Disability Community Hearing on Artificial Intelligence (PM Session)
>> Good afternoon, Madam Chair. You may begin the hearing. >> Katie Kale: Thank you. Good afternoon, and welcome to the U.S. Access Board's artificial intelligence virtual hearing for the disability community. My name is Katie Kale. My pronouns are she/her and as a visual description I am a white woman with long light brown hair.
I'm wearing a blacktop under a corral Blazer, a very bright corral Blazer, and I'm wearing dark rimmed glasses. I'm sitting in my office in front of the American flag and the General Services Administration flag. Thank you for joining us today. I'm a federal board member of the U.S. Access Board, as well as the Deputy Administrator of the U.S. General Services Administration or GSA. I am proud to serve as the Chair of the U.S. Access Board, as well as chair this hearing. The Access Board's mission is to create access for all the Access Board is an independent federal agency dedicated to accessible design for people with disabilities. The board is comprised of 25 individuals,
13 of whom are appointed by the President and the other 12 are representatives from various federal departments. I would like to acknowledge my fellow board members who have joined us here today for the hearing and thank them for making time. I would also like to take a moment on behalf of the board to additionally thank today's presenters and all of those in attendance for being with us today as we discuss artificial intelligence in relation to the disability community. So a few housekeeping notes as we begin. The hearing is being recorded. American Sign Language
interpretation and realtime captions are being provided. All attendees will remain muted with your cameras off until you are being called upon to speak. You may then use the chat feature to communicate with the host if you need additional assistance. For all board members, presenters, and for those that have preregistered to offer public comments, please take time now to ensure that the name that is listed on your Zoom screen is your full name so that we can easily identify you and provide you the permissions needed to speak during the hearing. On the screen, you'll find the agenda for today's hearing. To go over the agenda, after my welcome remarks, we will
begin the hearing with Alison Levy, Director of the Office of technical and information services for the Access Board. She will provide some foundational background on our AI series. After Alison, we will hear from a wide range of presenters on the uses, benefits, and risks and barriers for people with disabilities regarding AI. Following the presentations, Access Board members and staff will be able to and panelists questions. Afterward,
there will be time for public comments from those who have preregistered to provide public comments. Let us begin by welcoming Alison. >> Alison Levy: Thank you, Katie. It's a pleasure to kick things off with a little background information. Let's start off with the first slide. First, for those of you who are new to the Access Board and who we are and what we do, I'm just going to provide a little information about our roles and responsibilities. We were established to develop accessible design guidelines and standards, including Architectural Barriers Act, the Americans With Disabilities Act, section 508 of the Rehabilitation Act, and others. We also provide technical assistance and training on these standards and guidelines and we also enforce Architectural Barriers Act standards, which apply to federal, facilities, and federally leased facilities. Next slide, please.
So last fall, President Biden issued executive order on the safe, secure, and trustworthy development and use of artificial intelligence with the intention to maximize the benefits and minimize harms of emerging AI technologies. To that end, the Access Board was tasked with several asks. Next slide, please. Three items. First, we were asked to solicit public participation and conduct community engagement. Second, we were asked to issue technical assistance and recommendations on the risks and benefits of AI, including biometrics as a data input. And last, we're hoping to provide people with disabilities better access to information and communication technology, as well as transportation services. As a result of those asks, we developed a memorandum of understanding with two key partners, two national, nonprofit, disability based organizations, the Center for Democracy and Technology, otherwise known as CDT, and the American Association of People With Disabilities, AAPD. You'll hear from them
in just a few minutes, but I wanted to fill you in a little bit and bring you up to speed on where we are with our outreach efforts. Today's session or this afternoon's session is the third in a series in which we're hearing from different folks in the community about AI risk and benefits. We kicked off our series first with a foundation of information on Artificial Intelligence to help reach out to all of our communities to establish a foundation of understanding on artificial intelligence. The second was a hearing hosted this morning to the disability community, and we heard from some wonderful presentations which highlighted risks, benefits associated with the use of AI in the disability community. And then that brings us
to this afternoon where we are hearing from a second group of people. We're very excited about the panelists coming up. And our next issuance of a will be on August 22nd where we'll hear from partitioners in the artificial intelligence field, as well as federal agencies, what they're doing in the AI space. So with that, let me kick things back to
Katie for the next introduction. Thank you. >> Katie Kale: Thank you, Alison, for that foundational information. I'm excited that we will now transition to the panel presentations part of the hearing. Our panelists will be sharing general AI benefits,
AI risks, and touch on the current research, employment, and other related AI topics. We ask that all panelists keep their presentations to around eight minutes, and if they have slides, they should be prepared to share their screen at the time of their introduction. As a reminder, please keep your cameras off until it is your time to present. We ask that all others remain muted with cameras off. Okay. So we now welcome Henry Claypool, Technology
Consultant from the American Association of People With Disabilities and Arianna Aboulafia, the Policy Counsel for the Disability Rights and Technology Policy Center for Democracy and Technology for their presentation on MOU partner insights on AI for people with disabilities. Henry and Arianna, you're welcome to begin. >> Arianna Aboulafia: Good afternoon. I'm Ariana Aboulafia. I'll give a brief visual description. I am a young ish person with sort of past shoulder length brown curly heir. I have gray rimmed glasses and I'm wearing a black T shirt and a gray Blazer. So I'm going to be speaking very briefly today just about our partnership a little bit. SI am the policy Council for Disability Rights and policy
for center for Democracy and Technology and I lead disability rights work. I am so grateful for the partnership with AAPD and with the Access Board and our work with Executive Order implementation to engage the disability community and those interested in understanding how to ensure that AI is deployed in a way that is helpful for people with disabilities and to sort of paraphrase the Executive Order in a way that increases benefits and also minimizes and mitigates the harms for people with disabilities. So this is our third, I guess, session in this group of sessions on AI and disability. We had a disability community session earlier this morning and we also had a session a couple of weeks ago that was entitled a foundation on AI and disability. And that one is on the Access Board website.
