supposed to read picture books. So I'm just going to read word by word. And I'm not going to model on my device. Someone like you or I, Claire, tells them explicitly, we need you to model on the device. Here's what you do. When you're sitting on the couch with that, with the student next to you, comment on some of the pictures that you see. Ask them a question about what they're seeing, and then respond to your own question. If they don't respond right, do that now do that with gender bi, sort of listening into the background and that it gives you feedback. Hey, you could have asked a question here. Here's some questions you could have asked. Here's some comments you could have made. You made four comments. See next time, if you could make six comments, you know, you modeled on the device six times. Your goal was to get to eight. Here's how you can do two more. Like the generative AI being your thought partner as a communication partner is nothing. Any. I've, I've, I don't hear anyone in our space talking about that use of it. And so I know that was me rambling a lot. But what are your thoughts of using generative AI as a communication partner Support?

00:35:10 Claire O'Connor
I think that it would have potential in terms of maybe measuring the progress of the support partner. But in terms of offering ideas or input, I would say that's still, I would say that's still questionable because not only again, are they getting this from like global knowledge sources, like, you know, on the Internet or like, you know, from a blog post. But I think in terms of subjective support, I think that should come from the clinician because then you're talking about the clinician kind of slow or the generative AI slowly kind of taking the clinician's position. And that's what is a little frightening to me. So I would say that when thinking about how to frame generative AI as a support for the communication partner, I think think, yes, it could be used as a measuring tool to kind of like monitor the progress of the communication partner. But in terms of offering specific ideas, I think it would have the potential if clinician were behind offering those ideas. But to solely rely on the ideas that the generative AI Is, you know, showing the communication partner. I think that's something that should definitely be questioned because I don't think again, we entirely know the capabilities of generative AI yet. And I think again, we want to make sure that it's a support and not a replacement.

00:36:39 Chris Bugaj
In your co design. Right. Model, you're thinking there'll be some sort of product that comes out of it?

00:36:47 Claire O'Connor
We're not sure yet. I think we'll have a better idea once we run it and once we have data, then we could kind of further assess and see where it goes. But right now we're kind of just in the recruitment phase of, you know, looking for AAC users interested and also looking to create our support group. So I would say that's kind of in the wraps right now. We haven't really quite like gotten that far, but I think it would be pretty cool to explore it.

00:37:14 Chris Bugaj
Yeah, I mean, one of the things that I wonder about is more specific GPTs. So again, you, you did a great job sort of describing ChatGPT in general. And so what a GPT is again for, for our listening audience is something that has been designed by humans to be used with ChatGPT that is more specific user base. So for instance, I created one for our podcast where you could, you know, I've sort of trained it to what to look for. So I've sort of put some guardrails around the, the generative AI and said I need you to do these things. Right. And I wonder and not do these things like specifically search through our podcast inventory or our podcast history to look for answers on blah blah blah, as opposed to search your whole, your whole large language model for answers on blah blah blah. And I wonder about a product like that so that it is designed with people who are sort of experts in the knowledge and good clinicians at knowing how to do this kind of, of stuff with ACE user feedback as well. You know, if it was designed in the right way, I might mitigate some of those concerns about replacement and be used as a, as a thought partner with guardrails, you know.

00:38:34 Claire O'Connor
Yeah. Actually now that you're kind of talking about this in terms of designing a product, my advisor, Dr. Valencia also has a product called Compa and what that does is it actually monitors there's AC communication via virtual calls like Zoom. Because when you're video chatting an AAC user, you can't see, you know, what they're typing. You can't see they're trying to respond to a message. Like you could only really observe those Body movements of them, like looking down at their device, typing it, or using the chat feature on maybe like zoom to type their response. What COMPA does is it lets you know when the AAC user is typing so that you could kind of get an idea of how long you should wait until they AAC user respond. Because again, timing is always a huge barrier with AAC communication. So this platform helps mitigate those barriers. And then I believe she also designed a platform where AAC users can choose between generated responses based on. What's it called? Just a couple words. So it kind of just expands their responses. Same idea for my co design workshop. It provides those alternative responses is using a generative AI to allow the AAC user to use more timely communication. So there are definitely products out there and unfortunately they're, you know, I don't think they're as well known as they should be. But you know, I'm hoping I use this opportunity to really like project these ideas out there. So I think like there are certain products out there that I, I'm not sure if they're public yet. I could definitely inquire about that. But I just want to make like reiterate that they do exist. There are certain products there for AAC users to leverage. Generative AI for sure, for sure.

