Blind and low-vision people have experienced remarkable gains in information literacy because of digital technologies, like being able to access an online library offering more than 1.2 million books that can be translated into text-to-speech or digital Braille. But it can be a lot harder to come by an accessible map of a neighborhood they want to visit, or any simple diagram, due to limited availability of tactile graphics equipment, design inaccessibility, and publishing practices.
(You can also find this episode on the Internet Archive and on YouTube.)
Chancey Fleet wants a technological future that’s more organically attuned to people’s needs, which requires including people with disabilities in every step of the development and deployment process. She speaks with EFF’s Cindy Cohn and Jason Kelley about building an internet that’s just and useful for all, and why this must include giving blind and low-vision people the discretion to decide when and how to engage artificial intelligence tools to solve accessibility problems and surmount barriers.
In this episode you’ll learn about:
- The importance of creating an internet that’s not text-only, but that incorporates tactile images and other technology to give everyone a richer, more fulfilling experience.
- Why AI-powered visual description apps still need human auditing.
- How inclusiveness in tech development is always a work in progress.
- Why we must prepare people with the self-confidence, literacy, and low-tech skills they need to get everything they can out of even the most optimally designed technology.
- Making it easier for everyone to travel the two-way street between enjoyment and productivity online.
Chancey Fleet’s writing, organizing and advocacy explores how cloud-connected accessibility tools benefit and harm, empower and expose communities of disability. She is the Assistive Technology Coordinator at the New York Public Library’s Andrew Heiskell Braille and Talking Book Library, where she founded and maintains the Dimensions Project, a free open lab for the exploration and creation of accessible images, models and data representations through tactile graphics, 3D models and nonvisual approaches to coding, CAD and “visual” arts. She is a former fellow and current affiliate-in-residence at Data & Society; she is president of the National Federation of the Blind’s Assistive Technology Trainers Division; and she was recognized as a 2017 Library Journal Mover and Shaker.
Resources:
- Bookshare.org
- Itsjustabitofpaper.co.uk: “Origami for the Blind and Visually Impaired”
- Access Origami
- American Printing House for the Blind: Mastering the Monarch
- Orbit Research: Graphiti—A Breakthrough in Non-Visual Access to All Forms of Graphical Information
- Data & Society: Dark Patterns in Accessibility Tech with Chancey Fleet (June 2019)
- Data & Society: Community and Accessibility Online with Chancey Fleet and Taeyoon Choi (April 2020)
What do you think of “How to Fix the Internet?” Share your feedback here.
Transcript
CHANCEY FLEET
The fact is, as I see it, that if you are presented with what seems on a quick read, like good enough alt text, you're unlikely to do much labor to make it better, more nuanced, or more complete. What I've already noticed is blind people in droves dumping their descriptions of personal images, sentimental images, generated by AI onto social media, and there is a certain hyper-normative quality to the language. Any scene that contains a child or a dog is heartwarming. Any sunset or sunrise is vibrant. Anything with a couch and a lamp is calm or cozy. Idiosyncrasies are left by the wayside.
Unflattering little aspects of an image are often unremarked upon, and I feel like I'm being served some Ikea pressboard of reality, and it is so much better than anything that we've had before on demand without having to involve a sighted human being. And it's good enough to mail, kind of like a Hallmark card, but do I want the totality of digital description online to slide into this hyper normative, serene anodyne description? I do not. I think that we need to do something about it.
CINDY COHN
That's Chancey Fleet describing one of the problems that has arisen as AI is increasingly used in assistive technologies.
I’m Cindy Cohn, the executive director of the Electronic Frontier Foundation.
JASON KELLEY
And I’m Jason Kelley, EFF’s Activism Director. This is our podcast, How to Fix the Internet.
CINDY COHN
On this show, we’re trying to fix the internet – or at least trying to envision what the world could look like if we start to get things right online. At EFF we spend a lot of time pointing out the way things could go wrong – and jumping in to the fight when they DO go wrong. But this show is about optimism, hope and bright ideas for the future.
According to a National Health Interview Survey from 2018, more than 32 million Americans reported that they had vision loss, including blindness. And as our population continues to age, this number only increases. And a big part of fixing the internet means fixing it so that it works properly for everyone who needs and wants to use it – blind, sighted, and everyone in between.
