The Outer Circle

Ben Tillyer – University of Oxford – Reflections on WCAG, AI, and Inclusion

• Silktide • Season 1 • Episode 1

Share episode

Ben Tillyer, Head of Digital Accessibility at Oxford, talks with Jessica about WCAG’s limits, the messy ethics of unpaid testing by disabled staff, and whether AI could ever help fix it. 

From alt text to overlays to Oxford’s own machine learning research, this episode pulls no punches—and yes, there’s a cat.

[Jessica] Welcome to the Outer Circle, a podcast about digital accessibility, what it is, why it matters, and the people working to make the web a better place for everyone. I'm Jessica Chambers, accessibility evangelist at Silktide, and each episode, I sit down with someone in the field to hear their insights, experiences, and challenges. Let’s see where today takes us. And today I'm joined by Ben Tillyer, the University of Oxford’s head of digital accessibility. Ben has been working in digital accessibility for over a decade, and is now part of the advisory committee at the W3C. They're the people who make WCAG. We'll be talking about his first eight months in post, other bits and pieces, and what he's learned over the last ten years.[Ben] I remember the first time I presented for CSUN. It was during the pandemic and everything had to be recorded remotely, of course. So I think I ended up doing eight takes of my talk before they got one that they accepted, so... it's fine.[Jessica] I had never given an in-person presentation,[Ben] Okay.[Jessica] when I went to CSUN, and I was all like... Hi. Y’know, I like... And I was wearing these overalls and the, the microphone, it just kept rubbing against me and like, making noise. And I was like, just give me a microphone or something. Trying to hit the buttons on the laptop, trying to hold the microphone. And I was like... It was just honestly a bit silly. The second time around, I was like, I bought a clicker, right? And I was like, this is fab. I'm so smart. I don't know what I did with it. I went to the first presentation. Smooth as silk. Perfectly great. Go to the next presentation like two days later, don't have the clicker or can't find the clicker. And God bless him, my CEO actually was like, I will be your human clicker.[Ben] Amazing.[Jessica] And like, click the button. Which made it hilarious when he wasn't listening because it'd be like,“Next slide. NEXT SLIDE!” It's like the audience is laughing. The whole thing was just really kind of... I think it added something instead of taking it away. So, Ben. The University of Oxford, I'm just going to say it. The University of Oxford! How did you get here? [Ben] Yeah exactly. How did I get here? That's a really good question. And it was just a colleague of mine, someone from the W3C that is on the same working group as me said, just in case you're not happy with your current job or you're looking for something else, something exciting, what about this? And, yeah, like you said, the University of Oxford is exactly what my first thought was as well. So that's how I found out about it. And I didn't think for a second that I’d progress past the first round of interviews, because I have no university degree. I'm a university dropout. I attended university in Sheffield, but I didn't make it through for a variety of reasons that we won't go into here, but I thought that a prestigious university like Oxford would require someone to have a degree. But, I’m sat before you now. So, I must have done something right during the interview process.[Jessica] They did seem happy to have you when I was there. Like, honestly, the University of Oxford is the kind of the pinnacle of education. So, as you said, I am also a university dropout. I’m a high school dropout, too. So it's kind of amazing because that means that they're looking for talent, right? They're not setting a barrier by saying, if you don't have a university degree, you can't work here. [Ben] Exactly. And I think that's an accessibility accommodation in itself. The fact that if you look 10, 20 years ago, a lot of the jobs would have said at many places the seniority, like a degree, a an undergraduate degree or a master's degree or even higher, or certain qualifications. But now we're seeing more and more a degree or equivalent experience. And I think that has just made the job market easier for people. I mean, I'm sure you know that people with disabilities tend to drop out of higher education more than the general population on average. So the fact that they're doing equivalents and allowing people to just showcase their talent rather than qualifications opens the job market up to a whole different sector of people. I do believe that degrees have their place, but they're not for everyone, and they certainly weren't for me.[Jessica] Okay, let's see. What was your a-ha moment that started you down the accessibility path?[Ben] That's a good question. So I don't think there was an a-ha moment. There wasn't a light bulb that suddenly switched on and made me think that I really needed to work in this space, but it was... It was just the right place, right time situation. So when I was in my first job, which was an insurance call center, I was asked to sit down next to a gentleman called Chris, who was completely blind. Well, he could see... he could detect where light sources were. So he had some concept of light and dark, but apart from that, there was no useful vision other than maybe seeing where the windows are. And if the light is switched on or not. So they sat me next to Chris because they knew how I had an interest in computers and coding and programing. I've been coding for a while since I was a child, from Windows 3.1 I think I wrote my first Visual Basic app on, and they knew that I was always making macros in Excel and things to speed the job up. So they sat me down with him because he always had so many technical problems, his setup. He was using the same sort of computer as everyone else. It was a Windows PC, but he had a screen reader called Jaws installed on his machine that he used to read aloud and navigate through the insurance systems that we used for our job. So when customers rang us, instead of seeing the insurance claim information on the screen, he would press a series of commands and it would read out what was on the screen and