00:40:21 Chris Bugaj
You know, one of the things that I also hear people talking about in this space is sort of the generative AI listening to the communication partner. So you know, I'm an ace user. I come to talk to someone who's using their mouth words, that person using their mouth words uses them and the generative AI listens and generates potential options for my response. Again, I get to choose if I want one of those or not, or if I want to pick one and then edit, edit it, you know, as a, as a start, as another way to potentially increase my rate. Where do I have to generate the entire sentence? And I mean, and I could see that, you know, with like, I could see that rolling into mainstream use. I'm, I'm wearing glasses right now. Sinclair. People can't see me, but you can see me pointing to my glasses. Imagine they're smart glasses and they come up with potential responses in my lenses of like, oh, here's two. Mouth users could, mouth speakers could be having a conversation where they're getting prompts from the generative AI. Again as a, maybe as a social cue. Hey, here's some things you could say that would be polite, you know, as you're learning those social skills. Or here's some things that you could say or. Or big red flashes. Don't say this. That would be a problem. Like, there's lots of ways it could be used to help enhance communication to further social relationships as opposed to. Opposed to a barrier in them.

00:41:49 Claire O'Connor
Yeah, I'm not sure because it's been around for a long time. We have word prediction, right?

00:41:53 Chris Bugaj
Yeah.

00:41:54 Claire O'Connor
On Tolby Devices, we see our AAC users typing and the words come up right away. Yeah. It's not as fast as generative, the responses that come from generative AI, but there is that kind of component of predicting the next word, what the AAC user is trying to say without them having to type the entire thing. So it has. The idea itself has always, like, manifested itself in aac, but I think with AI becoming more prominent now, there's more potential for it.

00:42:24 Chris Bugaj
Yeah, I could not agree more. Yeah, that's. It's just figuring out what the right interface is and making sure that it is truly the person that is making the final call. I'm the final submitter of my paper that I'm writing. I'm the final person who hits the send on the email. I'm the one that is decided that this is the message I want to communicate and that it's, like you said, it's a support. So with all that said, how do people reach out to you? Like, you're looking. What's the. What's the time frame here? Like, when are you trying to get started and how can we help you? You know, sort of spread the word.

00:43:04 Claire O'Connor
I would like to start off with. I kind of want to see if there are interested AAC users. I'm looking for 16 and older. Older. I myself work with a lot of kiddos, but right now I kind of want to only focus on, like, teens, young adults. Adults. So I would say maybe like, you know, 16 and up in terms of the diagnosis. I would. I'm really interested in exploring autistic AAC users. I feel like they're becoming like, you know, autism is a very predominant neurodevelopmental disorder. And, you know, it's really interesting to see, like, you know, I see so many come into my practice. Like, it's just becoming more prevalent now. So I would like to explore more autistic users. And also I, again, I'm really pushing this support group because I really do want to try to make this a comfortable environment for the AAC users. And I want to make sure that they feel comfortable with us to share their ideas, because I don't want to just kind of like, Invite them on board and be like, oh, like, hey, you know, we want you to do this, we want you to do that. And, you know, we barely know them, so, you know, we want to kind of just like, build that connection with them. We want to kind of build that rapport with them before kind of just throwing them into these tasks. Because that's not what research is about. Searching for solutions for your population while also considering the relationship the population has to those issues. So, for example, we could think of a hypothesis. Right. Right now, AI can be used to address gaps in AAC communication. Yeah, sure. How. How do the AAC users feel about using AI? Do they not like it? Do they, you know, use it a lot? Do they have some, you know, skepticism about AI? So we don't know these things unless we build that rapport with the community and we really understand the user. Because if we just kind of recruit users and they, you know, go into the study without kind of knowing, like, you know, what exactly we want them to do, then what's the point? So I really am looking to advertise that we are here to support them and we are here to use our research for good. So, yeah, ideally, I'm looking for autistic AAC users 16 and up. If they are able to bring a caregiver, that would be wonderful. We do encourage caregiver participation as well. And hopefully there are AAC users, users living in the DMV area and beyond the DMV area. So we would. We. Unless they're willing to make the drive to umd, we don't mind recruiting from, like, Maryland, like, you know, the state itself, Virginia. I don't know if Delaware, New Jersey is a little far, but their preference that they want to come all the way down here. But again, we want to make this a safe space and we want to make sure that, you know, the AAC user can use this as an outlet to kind of share their, you know, know, ideas with others and also connect with other AAC users who share the same challenges and also share the same interests. So, again, we want to make sure that it is an opportunity for them, and it's also an opportunity for them to kind of just share their ideas and also make those connections with other AAC users. So, yeah, that's awesome. What I'm looking for.