JASON KELLEY
Our guest today is Chancey Fleet. She is the Assistive Technology Coordinator for the New York Public Library, where she teaches people how to use assistive technology to make their lives easier and more accessible. She’s also the president of the Assistive Technology Trainer’s Division for the National Federation of the Blind.
CINDY COHN
We started our conversation as we often do – by asking Chancey what the world could be like if we started getting it right for blind and low vision people.
CHANCEY FLEET
The unifying feature of rightness for blind and low vision folks is that we encounter a digital commons that plays to our strengths, and that means that it's easy for us to find information that we can access and understand. That might mean that web content always has semantic structure that includes things like headings for navigation.
But it also includes things that we don't have much of right now, like a non-visual way to access maps and diagrams and images, because of course, the internet hasn't been in text only mode for the rest of us for a really long time.
I think getting the internet right also means that we're able to find each other and build community because we're a really low incidence disability. So odds are your colleague, your neighbor, your family members aren't blind or low-vision, and so we really have to learn and produce knowledge and circulate knowledge with each other. And when the internet gets it right, that's something that's easy for us to do.
CINDY COHN
I think that's so right. And it's honestly consistent with, I think, what every community wants, right? I mean, the Internet's highest and best use is to connect us to the people we wanna be connected to. And the way that it works best is if the people who are the users of it, the people who are relying on it have, not just a voice, but a role in how this works.
I've heard you talk about that in the context of what you call ‘ghostwritten code.’ Do you wanna explain what that is? Am I right? I think that's one of the things that has concerned you.
CHANCEY FLEET
Yeah, you are right. A lot of people who work in design and development are used to thinking of blind and disabled people in terms of user stories and personas, and they may know on paper what the web content accessibility guidelines, for instance, say that a blind or low vision user or a keyboard-only user, or a switch user needs. The problems crop up when they interpret the concrete aspects of those guidelines without having a lived experience that leads them to understand usability in the real world.
I can give you one example. A few years ago, Google rolled out a transcribe feature within Google Translate, which I was personally super excited about. And by the way, I'm a refreshable Braille user, which means I use a Braille display with my iPhone. And if you were running VoiceOver, the screen reader for iPhone, when you launched the transcribed feature, it actually scolded you that it would not proceed, that it would not transcribe, until you plugged in headphones because well-meaning developers and designers thought, well, VoiceOver users have phones that talk, and if those phones are talking, it's going to ruin the transcription, so we'll just prevent that from happening. They didn't know about me. They didn't know about refreshable Braille users or users that might have another way to use VoiceOver that didn't involve speech out loud.
And so that, I guess you could call it a bug, I would call it a service denial, was around for a few weeks until our community communicated back about it, and if there had been blind people in the room or Braille users in the room, that would've never happened.
JASON KELLEY
I think this will be really interesting and useful for the designers at EFF who think a lot in user personas and also about accessibility. And I think just hearing what happens when you get it wrong and how simple the mistake can be is really useful I think for folks to think about inclusion and also just how essential it is to make sure there's more in-depth testing and personas as you're saying.
I wanna talk a little bit about the variety of things you brought up in your opening salvo, which I think we're gonna cover a lot of. But one of the points you mentioned was, or maybe you didn't say it this way in the opening, but you've written about it, and talked about it, which is tactile graphics and something that's called the problem of image poverty online.
And that basically, as you mentioned, the internet is a primarily text-based experience for blind and low-vision users. But there are these tools that, in a better future, will be more accessible, both available and usable and effective. And I wonder if you could talk about some of those tools like tablets and 3D printers and things like that.
CHANCEY FLEET
So it's wild to me the way that our access to information as blind folks has evolved given the tools that we've had. So, since the eighties or nineties we've had Braille embossers that are also capable of creating tactile graphics, which is a fancy way to say raise drawings.
A graphics-capable embosser can emboss up to a hundred dots per inch. So if you look at it. Visually, it's a bit pixelated, but it approaches the limits of tactile perception. And in this way, we can experience media that includes maybe braille in the form of labels, but also different line types, dotted lines, dashed lines, textured infills.