And I wanted to give a quick little recap of some of the things we spoke about there. So we gave definitions of AI, because it's really important to sort of couch these conversations in what it is that we're talking about, what it is that we mean when we talk about the impact of AI and algorithmic systems on people with disabilities. So we defined AI and machine learning and generative AI. We discussed some of the uses of AI. We mentioned that, you know, oftentimes federal agencies will use AI as ways to hopefully make things more streamlined and faster and easier and sometimes even less biased and that, unfortunately, that is not always the outcome. We mentioned some of the ways in which people with disabilities are using AI. Sometimes in
assistive technologies. And we also mentioned how AI can impact people with disabilities right now. We talked about how Artificial Intelligence tools are being incorporated into all sorts of different systems that people with disabilities interact with over single day. That can include education and employment. That can include healthcare and benefits determinations. And we mention some of the ways in which those sorts of tools, when used in those contexts, can have discriminatory outcomes and problematic outcomes for folks with disabilities.
What you're going to hear in today's session is hopefully more about some of the ways in which people with disabilities use AI, some of the benefits, some of the risks, some of the potential concerns, and hopefully some ways to recommend mitigation of those concerns. These listening sessions provide the Access Board, AAPD, another opportunity to center the experiences of people with disabilities and to understand the impact of AI and emerging technologies on our community, because it's an opportunity to foster dialogue with federal agencies and with industry on the use of AI. And it gives us the opportunity to continue to raise awareness for how disabled people experience AI, because AI tools, they are being used right now, which means that some of the benefits and also some of the concerns, they are happening right now.
But these tools will only continue to be incorporated into more and more aspects of our everyday life. So it's really important that we hear from the community now, that we start to hearing from the community, that we continue to hear from the community as to how folks are interacting with and experiencing these tools. Our next to last session and our next session is going to be on August 22nd where we look forward to hearing from federal agencies and other practitioners, and then we'll have a final session in mid November with the Access Board to release a report that will hopefully support future work on AI from both within and outside the disability community. So we are really looking forward to this hearing and to continuing our partnership with the Access Board and with our community as well. And thank you so
much. I will kick it back over to you, Katie. >> Katie Kale: Thank you, Arianna and Henry for your work on this. Very important topic, and thank you for your presentation. So next we are going to welcome Erie Meyers, chief technologist at the Consumer Federal Protection Bureau, CF PB, on Biometrics of Data, people with disabilities, and the impact on credit ratings. Erie, you are welcome to begin when you're ready. >> Erie Meyers: Thank you so much, and good
afternoon. My name is Erie Meyers, as you heard. I'm chief technologist from the Consumer Federal Protection Bureau. I am a white woman with brown hair. I'm sitting in my home in a floral chair in front of a map and lamp. I'm so glad to be here today. My agency regulates consumer finance. These are things like credit cards, but also debt collection and student loans, and there are a huge number of things that we regulate or oversee that impact communities such as the one we're talking about today. So I want to start with the really good news, which is that there's actually no exception in the law for fancy technology. That
means that the rights that you have won in sometimes very long fights are still yours. That means that the obligations of companies to follow the law don't go away just because they're using AI or fancy technology. And the federal government is building more technical capacity to make sure that companies are following that law. So I want to give you a couple examples of places where the CFPB is working hard to make sure that that is the case today and special areas of interest for the community of folks with disabilities. The first is around housing. Federal regulators have levied allegations of potential collusion in rental markets around the price of rent. One of the other things we've also seen are
allegations around the use of automated valuation models for the pricing of homes. These technologies have challenges even in traditional settings with the addition of advanced technology. It's critical that the firms understand how the technology is being used, how the data that is training the models is selected, and how those models are unleashed on the public. Our agency is working hard to make sure that these methods are not used to evade the law; that your homes are not rated as less valuable because of the color of your skin; that the cost of your rent is not going up because of illegal conduct. Another area is around data brokers and fraud. We are working on a regulation around the fair credit reporting act that says if a firm is collecting data about you and using that data to make a really important decision about your life, they need to follow the law. So when this law was passed 50 years ago, I think folks traditionally thought of the Credit reporting agencies. Many people know about your
credit report or sort of the big three. But in 2024, the data economy looks different. So our job is to make sure that firms, regardless of, again, how fancy their technology is, are really following the law. You may have seen allegations levied by the Department of Justice against several data brokers for knowingly selling the data of people with diminished capacity to firms that were known fraudsters. Those are the types of fact patterns we're very interested in ensuring don't proliferate or expand, and we're really looking forward to our forthcoming rule making on data brokers to address things like fraud or potential threats to our national security, but also to make sure that personal health data and biometrics are not used to decide what opportunities are available to you, what options you have when it comes to getting a job or finding a home, anything where you would need a credit report or a background check in order to get access to those services.