what he needed to know at that time. And I never heard of screen readers or assistive technology before. I was in my early 20s at the time, and come to think of it, I don't think I'd had a conversation longer than a few sentences with someone with a visible disability, either. So... I'll try and get to the a-ha moment, right. So— although the thing that is closest to an a-ha moment. I realized that the screen reader would only work if the insurance system was coded in a way that the screen reader was expecting, because he had all of these different screens and different paths that he went through on a day to day basis. But as soon as new screens were added, new processes were added that looked and felt different to me, using the system visually, it also looked and felt different to Jaws and sadly, it just fell down in some circumstances and if a system falls down when I was using it, I would just, you know, refresh the page until it looked right, or click around and try and investigate, to try and force my way through this process. But for his software, you have to be so precise that he couldn't do workarounds to get through. There weren’t any workarounds. It all relied on the author of the software to get it right for him, and it just made me see... The more software that I saw him use, the more I saw that the software developers either didn't know or didn't care about assistive technology and the importance of semantics and good naming. So maybe it is not a light bulb, but it just... just made me feel a bit like warm helping him and doing that. It seemed like the right thing to do, and I just so happened to have an opportunity to, have a full time job, for a company called Blazie Engineering, who are digital accessibility consultants and, Jaws specialist. Jaws scriptors. They've been doing this since since I, since I was the I don't know, they started in the 90s. So they've been doing it a long time. So that's how I started in it. And maybe, yeah, just the feeling of empathy, that was the closest thing to an a-ha moment.[Jessica] Let's talk about alt text. So part of the reason I bring this up is that, there doesn't seem to be a universal set of opinions, shall we say. There doesn't seem to even be a best practice. And even here in the UK, the BBC and Gov. UK disagree, you can line it up and they just have completely conflicting rules for how to deal with alt text. So one of the things I want to do is talk to people about what they think about alt text and how they approach dealing with it, like, what defines a decorative image?[Ben] I mean, you’re starting at the top, right? It’s WCAG 1.1.1. Are we going to go for all of these? All of the criteria? We might need to extend this podcast— But yeah, it's a good place to start. Because yeah, everyone has an opinion and I definitely have an opinion on this. So, decorative. Let's start with that. Yeah I think that's a good point. Decorative means so many different things to different people. And I think it should, because if you're in the marketing department and you're tasked with making your home page look and feel a certain way that reflects your brand, you are putting so much thought into the so-called decorative images which don't convey meaning functionally to the user, but are there to evoke a kind of emotion or feeling in the user. But if you were going to do a strict down the middle, decorative or functional or informative, there isn't anything there for the tone of voice or feeling or emotion. So I think this comes down to the HTML specification. And as well as spending so much time reading over WCAG, I've sadly spent so much time reading over the HTML5 specification and you either have alt text or you don't have alt text. There is no attribute on an image to say, is this decorative? Is this functional? Is this for emotion? And I think that's a problem. So I think the way we solve this is to petition the people who write HTML to add some more attributes to that effect. And then you can provide different alt text for different circumstances. Maybe you want an alt text to say exactly what the contents of the image are. There's a family having a picnic on a hill. There's a bottle of orange juice, you know, next to the strawberries. That is going to be interesting and important to some people that want that information. But if people want to know that it's designed to look modest and happy and trustworthy, I think the authors of the webpage should be able to tell the user that through some sort of metadata, whereas you need to give users the choice to ignore it. And if it is like on a banking platform, for example, and is an image just next to the enter your username and password field that has no relation to it, give the users the choice. If they don't want to hear the exact content, they don't want to hear the feeling, just let them ignore it. So hopefully that summarizes my thoughts on alt text. What do you think? [Jessica] I think it does. But I'm actually going to follow up with you on this. I actually love this idea. I think we should talk to people who make HTML5 about giving people some options. You obviously then need to talk to Freedom Scientific and, NVDA I guess is open source, so maybe we can impact that. We got to get all the screen readers to support the new spec, right? When you talk about emotions, that's something I hear a lot. Is that like the image is chosen to evoke a feeling. Do you think it's better to literally like, say, this is supposed to build trust or make you feel happy, you know, or, do you think we should be looking at how we write the alt text, like the... when we go for a very literal, this is what's in the picture. There's, you know, a red headed woman with a black cat, right? That is literally what's in the image, right? But it's different if you use more colorful language, shall we say. Do you think that's like a middle ground?[Ben] Well, I think the alt text is one piece of the puzzle, because if you have a different tone of voice in your alt text to the rest of the visible text on the screen, it's going to be really strange if you have a, if you have a really like, I'm thinking like a scientific journal that's very matter of fact. And and then you have the most— [Jessica] Flowery.