Building Disability Inclusion Into AI Policy And Practice

The scariest part of AI isn’t the sci-fi stuff, it’s how fast it can change the tools people already rely on to live and communicate. We sit down with Dr Julie Eshleman, a postdoctoral researcher working across Georgia Tech and Georgia State, to connect the dots between AI-powered assistive technology, Medicaid waiver systems, and the policies that will decide what comes next for disabled people and their families.

Julie breaks down her participatory action research project building a RAG model chatbot trained on state guidance to help families navigate Medicaid waiver applications for home and community-based services. We talk candidly about what makes the process so brutal, why waitlists can stretch for years, and how practical support like clear FAQs and resource signposting can reduce friction while people wait. From there, we zoom out to AI policy and regulation, and the problem of disabled voices not being treated as essential stakeholders even when AI rules directly affect accessibility, privacy, and discrimination risk.

We also dig into real-world AI accessibility wins: large language models on AAC devices that speed up communication, smart home technology that restores control over one’s space, and everyday AI features that act like external working memory for neurodivergent users. Then we tackle the messy middle: AI literacy, misinformation, hallucinations, and why the better question is often whether something is correct rather than whether it looks real. If you care about disability inclusion, assistive tech, and responsible AI, share this conversation, subscribe, and leave us a review so more people can find it.

Transcript

Welcome To Access Chat

Neil Milliken 0:00

Hello and welcome to AXSChat. Today we're joined by Dr. Julie Eshleman. Julie is a postdoctoral researcher, research fellow at Georgia Institute of Tech and a Lend Fellow at Georgia State University. Now I'm assuming that LEND is nothing to do with uh interest rates and money. So maybe you can tell us a little bit about your work on systems level disability inclusion and AI.

Julie Eshleman 0:27

Definitely.

Neil Milliken 0:27

So welcome, Julie.

What The LEND Fellowship Teaches

A Chatbot For Medicaid Waivers

AI Policy Risks For Assistive Tech

Julie Eshleman 0:28

Thank you. I'm glad to be here chatting with all of you. So the LEND Fellowship, as you know, we love acronyms. So LEND is the leadership education in neurodevelopmental and related disabilities. So LEND and fellow, because I'm postdoctoral. And what it is, is an interdisciplinary training opportunity that lasts for two semesters. So from the fall semester in through the spring semester, their graduate courses at Georgia State. And they invite self-advocates from the community, family members, family advocates, and allied health professionals. So people like occupational therapists, audiologists, physical therapists, school psychologists, you know, all of the different kinds of people who are in the disability space. And all together we get to have really challenging conversations about disability and culture and education and all of the different issues that impact disabled people and their families and the communities that are around them. So it's a really unique opportunity to have hard conversations with people you wouldn't really normally have the opportunity to really dig into tough topics with in an incredibly safe feeling space. So it's been absolutely life-changing experience-wise. I have really, really enjoyed it. And then let's see. So I have two AI, I'm doing AI research at both of the universities. So at Georgia State, as part of that LEND training, everybody has the opportunity to participate in a participatory action research project. And my action research project has built on previous cohorts of LEND trainees who have identified a need in the community in Georgia to help families navigate the application process to get certain Medicaid waivers so that they can have home and community-based services instead of institutionalizing their loved ones. But the application process is brutal. And the wait list in Georgia is very, very long. So my project group has built a closed called a RAG model chatbot that is trained on updated state guidance from the Department of Behavioral Health and Developmental Disabilities. And it answers kind of frequently asked questions that people's case managers and their wait list facilitators and administrators don't necessarily have the time to answer over and over again for all different people. So we have just made a chat bot that can answer basic questions, tell people where they might be in the process, and then signpost them to resources that they can use in their community while they wait for a decision on their waiver application. So that's the work that I'm doing at Georgia State. Just that little project. And then at Georgia Tech, I'm doing kind of bigger picture research and looking at what kinds of AI policies and regulations are being considered at the federal and state levels, and what those policy mechanisms, whether they're kind of incentive-based strategies or they are deterrent-based strategies and guardrails for AI, what those policies will mean for disabled people and especially disabled people who are using AI-powered assistive technologies. So if we start regulating AI in certain ways, it's going to impact the tools that people have probably built into their lives to the point that they kind of rely on them. And so just understanding what those impacts will be so that we can be watching, we as disability advocates can be watching for those policies to come across and will know what those impacts are and can speak to what will happen to the disabled community just so that people are aware.