Additionally, our agency has done work around chat bots and potential violations of law. I think you all may have had the experience where you're trying to get help from a company or even just a straight answer and you can get caught in sort of a doom loop. You ask a question on, you ask for help, and you get passed from person to person or in a web chat bot I ask a question, ask for help, and you get an answer that doesn't make sense. Maybe it's robotic. The firm might want you to think it's simply a hallucination. But when you're relying on a firm to give you a straight answer so
you can access your rights, it's much more serious than that. So my agency has put out a paper around the use of chat pots, both generative AI, cutting edge technology chat pots, and also very simple wins to say that the existing rights that everyone has in the United States when it comes to getting a straight answer from a financial firm remain in place, despite any fancy technology being used. The last thing I'm going to mention and then I'll turn it back over is that my agency accepts consumer complaints on these topics. If your bank, if a debt collector, if a credit reporting agency, if your student loan servicer isn't giving you a straight answer, if you're concerned about having meaningful access to the services you are due, we are very interested in hearing from you about what is a happened and how we can help. At consumer finance.gov/complaint or on our phone line at 866 411 CFPB, we have folks ready to help. You can
file online. You can file on the phone I believe at 160 different languages. We have materials, including videos in American Sign Language and other products to increase the accessibility of our complaint system, and it's critical we hear your stories and where you're stuck, because it is our obligation to ensure that these laws are followed. So thank you for the time today. I'm excited to hear more and learn more. Thank you. >> Katie Kale: Thank you, Erie, for that presentation and all of that great information. We will now welcome Lydia Brown,
Director of Public Policy, National Disability Institute. They will be presenting on AI Risks for People With Disabilities. Lydia, you may begin. >> Lydia Brown: Hello. This is Lydia XZ Brown, pronouns they/them. I'm a youngest
each Asian person with short brown hair and glasses. I am wearing a dark green and blue top and behind me there is a fake background that shows wall to wall and floor to ceiling bookcases in a light filled room. People with disabilities have long experienced discrimination in all sectors of life, resulting in disparities in outcomes economically and otherwise, including a poverty rate that is twice that of non disabled people, as well as unemployment rates that are about double that of non disabled people, disparities that are greatly exacerbated when data is disaggregated to account for racial and gender based differences alongside disability. So it is no surprise that the presence of AI and
algorithmically enabled decision making systems can further exacerbate and amplify those existing disparities. This is for a number of reasons. Win is reliance upon data that represents existing inequities and disparities. If an algorithmically determined tool, for example, is making decisions about job candidates, housing, or about a credit determination and that algorithmic determined tool is relying upon existing data, it will replicate existing inequality. Conversely, if the data that a particular tool is using is actually unreliable to begin with, the data is not necessarily accurate, the data is not necessarily representative of the full range of experiences of disability of disabled people and of the full diversity of the disabled community, then even an otherwise well designed system will be relying on un useable and reliable data to make its determinations or its assessments. The purposes for which AI generated tools might
be used very widely as well and generally fall into the categories of assessment, evaluation, and assessment that operate together. Of prediction or of decision making. And in all of those realms, reliance upon data that either replicates existing inequality or data that is unreliable will result in further inequitable and potentially harmful and discriminatory outcomes for disabled people. Another way in which AI tools can be particularly dangerous for people with disabilities is if they are not designed with the needs and the functionaries of disabled people in mind. In particular, if a particular tool requires input from or engagement of a user that assumes that all people communicate in certain ways, that all people's bodies move in certain ways or have certain functions, then all people receive and act upon sensory input, whether visual, auditory, tactile, or otherwise in the same way and with the same mechanisms for accessing a device or accessing a software program, then those programs will inevitably end up discriminating against people with disabilities, because in the worst case scenario, they will be entire inaccessible to a disabled person and in the best case scenario, they may tend to exclude people with disabilities or at least make it much more difficult for people with a range of disabilities to interface with and engage where that particular tool. with that particular tool. AI based technologies further have the potential to harm people with disabilities when they are supporting Public Policy aims that are not in line with goals of people with disabilities and the Civil Rights and independent living movements and the self advocacy movement led by disabled people. So for instance, if a state agency adopts an algorithmic tool with the explicit
aim of reducing the number of individual people who are recipients of public benefits programs, that is a goal that is explicitly at odds with a Disability Rights Movement that aims to ensure that all people with disabilities are equipped with the support and the services that they need to be able to participate in society. And given the currently high rates of poverty and unemployment for people with disabilities, rates that are disproportionately high for our community and higher still for those who face multiple forms of marginalization, that is a goal that will inevitably result in a greater proportion of people with disabilities from being removed from a public benefits program. If the Public Policy aim surrounding the use of a particular tool is not in alignment with goals of community integration, civic participation, and engagement in the mainstream economy, then that tool is posing a very real and immediate risk of harm individually, as well as at a macro level for the disability community. Further, AI tools that tend to have a discriminatory impact on disabled people will entrench discriminatory practices and policies in society. If it becomes more acceptable, for instance, to use an algorithmically determined method of choosing who will be considered for a job, for access to a rent stabilized unit or a subsidized housing program, or who will be able to maintain access to publicly funded healthcare benefits and long term services, then it may become acceptable in the public mind to allow existing inequities to grow. Without addressing