00:46:28 Chris Bugaj
So if someone's listening right now and goes, I'm in. I want to. I want to learn more. I live in Maryland. You know, I'm in Silver Springs. I want to come be there. What's the way they reach out to.

00:46:38 Claire O'Connor
You, they can definitely send me an email. It's C O c o n n o 5 m d edu. So they could send me an email. I want to make sure that the support group is preferably, maybe like five AAC users. I want to keep it very small and very intimate. I want to have the. The user reach out or user caregiver reach out to me first. I feel like if I advertise like the police and like the time and all that, like, you know, anybody could show up and like, I don't want it to be too overwhelming. So I would rather have them email me directly to get all the details and then we could kind of discuss like, in terms of what's feasible for them. And, you know, do they. Do they want to, you know, travel some time to get here? Because if we do get somebody who's far away, I would want to make sure it's worthwhile coming here and not just like, wasting their time. So, you know, definitely reach out to me at my email.

00:47:33 Chris Bugaj
Sounds like you got a. You're putting a lot of effort into making sure everyone feels comfortable and that it is respectful and that it works for everybody, that everybody leaves with a good experience. Right. So, okay, one other detail here, and that is these podcast episodes tend to live on. On the Internet for impertuity. So when is sort of the cutoff date that you're looking for? Like, oh, man, I missed it. I just listened to this podcast and o, I missed it by this date. You know what, when are the. When's your deadline?

00:48:04 Claire O'Connor
So in terms of the support group, I'm hoping to kind of get that up and running by within the next couple of weeks. Then with the co Design workshop, I'm hoping to start that this summer. So like, June or July.

00:48:16 Chris Bugaj
Gotcha.

00:48:17 Claire O'Connor
Okay.

00:48:17 Chris Bugaj
June or July of 2025.

00:48:20 Claire O'Connor
Yes. You don't have to wait another year.

00:48:23 Chris Bugaj
Awesome. Awesome. All right, well, we'll make sure we have that information, your email address, address in the show notes, if that works for you, so that people are like, wait, what was that? They can just click on the link and grab your email and. And send you an email asking you other questions or telling you that they're interested. And then I'll be interested to hear how things go, you know, after it all comes together, I'd love to. To catch back up with you and see what the results of the. Of all of this turns out to be.

00:48:51 Claire O'Connor
Yeah, no, thank you. I'll definitely follow up with you as well. And, you know, hopefully we'll have a better idea of where this goes.

00:48:59 Chris Bugaj
Awesome. All right. Any. Any last words, Claire?

00:49:03 Claire O'Connor
No, I. I can't wait to see who reaches out, and I hope that, you know there are people interested in, like, helping pursue this initiative, so thank you so much.

00:49:12 Chris Bugaj
You're very welcome. All right. Thanks, Claire.

00:49:14 Claire O'Connor
All right.