Antonio Vieira Santos 4:05

So my question goes exactly into that. But no, who is actually shaping the policies that could govern how disabled people can use them? Are disabled voices at the table?

Waitlists That Harm Families

Julie Eshleman 4:20

No, not really, as is usually the case. I think that it's not seen as a disability issue. You know, we talk about in Lend all the time, but all policy is disability policy. All policy. So it doesn't have to be a separate thing to be disability specific. All policy impacts disabled people. And so I think this is just one of those spaces where because of the lack of diversity in the pipeline into technology and sciences, there are just not that many disabled people who make it to the table there. There are not that many disabled people who have had access to the entire education pipeline and support pipeline to make it into that level of expertise to be included in those conversations. And that is one of the things that I'm trying to work on is identifying those voices and trying to amplify them so that there is more knowledge in that space. But there are disabled people using these tools. They deserve to understand what is going to happen to them too.

Debra Ruh 5:13

Julie, thank you for being on the program today. I was very excited. And I just want to comment on one of the things you said earlier when you were talking about the wait list is very high in Georgia. Sadly, the wait list is very high in every single state in the United States. I'm in Virginia. And Julie, I walked that with my daughter when my husband was dying from dementia. And I was in, you know, I own my own business. So it made our family a little stable as one of the partners was leaving. What we had to do to get that Medicaid, so that my daughter, who now lives in a group home and she's so happy. She's so happy. I'm so happy because she's so happy. But it was so intense. And actually, Julie, there were people that were telling me to get it. I had to say I would kick my daughter out of my house and let her just be homeless if they didn't help me. And I was like, I'm not going to say that. And they said, if you want this, Deborah, you better say that. So I just want to say that's so wrong, and I'm so glad about the work you're doing. But it it is so bad for families in the United States, which is so very sad to say.

Julie Eshleman 6:19

It's horrible. There, there are people in Georgia who have been on the wait list for 12 years, like have applied, have acknowledged that they have a really pressing need for supports and have not gotten them for 12 years. Frequently, there are administrators kind of at the state level who will just kind of secretly say, you need to take your kid to an emergency room and leave them there. Yes, yes. I was told things not come pick them up, which is horrible and is abandonment and feels horrible for a parent and a family. But that's pretty much the only way to get services fast. And it's really traumatizing. The entire process, even if it goes smoothly, is really traumatizing for families. So we're trying to help signpost to uh, you know, parent-to-parent resources and support groups and you know, day programs and things like that that people can access while they're going through that really traumatic time, waiting to have the right supports to live in their community and their homes and stay together.

Debra Ruh 7:16

Yeah. And I will also just make this comment and then stop on this topic. I was all I also, because my family was stable, we were doing good, my daughter was healthy until she wasn't, right? And so I didn't get on the waiting list because I said, I'm not gonna be on the waiting list. There's too many families that are more in trouble. And I was told, Deborah, please get on the waiting list so we can see what the real numbers are. Thank you. But I wasn't gonna do it because it was like so many people are in trouble. But those are kind of the kind of decisions that families are having to make. And then of course, unfortunately, I did get in trouble because no, we're human. I am so thankful that my daughter has the community waiver. I am so grateful for that. But every year they threaten to take it away from her, even though she has Down syndrome and pretty significant health issues. Anyway, this isn't about me, but I am just so grateful for your work. Go ahead, Neil.