the underlying Public Policy inadequacies of current research, of current practice, of current services, and of the ways in which disabled people are able to live their everyday lives, not just interface with systems, then AI tools that we develop will not exist outside of that context. And that ultimately is what is necessary for policy makers to understand if they don't always know what is happening, because the average policymaker cannot be expected to be an expert on every aspect to the disability community, and certainly not on AI and other algorithm driven tools. But policy makers do have a responsibility to know that the policies that they are working to develop and implement are done with the needs. People who have the most to lose, who are most vulnerable, and most marginalized, front and center in mind. They have a responsibility to ensure that the policies that they adopt will not result in reinforcing, perpetuating, and amplifying inequity injustice. And so AI tools that we have public funding for,
public support for, or public procurement of, in particular, should be held to a high standard of accountability and should be held to standards of accountability that reflect back what the priorities and needs are of the communities that are directly affected by them, whether that is in relation to credit decisions, in relation to access to housing, in relation to access to employment, or in relation to access to public benefits programs, as well as a myriad number of other applications possible for algorithm driven tools. It makes sense when building and designing policies to collaborate with people who are from impacted community. And both technologists and policy makers have a great deal to learn in that respect from members of the disability community who are well represented here today in this discussion. AI tools exist in context, and that means that to the extent that there
are many potential opportunities for benefits, for disabled people from growing applicability of AI tools to everyday life functions, that it is all the more important for us to be cognizant and conscious of the great risks that occur with the existing utilization of AI tools, as well as with continued development of AI tools where we are still working to create policy that works for human beings, where we are still working to create policy that makes sure that people have access to the workforce and education and housing and healthcare. And work that we spend time and attention that we spend in relation to AI needs to reflect at the end of the day, it is human lives that matter, and risks of AI are enormous. And if we work to address them now, we'll be better positioned to benefit for the potential opportunities that AI can offer to making the world more accessible and inclusive for people with disabilities instead.
I'll turn my time back over. >> Katie Kale: Thank you, Lydia. We will now hear from Theo Brady, Executive Director, National Council on Independent Living on AI benefits and impacts on independent living. Theo, you may begin. >> Theo Brady: Great. Let me get started. Again, my name is Theo Brady. I want to talk a little bit about myself. My pronouns are he/him/his. My video description, I'm a black many, male,
bald headed with a salt and pepper beard and glasses, blue glasses. I'm a C4 quadriplegic who uses a complex motorized wheelchair. I've been a person with disabilities since age 15 due to a football accident. I ran the Center for independent living for 31 years and I am currently Executive Director of the National Council on Independent Living, as well as a member of the national Council on disability. So let me briefly take to you about a number of things actually Lydia mentioned in regard to the benefits and opportunities for people with disabilities in regard to AI. And so I have a few of those things to share with you. And keep this
in mind. AI is moving so quickly that probably by the time this presentation is over with, there's going to be some advancement. So mine is definitely not inclusive of everything, but I want to point out a few things. Assistive technology. Speech recognition. Right? Tunes like Apple series and Google systems and Amazon Alexa help individuals with mobility, disabilities, control devices and access, information using voice commands. So it's very important. Text to speech and speech to text. Software associated with dragging, naturally speaking, assist those with
individuals disabilities or more disabilities by converting spoken words into text and vice versa. There's mobility and a half irrigation. Autonomous vehicles. It's going to be a game changer. Right? Self driving cars like those being developed by Waymo and Cruz and other companies who provide transportation options for people with physical disabilities. And again, I say it's going to be a game changer. It's going to be something that's happening right now and certainly in my lifetime. Navigation apps. Apps like AIRA and BeMyEyes and human assistance to help individuals with visual disabilities navigate their surroundings. Very key. Screenreaders and keyboard navigation
allows people with mobility disabilities to also navigate and communicate augmentative and alternative communication. If anybody knows Bob Williams, he's doing a lot of great work with community first with this. AI powered devices help individuals with speech, disability communicate more effectively. Speech assistance in regard to particular text, to assisting and forming
sentences. Realtime translation. AI tools like Google Translate facilitate communication for those who are deaf or hard of hearing by providing real time translation, spoken languages into text. Very useful to people. Home automation. Very important. We saw this even more so during Covid. Smart home devices. AI driven devices like smart thermostats, lights, air purifiers and security systems. Something that was very beneficial in helping people with is mobility disabilities manage their home on a daily basis. This stuff
is advancing so quickly. I encourage you all to stay and use these kinds of smart devices. Lotus ring. Something fairly new was presented at the National Council on Independent Living conference R a wearable ring that controls objects at your home by pointing at it. Again,
technology is moving so fast. Health and wellness. Again, wearable devices. Like smart watches and finger rings can monitor a person's health on and alert them when potential health issues come up. Again, something I'm constantly using myself
and encourage other people to use this, this kind of technology. Telehealth services. AI enhanced telehealth platforms by remote consultation. Reducing the need for logistic, you know, getting back and forth. We all know that people with disabilities, some cannot afford affordable, accessible transportation and can't get to a doctor's appointment like they should, and again, Covid 19 proved this to be very valuable and those things are continual post Covid.