[Ben] Yes. I was trying to avoid saying the word flowery. I couldn’t think, so I chose silence instead. Yeah. So if you have the most— and I was also trying to avoid black and white —if you have the most functional scientific text and then the most convoluted, that's a good one, the most convoluted text in your alt text, it might be disconcerting to the end user.[Jessica] It can be quite jarring.[Ben] So... there's two things here, I think. One is that you should ask the users. User research is key here. If you find fifty of your customers or 100 of your customers that make use of alt text in some way, and it doesn't have to be screen reader users. Alt text is made use of by people who have very low bandwidth and disable images coming through and being downloaded to their machine. I mean, they talk about the cost of data being almost free in the UK now, every major internet provider kind of has an option for unlimited data for less than 100 pounds a month. But if you look at the global south, data is extremely expensive. So your choices for alt text not only impact the blind and visually impaired, but some other groups as well that you need to factor in. And so, find some users who make use of alt text and ask them. If all of them agree on whether to include a certain tone of voice or a certain level to include or exclude, fantastic. But I don't think that'll be the case. All the users that I've ever spoken to or the like, consumers of alt text, that I've spoken to, they all have different thoughts. They all have different needs. It depends on whether they're in a rush. It depends on their mood that day. If, you know, there's some days that I really enjoy seeing adverts on pages, but there's some times that I just try and disable them all and that's a choice I have. I have a Chrome extension that gets rid of ads and colors and all sorts to lower my levels of distraction, so maybe the choice should be up to the user in terms of personalization.[Jessica] One of the things I hear a lot is, bad alt text is better than no alt text. What's your take on that?[Ben] Well, as long as it's bad alt text and not wrong alt text I think is a key one.[Jessica] That's an important... [Ben] Yeah. So you mentioned the lady with the black hat. If you had a lady with a black cat on your alt text, but it was actually, a picture of a tree, you know, maybe it would be better to have no alt text at all in that situation. It's, I think going through this podcast, I think I could answer every question with, it depends. Which is, you know, the accessibility professional's ultimate answer is, it depends. So... I'm just thinking more on this thing about functional versus emotional versus decorative. You struggle to get some content authors to put one piece of alt text in. So if we then change the HTML spec to say, oh, you need four. You need its functional use, its emotional use, its decorative use, and another one, we're never going to get there, are we? We're going to have three sites that will meet this level of compliance. And no one else is going to bother, sadly.[Jessica] Okay, I totally agree with you, but I still think that the concept is worth exploring. One of the things I think and I look forward to is that when we set an image as decorative, it can create FOMO. You know, the fear of missing out. Because so many people who use screen readers have some vision. So they know there's an image there and you're not telling them what it is, and it makes them frustrated, which I think is actually valid. Right? Yeah. I would be like, dude, I can see there's an image. You know, what is in it? But more of them, like VoiceOver has this option, you can ask it to give you an AI description of what's in it. And those tend to be quite literal. Right. So it's like the, the literal descriptions are definitely coming. And I think that should be an option though, I think anyone who wants to know what's in an image that someone hasn’t made alt text for, we need to find a way to use AI to bring that in, because that kind of hits that, Is it better to have bad alt text or no alt text? Well, like if you remove the, it's just wrong, then like, you know. One of my favorite examples is a picture of the Queen at somewhere in what looks like Parliament. It's not. It's the House of Lords, right? You know this because of the color of the seat she's in. And I was like, really? But the AI was so certain. It was like, this is the Queen at the House of Parliament. And I was like, what? And I don't think you're supposed to have to be an expert in these things, right? You know what I mean? Like, I just thought the color looked weird, so I looked it up right. There's like, you know, also, apparently the queen isn't allowed into the House of Commons, so it's just like, why would you say that? But it delivers it with such a authority that you can't really be sure. And I think we're getting better. But I do think it's still a little bit of a risk, but I think often about how many different ways there are to approach alt text and how many different opinions there are on what's decorative and what isn't.[Ben] I think the bottom line is, if you are asked for an opinion, for your company, if you're asked to write some guidance on alt text, think of the users and think of the spirit of these guides. So you mentioned GDS and BBC and their guidelines. They both have the spirit of alt text like baked into them. I know they’re different and I don't necessarily agree with either of them, but that's okay. It's okay to have your own thoughts on this and look to see what works best for your business or your University or your government department, whatever. Whatever works for you. Just yeah, get some data. Don't rely on your thoughts and intuition. Speak to people. Speak to developers. Designers. Speak to your marketing team and speak to people with disabilities and speak to people without disabilities as well. It might be that your website's too bloated with images and nobody likes them. Never mind talking about alt text. You might not need some of these images in the first place. Yeah, data. Data speaks.[Jessica] What's your top tip for someone who's new to accessibility?[Ben] So what I always say when I have a first conversation with someone who is wanting to learn more about accessibility, is that there is no such thing as a stupid question when it comes to digital accessibility.