Why Neurodivergent Users Adopt AI

Neil Milliken 8:09

Yeah, I mean, obviously we have different systems in the UK and in in Europe as as well. But I think there's a common trend that there is a tendency of governments to see disability provision as an area where they can save money by cutting services. And that's a that's a worrying and disappointing trend. Even governments that purport not to be right wing are trying to find savings in disability benefits and healthcare support and stuff. Whilst we don't have the same structure in the UK, we we do have the long wait times for access to services. So for for example, if you can't afford to pay, depending on which area you're in and which region you're in, you may wait for maybe five years for an ADHD diagnosis. So it's everywhere and it is, as you say, traumatizing. But that that that that topic was not the one I really wanted to to ask Julie about. You were talking about sort of AI-powered assistive tech in particular, sort of the neurodivergent community. And and one thing that I've observed is that this community, I'll say our community, because I'm part of it, is really uh finding new utility in AI-powered tools. I've I've worked in assistive technology in the in and around the sort of dyslexia and neurodivergence space for 25 years. And for 15 to 20 of them, the technology, whilst it got better, hadn't really changed. What was on offer hadn't really changed. It was things like mind mapping, text-to-speech, spell checking, speech recognition. Right? And that those were the tools, right? And now what you've got are things that are much more context aware, that are able to support you in different ways. So, you know, things like I I was joking with Deborah before we came on air because there was a bit of confusion as as to whether which meeting platform we should be on. And two note-takers had joined the meeting and and no people, right? And I I keep I keep throwing otters out of my meetings. But as an example of a tool that is beloved of the neurodivergent community, the note-taking, the inferred actions, the sort of organizational capabilities. So when when we see sort of early adoption and and and AI programs, quite often there is quite a lot of sort of failures in some of these early programs because they're not structured. But the one group that is really wedded to the idea of AI is the neurodivergent community. So, what are some of the things that you're working on and investigating in terms of AI-powered assistive tech for this community?

Julie Eshleman 11:06

So the biggest, the biggest use that I can see in literature at least, is the large language models on AAC devices. So it is that word recognition and being able to like type a couple of letters and have word suggestions pop up and then be increasingly accurate as a person continues to use that program. That is an absolute game changer for people who have never been able to communicate at a regular conversational pace before. It has usually been very, very effortful for them to communicate, not just for the person, the disabled person who is communicating, but for the recipient of that communication to have to wait or to have to have a certain amount of attention span to watch them touch letters that aren't like are on a letter board rather than on like a keyboard and a screen. It takes a lot of cognitive effort and it has really, really changed how people can communicate with each other at a more kind of social, socially accepted pace. So that is one really big one. The other one is the smart home technologies that kind of make an ecosystem that has some predictive health analytics or predictive behavioral analytics. And so by that I mean things like an Eco-B or the Google Nest or, you know, whatever the Amazon one is, that they can kind of learn between the pods, the smart speakers in the house, and the temperature controls and windows and doorbells and power locks have given people a hell of a lot more control over their own spaces. We were implementing with a charity in the UK that I work with, have been implementing assistive technologies like that to give people more control over their space, even in really high support settings where inherently there's not a lot of independence built in. There's a lot of stuff that is done for people. And so they don't have a lot of ownership and autonomy of their own space. So we've been building in things like ways for them to control their television, their music, their windows, their doors, their, you know, phones and communicate and lights, fans, all of those things. So those are the other ones that can really reduce the effort required to make your space the way you want it because it can start to learn that you like these, you like this temperature at this time and you like this temperature at this time, and you like the lights to come on or turn off at these certain intervals, which really reduces the amount of effort that it takes for someone to hold in their memory, remember to turn the light on when it's getting dark outside, it will automatically do that. So it's it's reduced a lot of the cognitive effort and really increased the quality of life because it it improves how comfortable you are in your space for a lot of the time. So I think those are kind of the big ones. But like you mentioned, in our in our community, I mean, the AI-powered technologies that I use, and I I always kind of hesitate to call them assistive technologies, because at this point they're assisting everyone. It's not because I'm disabled and using it that has made it assistive. It's just it's just technology now that is accessible. But I'm gonna call her Iris, spelled backwards so that I don't wake her up. But my goal, Iris, is my entire working memory. Like every time something flips into my brain, I'm like, Iris, remind me at three o'clock to email Mark about the thing. Iris, remind me at five or remind me in an hour to change my clothes over from the washer to the dryer. But is it little pieces of help like that throughout the day that keeps my brain just a little bit calmer because I'm not holding this entire to-do list in my working memory. It's freeing that kind of cognitive bandwidth up to do deeper, more intense things. Whereas before, my brain would have been so bogged down by my to-do list, I wouldn't be able to do things in between those tasks.

unknown 14:41

Yes.