Educational tools. Personalized learning. AI can tailor educational content to meet the unique needs of students with disabilities, providing adaptive learning experiences. Accessibility features and education software can offer features, such as closed captioning. Audio descriptions, customized interface to improve, again, accessibility. Even if the workplace, we see AI taking effect, tools that remove biases of hiring. All right? And
one of the things that we got to be very careful about, and I think Lydia mentioned it, we've got to be very careful with that, that these AI tools are not being developed based on ableist thinking so they create another barrier that people with disabilities have to face on a day to day basis. These applications help with task automation and making things more accessible in the workplace in regard to assistive AI and employment. Lastly, I want to finish up with talking to you, something personally I can testify to in my own life, during emergency situations. Dropping my phone or falling, but still being able to contact someone using voice command, voice activated commands. I can personally testify that it worked in my life. During the pandemic, it was very difficult to a period of my life and I'm sure many others, but I was still able to use AI, assistive devices, to do my grocery shopping, my banking, and so on. Household devices. I use it for my lights, my thermostats, my air pure fire, and my heater. Heat and coolness can be very did he have state to go a person with a physical disability,
and I was able to use that AI to ensure my own safety. Then with cameras. I'd use ring video cameras to see who was at my door. Was able to lock it and unlock my door. Speak to the people before I even let them in. And we also know how important it was during Covid 19 or any other kind of emergency how important it is to access information in realtime. AI allows that to happen. All right? Air quality. We talked
about that already in regard to air purifiers. We know how difficult activities can affect a person with a disability. So you can also use that. And again, I mentioned telemedicine. My doctor was rarely available. I couldn't come out to see her, but when I brought her in on video technology, she was right there. Even my pharmacy was based on automation and they could send my medicine. Socializing during that stay at home period, we all know about that, how important Zoom calls was. We needed to be entertained, and again, we used AI for that purpose. And not only that, we visited and virtually visit our family members to ensure that they was also safe and also could be socialized. And I'm using right now, I'm working at home,
like many other people are working at home remotely based on technology, AI technology. And we saw post Covid that employment for people with disabilities increased simply because of AI technology. And again, we're talking about Alexa. We're talking about Amazon, Echo, Cube, Fire Sticks. All of these things are voice activated that allows you to function in your home safely
and securely. Apps can enable a number of skills, such as morning routines. Right? I'm talking my medicine. Sometimes this becomes very important for people to be able to take medicine on time and often people forget what these routines, once set up, can help you. And I can't even talk about how many times I used Audible books and Kindle books to read when I'm not in a sitting position as well as Facebook Portal, that can follow you around as you take care of business. Smartphones.
All of these things are very important, vital. Right? I wish I had stock in Amazon, because I use them all the time. DoorDash and Instacart. My wife don't even shop anymore. She makes me do it. All right? They had are the things that AI is doing that enhances our life. It used to be where AI and technology like this was an afterthought. People would develop this
stuff and they would say, oh, well, this is going to benefit people with disabilities. Now big businesses and companies are intentionally looking at what these things could do to enhance the lives of people with disabilities. So again, I say by the time this presentation is over with, there's going to be some technology that is going to improve the lives of people with disabilities, because it makes sense and it's good for business and people every day are contacting to say can you terrorist this or can you test that in regard to autonomous motorized wheelchairs are being developed. And these are the kinds of things that AI are going to benefit all types of people in the near future. So this was not completely everything, so just look forward and become educated in regards to all the things technology wise and AI wise that can benefit the lives of people with diverse disabilities. Thank you. >> Katie Kale: Thank you, Theo. I especially liked during the beginning when you said by the time your presentation would be done, there probably would be more AI uses that are out there. I think that's a great
reminder of how fast everything is moving. All right. Next we're going to hear from Robin Troutman, Deputy Director, national Association of Councils on Developmental Disabilities on the AI impacts for People With Developmental and Intellectual Disabilities. Robin, you may begin. >> Robin Troutman: Thank you so much,
everyone. And I just spilled my water. Give me just a moment. I'm Robin Troutman. I use she/her pronouns. I'm the Deputy Director of the National Association of Councils on Developmental Disabilities or NACDD. For visual description, I am a white woman and I'm wearing glasses, a blacktop, and I'm wearing a beige jacket. In. ACDD works across the 56 U.S. states and territories to ensure that people with intellectual and developmental disabilities can live in the community of their choice and lead a self directed life. I'm excited to join you today for this important hearing on artificial intelligence and the potential impacts on people with lived experience. Thank you to the U.S. Access Board for inviting NACDD today.