Tactile design is a little bit different from visual design because our perceptual acuity is lower. It's good to scale things up. And it's good to declutter items. We may separate layers of information out to separate graphics. If Braille were print, it would be a thirty-six point font, so we use abbreviations liberally when we need to squeeze some braille onto an image.
And of course, we can't use color to communicate anything semantic. So when the idea of a red line or a blue line goes away we start thinking about a solid line versus a dashed or dotted line. When we think about a pie chart, we think about maybe textures or labels in place of colors. But what's interesting to me is that although tactile graphics equipment has been on the market since at least the eighties, probably someone will come along and correct me that it's even sooner than that.
Most of that equipment is on the wrong side of an institutional locked door, so it belongs to a disability services office in a university. It belongs to the makers of standardized tests. It belongs to publishers. I've often heard my library patrons say something along the lines of, oh yeah, there was a graphics embosser in my school, but I never got to touch it, I never got to use it.
Sometimes the software that's used to produce tactile graphics is, in itself, inaccessible. And so I think blind people have experienced pretty remarkable gains in general in regard to our information literacy because of digital technologies and the internet. For example, I can go to Bookshare.org, which is an online library for people with print disabilities and have my choice of a million books right now.
And those can automatically be translated to text-to-speech or to digital braille. But if I want a map of the neighborhood that I'm going to visit tomorrow, or if I want a glimpse of how electoral races play out, that can be really hard to come by. And I think it is a combination of the limited availability of tactile graphics equipment, inaccessibility of design and publishing practices for tactile graphics, and then this sort of vicious circular lack of demand that happens when people don't have access.
When I ask most blind people, they'll say that they've maybe encountered two or three tactile graphics in the past year, maybe less. Um, a lot of us got more than that during our K-12 instruction. But what I find, at least for myself, is that when tactile graphics are so strongly associated with standardized testing and homework and never associated with my own curiosity or fun or playfulness or exploration, for a long time, that actually dampened down my desire to experience tactile graphics.
And so most of us would say probably, if I can be so bold as to think that I speak for the community for a second, most of us would say that yes, we have the right to an accessible web. Yes, we have the right to digital text. I think far fewer of us are comfortable saying, or understand the power of saying we also have a right to images and so in the best possible version of the internet that I imagine we have three things. We have tactile graphics equipment that is bought more frequently, and so there are economies of scale and the prices come down. We have tactile design and graphics design programs that are more accessible than what's on the market right now. And critically, we have enough access to tactile graphics online that people can find the kind of information that engages and compels them. And within 10 years or so, people are saying, we don't live in a text-only world, images aren't inherently visual, they are spacial, and we have a right to them.
JASON KELLEY
I read a piece that you had written about the kind of importance of data visualizations during the pandemic and how important it was for that sort of flatten the curve graph to be able to be seen or, or touched in this case, um, by as many people as possible. But, and, and that really struck me, but I also love this idea that we shouldn't have to get these tools only because they're necessary, but also because people deserve to be able to enjoy the experience of the internet.
CHANCEY FLEET
Right, and you never know when enjoyment is going to lead to something productive or when something productive you're doing spins out into enjoyment. Somebody sent me a book of tactile origami diagrams. It's a four volume book with maybe 40 models in it, and I've been working through them all. I can do almost all of them now, and it's really hard as a blind person to go online and find origami instructions that make any sense from an accessibility perspective.
There is a wonderful website called AccessOrigami.com. Lindy Vandermeer out of South Africa does great descriptive origami instruction. So it's all text directing you step by step by step. But the thing is, I'm a spatial thinker. I'm what you might think of as a visual thinker, and so I can get more out of a diagram that's showing me where to flip dot A to dot B, then I can in reading three paragraphs. It's faster, it's more fluid, it's more fun. And so I treasure this book and unfortunately every other blind person I show it to also treasures it and can't have it 'cause I've got one copy. And I just imagine a world in which, when there's a diagram on screen, we can use some kind of process to re-render it in a more optimal format for tactile exploration. That might mean AI or machine learning, and we can talk a little bit about that later. But a lot of what we learn about. What we're good at, what we enjoy, want, what we want more of in life. You know, we do find online these days, and I want to be able to dive into those moments of curiosity and interest without having to first engineer a seven step plan to get access to whatever it is that's on my screen.