Julie Eshleman 14:42

So I think in terms of organization and just keeping my life on track and just keeping it feeling less effortful to just live a regular day has been a huge game changer for me.

Debra Ruh 14:52

I agree. And and I think it's been a huge game changer for everybody. But as also someone that's also neurodiverse, it's made a huge, huge, huge difference for me, just because of my neurodiversity. But I also wanted to say, um Julie, that this is stuff that we've been doing or trying to do for a long time for our aging society as well, making sure that when my mother, who has passed, leaves the house, that the doors are locking and that the garage door shutting. And whenever they're coming back, it's coming on the lights. So we were trying to do that already because so many people are aging over a certain age now. And so once again, I think it goes back to what you said in that this is something that it helps all people. But AI has been a game changer for someone like me with my neurodiversity, which some of my symptoms seem to be getting worse because of the world, I guess, but also thinking maybe my age and my desire to get so much done. I I was just I'm hoping that the work you're doing is also getting to the aging community.

Julie Eshleman 15:53

I mean, I think a lot of people we we kind of say the disability is the only minority group that anybody can join at any time. And I think a lot of people age into the disability community, and some people find their way in a different way. But definitely technology that impacts that impacts disabled people and makes just life a little bit easier to manage for some people will make life possible to manage for other people.

Debra Ruh 16:17

Yeah, well said. Antonio?

Antonio Vieira Santos 16:19

You mentioned about participatory initiatives, and we're not talking about how AI is supporting people. I'm curious to know if there are any relevant data and information on AI and AI helping people to have voices. Like I give an example. Someone who never blogged can now use AI to blog so people can know what that person thinks about something, no, their views of life. Do you see cases like that happening? Or yeah, okay.

Julie Eshleman 16:51

Absolutely there are. Yeah. I I think for people who have really struggled, I mean, maybe with basic kinds of communication, but now because there are so many ways that someone can interact with a computer, it's not a matter of, you know, 15 years ago, maybe more. It would have been a matter of having someone who will sit next to you while you touch the ABCs on a piece of paper and then write down what you said, like that's effortful. And it's expensive to have people kind of sitting and supporting that way. But because now you can put a keyboard on your screen and use eye days, or you can use a head mouse, you know, a dot that will that will navigate you through your computer, or you can dictate with your voice. Because there are so many ways that you can interact with a computer now, the amount of people or the amount of content that different kinds of people can produce now is astounding. Uh, I think that that has been a game changer in terms of letting us hear voices that previously would not have been heard. There's a woman named Jordan Zimmerman, and she went all the way through compulsory education years without much of a communication system. And she now has a master's degree, and I think she works in DC and advocates for special education policy because she because she did not communicate vocally, didn't have a lot of access to appropriate education, because people assumed a certain amount of learning capability because of her not using vocal speech. So I think having devices like that have let a lot of people communicate in ways that they previously couldn't. And I think people whose brains are inter terribly disorganized like mine just have a very chaotic, autistic and ADHD brain. So just kind of at war with different kinds of diversity, neurodiversity all the time. But it helps me to kind of do a brain dump and say, I really want to talk about this or I want to write an article about this. Here are all of my ideas. Please help me organize it. And then it can give me some strategies for turning it into an outline and then what kinds of points I might want to stick to. And then once I write something, I ask it for feedback. Like, does this narrative make sense? Is this clear? Am I communicating well? Is there too much here? Is there anything I should take out? So now I can produce content that is a little bit more targeted and a little bit more organized, which means that more people maybe get access to it than before when it was a fairly chaotic, a little bit controlled brain dump. So I think that that has made a big difference in getting disabled voices out.