A little story. Back in January, I attended a conference for meeting an event planners as part of a professional development organization that I am a member of. They had three keynote speakers at different times during the event on the main stage talking about AI, how AI is here whether we are ready for it or not. How AI should be seen as a tool and not a replacement. And the biggest take away from all three in the exact same words, AI is not going to take your job or replace a human, but the person who uses AI as much as they use Word, Excel, or Canva will take the job or advance faster and further than those who do not. But what does that mean for people with intellectual and/or developmental disabilities? We have seen already that advancement in technology help people with intellectual and developmental disabilities live more independently, as Theo said, be able to work more efficiently and effectively and be able to hang out with their friends wherever, whenever they want. But these advancement only help those who can access them. We still have a large digital divide in our country and globally where more rural and
poorer areas do not have access to stable, high speed internet, which many of these tools and technologies require. According to the World Health Organization, there are more than 2.5 billion disabled people that will need one or more assistive technologies by 2030. However, the WHO, World Health Organization, also states that almost 1 billion of those people cannot access the products. So we have some work to do. As my colleagues have mentioned, since the Covid 19 pandemic, we have seen an increase in applications like Zoom or Teams and AI generated captions, website accessibility overlays, and other tools, and people, including the word accessibility in their Diversity, Equity, and Inclusion conversations as they should. But these decisions are being made not by the people who need to use them, but by technology experts as they should be at the same time, because they are experts in coding and programming. But people with intellectual and developmental disabilities are experts in their needs and what works for them and what does not work for them.
If we're able to take a line from the hit musical Hamilton, people with intellectual and developmental disabilities and other disabilities need to be in the room where it happens. They need to be included on research and development teams, included in user experience teams and demos, and in impact or evaluation teams. AI is only as unbiased as the data and algorithms it relies on. We need to ensure that AI is not only intelligent, but also ethical, inclusive, and aware of human diversity. AnneMarie I'm going to get her last name wrong. AnneMarie Imafidon is a British computer scientist and social entrepreneur and she states
that when a wider range of perspectives inform how new advancement such as AI are designed and used, the resulting technology is more equitable and beneficial for everyone. This isn't just a computer science thing. Understanding society is crucial when releasing technology into the world. For example, similar to what Lydia mentioned in her presentation in their presentation, I apologize, similar to what Lydia mentioned in their presentation, if you're designing an app for jobseekers, you need to ensure you have people actively seeking employment involved in user testing and have people with disclosed disabilities involved in that user testing. STEM, science, technology, engineering, and math, and STEAM, science, technology, engineering, arts, and math, need to represent an inclusive cross section of society to create the best outcomes possible. People with disabilities and especially those with intellectual, cognitive, developmental,
or neurodivergents disabilities must be included in the STEM and STEAM workforce. By including people with intellectual and developmental disabilities in the AI and computer sciences and information technology workforce, and let's be honest, all areas of employment, it is less likely to have bias in the AI algorithms. Currently, the algorithms are consciously or unconsciously discriminating against people with disabilities. AI systems learned from vast datasets, which often reflect societal biases. If these datasets predominantly feature able bodied individuals, the resulting algorithms may fail to accurately interpret or serve people with disabilities.
Developers might not fully consider the needs of people with disabilities, leading to products that are inaccessible. For instance, as Theo mentioned, voice activated systems might struggle to recognize the speech patterns of individuals with speech difficulties. In addition, AI driven hiring tools might screen out candidates with gaps in employment history due to medical conditions or autonomous vehicles might not recognize wheelchairs as obstacles, posing safety on risks. To combat ableism in AI, it is essential to adopt a more inclusive and ethical approach to AI development. AI systems must be trained on diverse datasets that include representation of people with disabilities and other intersectionalities. We must involve
people with disabilities in the design and testing phases of AI development and there must be ongoing training and education, as well as listening, communication, and transparency. This is not to say there aren't good tools out there currently. Microsoft's AI for accessibility, for example, projects include AI tools that help blind or low vision people navigate the world and apps that assist those who are deaf or hard of hearing. In addition, there is voice ITT, which is a speech recognition app designed to understand nonstandard speech patterns. The impact of AI on the intellectual and developmental disability population is still relatively new to determine, but if we are mindful of including people with intellectual and developmental disabilities early in the AI processes, then we can see it be used to improve the quality of someone's life, increase employment opportunities, and educate a greater understanding of people with disabilities, all of which creates a more inclusive society. Thank you.
>> Katie Kale: Thank you, Robin. We will now hear from AnneMarie Killian. She's executive officer, TDI for Access, Inc. and Jeff Shaul, software developer, GoSign AI on the research and impact of AI on sign language interpretation. AnneMarie and Jeff, you may begin. >> AnneMarie Killian: Hello. First, I'm just testing to make sure you can hear the voice interpreter. Good. You can hear. Okay. Perfect. So first of all, I want to introduce myself. I'm AnneMarie, CEO of TDI for access. And I am with my
partner, Jeffrey Shaul, who is cofounder of GoSign AI, and I am going to do my image description. I am a white woman, middle aged, brown hair wearing black glasses and brown shirt. And before we continued, I just really want to commend the U.S. Access Board for organizing this important event. As we've seen here, in the comments and presentations shared today, it's clear AI's presence in our daily lives requires that we selectively work together to ensure the safeguards and no undue harm on people with disabilities. Today we are presenting an emerging technology for language access representing the advisory group on AI and sign language interpreting. Our advisory group, primary goal is to develop inclusive language and access standards for all communities, including sign language users. Our goals, we act on our counterparts
to ensure interpreting standards and access to all communities, including sign language. As we act as counterparts to the interpreting and safe AI taskforce, both align and we collaborate together with interpreters and spoken languages. We've been working with significant resources through #DeafSafeAI, reports are available and you can see the guidance. We study, we do the research and we set that up, the policies and our findings to ensure that safe AI has standards. Engagement and activities involvement to promote and implement these standards. We engage with community through various activities, including webinars, sympathy pose ups, and workshops all over to gather the data for our findings. And it's clear to get the messaging and our mission
to technology to provide equitable for all disabilities. And I would like to transition to Jeff who will expand a little more on the technical considerations and motivations that are working. So I'm going to turn it over to him to discuss the efforts and the highlights, the disposition of the benefits of AI. Jeff? >> Jeff Shaul: Thank you, AnneMarie. And I just want to make sure everybody can see me. Perfect. Okay, thank you.