JASON KELLEY
Let’s pause for just a moment to say thank you to our sponsor. “How to Fix the Internet” is supported by The Alfred P. Sloan Foundation’s Program in Public Understanding of Science and Technology. Enriching people’s lives through a keener appreciation of our increasingly technological world and portraying the complex humanity of scientists, engineers, and mathematicians.
And now back to our conversation with Chancey Fleet.
CINDY COHN
So let's talk a little bit about AI and I'd love to hear your perspective on where AI is gonna be helpful and where we ought to be cautious.
CHANCEY FLEET
So if you are blind and reasonably online and you have a smartphone and you're somebody that's comfortable enough with your smartphone that like you download apps on a discretionary basis, there's a good chance that you've heard of a new feature in this app, be my eyes called be my AI, and it's a ChatGPT with computer vision powered describer.
You aim your camera at something, wait a few seconds, and a fairly rich description comes back. It's more detailed and nuanced than anything that AI or machine learning has delivered before, and so it strikes a lot of us as transformational and or uncanny, and it allows us to grab glimpses of what I would call a hypothesized visual world because as we all know, these AI make up stories out of whole cloth and include details that aren't there, and skip details that to the average human observer would be obviously relevant. So I can know that the description I'm getting is probably not prioritized and detailed in quite the same way that a human describer would approach it.
So what's interesting to me is that, since interconnected blind folks have such a dense social graph, we are all sort of diving into this together and advising each other on what's going well and what's not. And I think that a lot of us are deriving authentic value from this experience as bounded by caveats as it is. At the same time, I fear that when this technology scales, which it will, if other forces don't counteract it, it may become a convincing enough business case that organizations and institutions can skip. Human authoring of alt text to describe images online and substitute these rich seeming descriptions that are generated by an AI, and even if that's done in such a way that a human auditor can go in and make changes.
The fact is, as I see it, that if you are presented with. What seems on a quick read, like good enough alt text, you're unlikely to do much labor to make it better, more nuanced, or more complete.
CINDY COHN
I think what I hear in the answer is it can be an augment to the humans doing the describing, um, but not a replacement for, and that's where the, you know, but it's cheaper part comes in. Right. And I think keeping our North Star on the, you know, using these systems in ways that assist people rather than replace people is coming up over and over again in the conversations around AI, and I'm hearing it in what you're saying as well.
CHANCEY FLEET
Absolutely, and let me say as a positive it is both my due diligence as an educator and my personal joy to experiment with moments where AI technologies can make it easier for me to find information or learn things. For example, if I wanna get a quick visual description of the Bluebird trains that the MTA used to run, that's a question that I might ask AI.
I never would've bothered a human being with it. It was not central enough. But if I'm reading something and I want a quick visual description to fill it in, I'll do that.
I also really love using AI tools to look up questions about different artistic or architectural styles, or even questions about code.
I'm studying Python right now because when I go to look for information online on these subjects, often I'm finding websites that are riddled with. Lack of semantic structure that have graphics that are totally unlabeled, that have carousels, that are hard for screen reader users to navigate. And so one really powerful and compelling thing that current Conversational AI offers is that it lives in a text box and it won't violate the conventions of a chat by throwing a bunch of unwanted visual or structural clutter my way.
And when I just want an answer and I'm willing to grant myself that I'm going to have to live with the consequences of trusting that answer, or do some lateral reference, do some double checking, it can be worth my while. And in the best possible world moving forward, I'd like us to be able to harness that efficiency and that facility that conversational AI has for avoiding the hyper visual in a way that empowers us, but doesn't foreclose opportunities to find things out in other ways.
CINDY COHN
As you're describing it, I'm envisioning, you know, my drunk friend, right? They might do okay telling me stuff, but I wouldn't rely on them for stuff that really matters.
CHANCEY FLEET
Exactly.
CINDY COHN
You've also talked a little bit about the role of data privacy and consent and the special concerns that blind people have around some of the technologies that are offered to them. But making sure that consent is real. I'd love for you to talk a little bit about that.