Neil Milliken 19:18

So I think I think that that's these are really nice use cases. At the same time, on the flip side of that, what we're seeing is a proliferation and explosion of content in general. Right. And so people are deluged with information. The is also they're deluged with misinformation. And how can we use AI to address some of that sort of surface of uh information and also assess and fact check and everything else when in itself sometimes AI is not entirely truthful because we have the the whole issue of AI hallucinations. So so we're in a new era of managing information. Is that an area that you're you're looking into in your studies?

Julie Eshleman 20:09

It's not an area that I have gotten to address yet, but having previously been a teacher, I I can remember when I was teaching primary school, the librarians being in charge of teaching information literacy, you know, not just how to just getting excited about reading and finding books that you like and things like that, but also just how do you know that the information that you are looking at is correct? Because when I was a kid, this will age me. If it was in an encyclopedia, it was correct. Like if you could find it in a book, it was real. And that was a reliable source. But now there is a completely different kind of teaching that has to happen so that people know the validity of the sources, how to know if what they're reading is real, because you don't have the luxury of if it's in a textbook and a respected publisher has put their money into it, then it's real. But but since they've made it easier for anybody to make content, anybody can make content now. So we do have to teach a different kind of information literacy so that people know how to check, how to recognize if it's you know AI generated, how to recognize if it's AI generated but was checked by a person. So it's still correct. How to tell if it is something that is completely made up. I think that that's an AI literacy that we have not addressed that well yet. And we kind of just went straight from crawling to running with AI because it got so exciting so fast to use those tools for so many things that we skipped this middle part where we think carefully about what are we using it for? How can we tell when a video is correct and incorrect, when we're looking at election candidates or you know, things like that. How do we know in politics what is real and what is not real? Because quite frankly, we're living in a very wild time right now, and I think things change very quickly. And it is very problematic that people don't really have that base literacy to tell what just got cranked out by an AI bot and what is actually happening. And I think that that is a is a public health crisis. Yes. Because of the way that information can get out so quickly. But I think that it's a matter of AI literacy and needing to work harder on people understanding the good usage cases and how to how to verify things, how to think about what you're reading before you act, those kinds of things.

Neil Milliken 22:29

So a very quick follow-up, and then I know Antonio wants to comment as well. I think that that that lack of literacy and that lack of ability to question and so on is is something that we're going to have to put into the education systems to help people. But we put the genies out of the bottle all already. So how we address that fact as societies and how the the mass media platforms address content moderation is a pressing issue for our age because it's not only does it affect broader society, but it definitely impacts the disability community in terms of how we're portrayed and how we communicate.

Julie Eshleman 23:15

For sure, yeah. And I I was just I was listening to MPR this morning, which is is a source that I generally trust. And they were talking about it's National Public Radio. They were talking about how there is a case in one of the states where a lawyer used AI to generate like a case that he filed. And so it cited, made up casework, which kind of invalidated the report that he's filed. And I I feel like that is going to keep happening because people are not being cautious about how they're using it. But we have warring incentives, right? Like Apple likes to teach us how to use cool features on iPhones because we'll buy iPhones. And, you know, AI will start advertising. You know, they've got, we saw during the Super Bowl, especially, like there are a lot of different AI platforms that are selling their product to people and teaching us how to use it. But their incentives are sell more of the things. It's not for it to be used well. So I think the sources that we would normally count on to train us how to use these tools don't actually have the right incentives. So we kind of have to find the parts of society that do have the right incentives, like places like the Center for Democracy and Technology love their reports. But places like that that are kind of doing the good work of saying, yes, yes, AI is is quite good, but proceed with caution and be aware of these things, and here's what we know, to kind of educate the public. But I think it's going to be important for us to find those resources and amplify them and talk about it. And I mean, everybody wants to talk about AI. We just need our talking points ready so that we can kind of help disseminate.