Thank you, AnneMarie. My name is Jeffrey Shaul. And I'm a software developer for AI company called GoSign AI specialists in signing captions and data gathering. My image description is I am a white male in my Thirties. I have short brown hair
and blue eyes and a little bit of stubble and I'm wearing a lavender collared shirt. All right. So one of my favorite quotes by a famous scientist and scholar is the future is already here it's just not evenly distributed. We've seen this happen many times throughout history. For example, the telephone. It became common in the early 1900s. The at large population reached the benefits, while the Deaf and Hard of Hearing community was left out, until the 1960s, when the TTY arrived. There are countless other examples now emerging throughout history. So with the emerging technology, we hold a great potential, but we have to be assertive in ensuring that it's evenly distributed among the communities. And that new technology is here
in artificial intelligence. And that word is so over used. It's mostly a buzzword for marketing. I prefer to think of it as a model, a model that detects patterns in new data using prior. Imagine if you asked a kid to draw a picture of a human. What is a kid going to draw? Possibly it's going to draw a stick figure. They haven't been exposed to the art and training and they haven't seen all the examples of the art and had that. So a kid internal, their model is not competent for that. It's not complete. So the same with sign language and the model for recognizing
and translating and creating and generating. How to make sure that it's complete? There's so many different considerations that need to be made. First of all, the datasets. They need to be inclusive. Sign language is just so diverse across the nation and the country.
People sign the same word many different ways. So the datasets need to be inclusive and include all different classes of signing. You know, gender, ages, sex, race. It doesn't matter. So imagine if the model training only was on College age applicants and volunteers. They wouldn't perform as well encountering to the senior citizens. Just like in English speaking, they have accents. There's many signers that have accents as well.
For example, in Maine, their sign for their Capitol City in Portland is this. It's the same sign as the color purple in other areas. So it's a conflict. So the model needs to be known that it's a nuance, and there's many more examples that contradict that. So the model is only as good
as the data that it is exposed to. There's many other considerations as well. Just like reliability, you know? How to make sure that it is robust and, you know, what if the power goes out? What if the internet drops? And lastly, how do we evaluate and make sure that we have the checklist and to guarantee that that model is inclusive and safe and effective? And how to communicate that with the public. So the considerations on our minds are when the advisory group and the taskforce developed policies that, quote/unquote, automatically interpret by artificial intelligence, quote ain't quote, AI, by AI, should follow. So now I will explain the principles that we developed with the feedback in the signing community. First of all, the user autonomy. It means that they should be able to decide
to use it or not. We need to empower the user to be able to make their own decision. Secondly, the model should improve the safety and well being for all users. They should feel empowered if they're technically not feeling trapped. And they should be able to not feel that they're stuck and they're in this box. The third model is strength and weakness. It
should be transparent and clearly communicated with everyone so that people can make informed decisions. And fourth of all, it's very important that providers should be accountable for their errors and harm that isn't cured by the models that they provide. If you want to learn more, please, you can refer to the QR code and the link to read the guidance document that establishes and justifies these principles. And you can learn more why we've established the principles and support the principles as well. So that is what the advisory group is all about. We call upon the U.S. board to take action that aligns with the strategic plan. I pulled out a few of the specifics,
objectives from their plan. First, your first objective is the technical specifications for the technologies. You know, today the technical specs for the new models, they're not mature. They are underway and they're being developed in the specs that they're trying to get ready to sample, but the models are already out there and into the wild. So it's really important to follow the design equity principles and to include Deaf and Hard of Hearing community in every part of every step of the design process. Second, we have been engaging with the community
and the public and the deaf community through webinars, symposia, and workshops. And they have provided effective learning and users' wants and needs and concerns. So we call on the Access Board to do similar efforts. It's a great start. Third and finally, we're here to make one very important point. There is a disparity between access to technology between Deaf and Hard of Hearing and the population at large. We need to work together to make sure that these amazing new technologies are evenly distributed. Thank you for listening. And Katie Kale: Thank
you, Jeff. And thank you, AnneMarie. Okay. Now we are going to hear from our final presenter, Melanie Fontes Rainer, who is the Director, the Office of Civil Rights, Health and Human Services on access to healthcare and agency services for the public. Director Fontes Rainer, you may begin. >> Melanie Fontes Rainer: Sure. And thanks for having me today. My name is Melanie Fontes Rainer. My pronouns are she/her/alla.