CHANCEY FLEET
When AI is deployed on the server side to fix accessibility problems in lieu of baking, accessibility in from the ground up in a website or an application, that does a couple of things. It avoids changing the culture at the company, the customer company itself, around accessibility. It also involves an ongoing cost and technology debt to the overlay company that an organization is using and it builds in the need for ongoing supervision of the AI. So in a lot of ways, I think that that's not optimal. What I think is optimal is for developers and designers, perhaps, to use AI tools to flag issues in need of human remediation, and to use AI tools for education to speed up their immersion into accessibility and usability concepts.
You know, AI can be used to make short work of things that used to take a little bit more time. When it comes to deploying AI tools to solve accessibility problems, I think that that is a suite of tools that is best left to the discretion of the user. So we can decide, on the user side, for example, when to turn on a browser extension that tries to make those remediations. Because when they're made for us at scale, that doesn't happen with our consent and it can have a lot of collateral impacts that organizations might not expect.
JASON KELLEY
The points you're making about being involved in different parts of the process. Right. It's clear that people that use these tools or that, that actually these tools are designed for should be able to decide when to deploy them.
And it's also clear that they should be more involved, as you've mentioned a few times, in the creation. And I wanted to talk a little bit about that idea of inclusion because it's sort of how we get to a place where consent is actually, truly given.
And it's also how we get to a place where these tools that are created do what they're supposed to do, and the companies that you're describing, um, build the, the web, the way that it should be built so that people can can access it.
We have to have inclusion in every step of the process to get to that place where these, all of these tools and the web and, and everything we're talking about actually works for everyone. Is inclusion sort of across the spectrum a solution that you see as well?
CHANCEY FLEET
I would say that inclusion is never a solution because inclusion is a practice and a process. It's something that's never done. It's never achieved, and it's never comprehensive and perfect.
What I see as my role as an educator, when it comes to inclusion, is meeting people where they are trying to raise awareness – among library patrons and everyone else – I serve about what technologies are available and the costs and benefits of each, and helping people road map a path from their goals and their intentions to achieving the things that they want to do.
And so I think of inclusion as sort of a guiding frame and a constant set of questions that I ask myself about what I'm noticing, what I may not be noticing, what I might be missing, who's coming in, for example, for tech lessons, versus who we're not reaching. And how the goals of the people I serve might differ from my goals for them.
And it's all kind of a spider web of things that add up to inclusion as far as I'm concerned.
CINDY COHN
I like that framing of inclusion as kind of a process rather than an end state. And I think that framing is good because I think it really moves away from the checkbox kind of approach to things like, you know, did we get the disabled person in the room? Check!
Everybody has different goals and different things that work for them and there isn't just one box that can be checked for a lot of these kinds of things.
CHANCEY FLEET
Blind library patrons and blind people in general are as diverse as any library patrons or people in general. And that impacts our literacy levels. It impacts our thoughts and the thoughts of our loved ones about disability. It impacts our educational attainment, and especially for those of us who lose our vision later in life, it impacts how we interact with systems and services.
I would venture to say that at this time in the U.S, if you lose your vision as an adult, or if you grow up blind in a school system, the quality of literacy and travel and independent living instruction you receive is heavily dependent on the quality of the systems and infrastructure around you, who you know, and who you know who is primed to be a disability advocate or a mentor.
And I see such different outcomes when it comes to technology based on those things. And so we can't talk about a best possible world in the technology sphere without also imagining a world that prepares people with the self-confidence, the literacy skills, and the supports for developing low tech skills that are necessary to get everything that one can get out of even the most optimally designed technology.
A step by step app for walking directions can be as perfect as it gets. But if the person that you are equipping with that app is afraid to step out of their front door and start moving their cane back and forth and listening to the traffic and trusting their reflexes and their instincts because they have been taught how to trust those things, the app won't be used and there'll be people who are unreached and so technology can only succeed to the extent that the people using it are set up to succeed. And I think that that is where a lot of our toughest work resides.
CINDY COHN
We're trying to fix the internet here, but the internet rests on the rest of the world. And if the rest of the world isn't setting people up for success, technology can't swoop in and solve a lot of these problems.
It needs to rest upon a solid foundation. I think that's just a wonderful place to close because all of us sit on top of what John Perry Barlow called meatspace, right, and if meatspace isn't serving us, then the digital world can only, you know, it can't solve for the problems that are not digital.
JASON KELLEY
I would have loved to talk to Chancey for another hour. That was fantastic.