Guardrails Plus Incentives For Accessibility

Antonio Vieira Santos 24:48

So uh we talk about the importance of digital literacy, but some of these platforms have been proud of themselves on doing a great job making sure to enable AI in their own platforms and making, you know, proud themselves. Oh, this is so making things so real that it's very difficult for everyone and to the user base to identify is this real? Is this not real? Because I know at some of I know there's recent documents from some court cases in the United States where behind the scenes they're being proud of themselves of making things look real when they were not. So I don't think we can do much on literacy there. We need more policy.

Julie Eshleman 25:30

I don't disagree with you, but I would challenge just by saying we might have to shift the narrative from telling if something is real to telling if something is correct. Because it doesn't have to be real to be accurate, right? You can generate a video that is not real, but is delivering good public health information or accurate information about a candidate or you know, something like that during election. So so I would kind of say one of the things that we can do is start to shift the narrative from this obsession with deciding if something is real or not real, and instead figuring out how can we tell if what it's saying is correct or what it's portraying is accurate, or you know what I mean? So I I feel like that's one way. Um but like there's there's an AI Civil Rights Act from of 2025 uh that that was proposed. And it is almost exclusively protections, which I think is great. You know, it's it's deterrent-based policy regulations. And I think that that's important, right? We have to consider, say, you know, the same thing that everybody talks about, right? There's privacy, there's data protections, there's disability discrimination that's built into these algorithms. I think that we know what the song is. You know, we're all singing from the same sheet from that one. But what it doesn't have is incentives to build the kind of market we want, right? So it's got guardrails blocking us from the things that we don't really want it to do, but there's nothing that's pushing forward some of the greatness, right? There's nothing saying we need this baseline accessibility for you to access the marketplace. How about that? There's nothing, you know, there's no incentive for them to build stuff that's really, really good. It's just, it's just disincentive and punishment for building stuff that's bad or harmful. So I think that we need a policy shift, not just in understanding what those guardrails are, should be, and what they mean for disabled people. But I think we also need some market-shaping incentives that make it profitable for companies to build stuff that does good public work.

Owning A Disabled Professional Identity

Debra Ruh 27:30

Agree, agree. I know that we're pretty much out of time, Julie. So I'm gonna hand it over. I know, I know, you have to come back on. And I also need to say Julie is one of our global advisors at Boogeam Strong, which I'm proud of. But I did want to do just a real quick before we hand it over to Neil to close, to say that we were so proud that you were in the United States, Julie. And I know you were born and as American, but you actually lived for a while in the UK and have worked with Leonard Cheshire and Steve Tyler, who we all love. And I just I just wanted to give you a moment just to address that because I liked how you were working on looking at it also autistic and neurodiversity with Leonard. So I just wanted to give them a moment of love. Sure.

Julie Eshleman 28:13

Yeah, absolutely. Yeah, and and actually I I credit Leonard Cheshi with a lot of my identity transformation. Kind of in America, I felt very much like I had to wear two very separate hats. And one of them was a clinical practitioner, you know, I am the professional, and the other one is I'm kind of the disabled person. But the disabled person part of me was very, very quiet because it wasn't really broadly accepted in the practitioner space. When I went to Leonard Cheshire, they were very openly disabled people doing really wonderful, awesome, impressive things. And I felt like that was the first time I had permission to have this dual identity and that it could actually be a benefit. So I credit them a lot with kind of feeling like I have stepped into my power in that way.

Thanks And Subscribe

Debra Ruh 28:54

They really walk the walk. All of their executives have lived experiences with disabilities. Yeah. All of their executives. I mean, I'm doing this disability power index, and it's like Leonard, wow. Yeah, I've done a good job. Over to you, Neil.

Neil Milliken 29:08

Well, thank you. It's been an amazing chat today. I need to thank Amazon for supporting us, keeping us on air, and and thank you, Julie, for joining us today and covering such a diverse range of topics. It's been a real pleasure, and we look forward to sharing this episode with our audience. If you are listening, do us a favor and subscribe to Access Chat because when you subscribe, it helps us get the message out there further.