I'm a mixed race Mexican American with light skin. My hair is black and pulled back in a bun. I have on a white jacket and bright red lipstick today. So for everyone, we put up the slide, maybe we can hold off if putting it up in a second. We wanted to talk a little bit about some of the things that our office has been working on. So we are the Civil Rights office for healthcare and human services, which means it is our job to advance, promulgate, work on, effectuate, implement, and enforcing Civil Rights. As part of that, we do that through rule making,
policy guidance and law enforcement. So we have a unique role in the space and we've done a lot of work already that we're excited to talk about today. So one of those roles is section 504 of the Rehabilitation Act of 1973. This is a law that prohibits discrimination on the basis of disability. When the programs are funded by HHS, so the Health and Human Services agency that is our national office, or when they're conducted by those agencies. So recently,
we updated our rules when it's for programs that are funded or activities that are funded by this department. And through that rule making, we actually took on this issue. So in that final rule, tools that used patient information like age or health condition and other factors to estimate the benefits of specific healthcare, also known as value assessment tools, often involve algorithms. Those are not be used to discriminate on the basis of disability. So what does that seen in so in healthcare, we know these types of tools, value assessment tools. They're used to advise or determine the value, the dollar value of a particular health service, for example, such as a drug, a device, or a surgery. And the outputs of these tools often inform care decisions, such as whether your health insurance company will cover it in the first instance, whether a state Medicaid agency will cover that service, under what circumstances might they cover that, meaning they'll pay for that service, and they might be used to contain costs, so to try to make costs lower, and they might be used for quality improvement efforts, but we also know from our experience that sometimes they might lead to discriminatory outcomes against people with disabilities when they place a lower value on the life of a person with a disability compared to the life of a person that does not have a disability.
So we know, for example, we've seen this before. It's not new. Right? So we saw this during Covid 19, for example, when different treatments were denied to people with disabilities, and we know, for example, if two patients have nearly identical characteristics, but one has Alzheimer's disease, a value assessment tool may assign a lower value to that person's overall health because of that illnesses, and they might have a lower quality of life in such a tool. And so we know that a provider that might use that tool, then, could use that tool to deny that person a ventilator or some other treatment that would violate the law and is discriminatory. So in short, this Rule would prohibit that, which we think is really important, because again, while artificial intelligence, clinical algorithms, whatever you want to call it, these things are new, but also, they're not new in healthcare. We've been seeing algorithms,
value assessments being used for a long time in healthcare. And in fact, my office has done enforcement work and we know that they can discriminatory on the basis of race, age, disability. Right? And we know that we, as human beings, we're intersectional. We're not just a woman. We're also a woman who is Mexican American. We're not just a person with a disability. We're a person with a disability who has gender identity or pregnancy status, et cetera, et cetera. And so in addition to that rule, we also extend that Rule and make very clear that it applies in the child welfare context. This is important, because again, we know, we've seen algorithms be used in the Human Services space to deny parents with disabilities placement with foster kids, to deny kids placement with their own parents because of a disability, and that's why this rule, which has both this value assessment tool and this child welfare provision, is so important in disability.
I'm glad that this rule has gone forward. We're continuing to implement it. But also, it's really important, because while AI is new, a lot of these tools, predictive analytics, they have been used in the healthcare space to make benefits decisions, Social Services decisions for a long time. The other Rule my staff worked on that is really important in this space is section 1557 of the Affordable Care Act. Section 1557 is part of the Affordable Care Act. It's a provision, it's a Civil Rights provision, and it literally says nondiscrimination in health programs and activities, and that applies across race, color, national origin, sex, and age and disability. And so unlike section 504, which is just focused on disability, 5057 has all of these other provisions and accounts for this intersectionality between these protected classes. In addition, 1557 gives us jurisdiction
over insurance, which is something that 504 does not, which we all know algorithms sometimes can be used for prior authorizations, medical management techniques, ways in which you mate have restrictions in how you access prescription drugs, healthcare services, and otherwise. So under this new rule, those recipients of HHS funding, so Health and Human Services funding, may not or must not discriminate against an individual through the use of a patient care decision support tool. That includes artificial intelligence tools, clinical algorithms, flowcharts, eligibility tools, risk prediction tools, value assessment tools, and more.
And to clarify, we know clinical algorithms are like a step by step guide that healthcare professionals use to make decisions about patient care, and those steps are guided by specific symptoms, specific diagnostic information about that patient, that person, test results, other medical information inputs to the algorithm. And that then those algorithms determine, again, we talk about whether something is covered, whether a particular treatment is needed or should be different, and again, so we know that those things don't always reflect the conversations with provide ends, and we know that they oftentimes may have input processes that might treat someone differently because of a disability. So we think that there is a promise in the use of these tools to reduce health disparities and increase access to healthcare, but where he also want to make sure that we also want to make sure that providers of healthcare use these tools responsibly so that they can be ethical and make sure they're not driving more harm and making sure that they're not inadvertently discriminating through bias. So through this rule, tools that have an elevated risk of discrimination, which might be a tool that has a lot of variables with age, origin, sex, or disability, that
2024-08-12 06:05