CINDY COHN
Yeah, that was a really fun conversation. And I have to say, I just love the idea of the internet going tactile, right? That right now it's all very visual, and that we have the technology to make it tactile so that maps and other things that are, you know, pretty hard for people with low vision or blindness to navigate now, but we have technology, some of the, tools that she talked about that really could make the internet something you could feel as well as see?
JASON KELLEY
Yeah, I didn't know before talking to her that these tools even existed. And when you hear about it, you're like, oh, of course they do. But it was clear, uh, It was clear from what she said that a lot of people don't have access to them. The tools are relatively new and they need to be spread out more. But when that happens, hopefully that does happen, it sort of then requires us to rethink how the internet is built in some ways in terms of the hierarchy of text and what kinds of graphics exist and protocols for converting that information into tactile experiences for people.
CINDY COHN
Yeah, I think so. And it does sit upon something that she mentioned. I mean, she said these machines exist and have existed for a long time, but they're mainly in libraries or other places where people can't use them in their everyday lives. And, and I think, you know, one of the things that we ended with in the conversation was really important, which is, you know, we're all sitting upon a society that doesn't make a lot of these tools as widely available as they need to.
And, you know, the good news in that is that the hard problem has been solved, which is how do you build a machine like this? The problem that we ought to be able to address as a society is how do we make it available much more broadly? I use this quote a lot, but you know, the future is here. It's just not evenly distributed. Seemed really, really clear in the way that she talked about these tools that like most blind people have used once or twice in school, but then don't get to use and turn part of their everyday life
JASON KELLEY
Yeah. The, the way I heard this was that we have this problem solved sort of at an institutional level where you can access these tools at an institution, but not at the individual level. And it's really. It is helpful to hear and and optimistic to hear that they will exist in theory in people's homes if we can just get that to happen. And I think what was really rare for this conversation is that it, like you said, we actually do have the technology to do these things a lot of times we're talking about what we need to improve or change about the technology and and how that technology doesn't quite exist or will always be problematic and in this case, sure, the technology can always get better, but it sounds like we're actually At a point where we have a lot of the problems solved, whether it's using tactile tablets or, um, creating ways for people to use technology to guide each other through places, whether that's through like a person, through Be My Eyes or even in some cases an AI with the Be My AI version of that.
But we just haven't gotten to the point where those things work for everyone. And everyone has a level of technological proficiency that lets them use those things. And that's something that clearly we'll need to work on in the future.
CINDY COHN
Yeah, but she also pointed out the work that needs to be done about making sure that we're continuing to build the tech that actually serves this community. And she, you know, and they're talking about, you know, ghostwritten code and things like that, where, you know, people who don't have the experience are writing things and building things based upon what they think the people who are blind might want. So, you know, on the one hand, there's good news because a lot of really good technology already exists, but I think she also didn't let us off the hook as a society about something that we, we see all across the board, which is, you know, it need, we need to have the direct input of the people who are going to be using the tools in the building of the tools, lest we end up on a whole other path with things that other than what people actually need. And, you know, this is one of the kind of old, you know, what did they say? The lessons will be repeated until they are learned. This is one of those things where over and over again, we find that the need for people who are building technologies to not just talk to the people who are going to be using them, but really embed those people in the development is one of the ways we stay true to our, to our goal, which is to build stuff that will actually be useful to people.
JASON KELLEY
Thanks for joining us for this episode of How to Fix the Internet.
If you have feedback, we'd love to hear from you. Visit EFF.org/podcast and click on listener feedback. While you're there, you can become a member, donate, maybe pick up some limited edition merch like tshirts or buttons or stickers and just see what's happening in digital rights this week and every week.
This podcast is licensed Creative Commons Attribution 4. 0 International and includes music licensed Creative Commons Attribution 3.0 unported by their creators. In this episode, you heard Probably Shouldn't by J.Lang, commonGround by airtone and Klaus by Skill_Borrower
Our theme music is by Nat Keefe of BeatMower with Reed Mathis
And How to Fix the Internet is supported by the Alfred P. Sloan Foundation's program in public understanding of science and technology.
We’ll see you next time.
I’m Jason Kelley…
CINDY COHN
And I’m Cindy Cohn.