If you don't know what 4.5:

1 means in color ratio, or you don't know what screen reader is, or you can't work out how someone that's hearing impaired can make use of Netflix, just ask. You can Google it. You can ask one of your large language models and hope it's right. Or you can talk to me. Or talk to someone in the accessibility community. The vast majority, if not all of us, are here to make the digital world work for people with disabilities. And the only way we're going to do that is by spreading knowledge throughout everyone who touches the digital world, whether that's a user or whether it's a designer, developer, quality assurance tester, whatever it is. So yeah, just ask. Don't be afraid to ask and don't stop asking. As you said in the introduction, I have been doing this over ten years and there are still things that I learn virtually every day. Yeah, if not every day, you know, a few times a week. Because there is so much to learn. So, so don't feel, don't feel bad if you don't know something. It's not your fault.[Jessica] Every day is a school day.[Ben] Every day is a school day. But they don't teach accessibility at schools for some reason.[Jessica] Do you have a favorite website?[Ben] My favorite accessibility related website is probably HeydonWorks.com. It is a blog and has a nice little online shop as well, from an accessibility professional. I will make sure you have the link, Jessica. I don't know if you've read some of the articles on that. You probably have, but might not realize that's what website you are on. But it's an amazing take on digital accessibility and just development and user experience in general. So that would probably be my top website in terms of accessibility.[Jessica] Do you have any pets?[Ben] I do, and it's completely on brand. I have a visually impaired cat. So his name is Denny. He's a black, British Shorthair, and we adopted him during the pandemic. I think it was March 2021. So it's coming up to four years of ownership now. He's amazing. He spends most of the time that I'm here in the office, sat in the office with me. So he, he keeps me company, and he is just... He's the best thing that happened during lockdown.[Jessica] What's one common accessibility mistake you find?[Ben] There are just so many. Sadly, there are so many common accessibility mistakes. The ones that I find the most frustrating are the ones that the developers will have been told about, either, when they're writing their code and compiling it, and they have a linting— that I'm assuming most developers using React have a linting tool. And there's some really basic HTML issues that just cause accessibility problems. Things like using the wrong ARIA. Making ARIA values and attributes up is something that I don't quite understand how it happens. The other is color contrast. I find the most websites I go to just for my own pleasure have color contrast issues all over the place, and all you need to do to discover these is run Google Lighthouse. If you've never used Google Lighthouse before, it's built right into the browser. If you open Developer Tools in the settings menu, there's a tab called Lighthouse, and if you use two clicks, you click Lighthouse, you click run and it throws back a score from 0 to 100. And most sites that you go on will find an automated result in less than 10 seconds for color contrast. And that's even before you get to more advanced tools, things like Silktide’s scanning software. This is something that's been delivered to you on your PC already. So, that's what I would probably want people to fix. Just things that are found within 10 seconds of running a test.[Jessica] It is a bit painful when you're like, any automated checker could have told you this was a problem. Even Figma has color contrast checking built in.[Ben] Well there you go.[Jessica] The design tools are picking it up, which is kind of cool.

But my rule is simple:

squint. If everything disappears, your color contrast is bad.[Ben] So I had a little foray into non-digital accessibility, a few years ago, and I helped my wife, my wife was a museum curator, and she asked me for some guidance on color contrast for the printed museum board, things like the text for each of the images in the gallery or displays. I looked online, and I was surprised that there weren't many pieces of support out there. There was one from the Smithsonian. The Smithsonian has an amazing accessibility guide. If you type in Smithsonian Accessibility Guide into your search engine of choice, I'm sure it will come up. But there wasn't anything from any of the UK galleries which I thought was odd. So I made one. And rather than, squinting, or taking my glasses off, what I said was, design it how you want it to look and then print it the size you want it to look, but then look at it from four times the distance that you are planning your guests to read it from. And if you can still read it, fantastic. That will cover a lot of the audience. I'm not saying it's going to be perfect, but pragmatism over idealism is always key for me. So four times the distance, can you still read it? And if the answer is no, go back to the drawing board to make things higher contrast or a bigger font size. So at one point during during lockdown, Sophie brought home massive like A1 and A2 printed displays that were going to go in the museum, and stood at the other side of the garden from me, and I was like, can I still read this? And we got there. So yeah, there's so much to think about in terms of when you print digital data. Yeah. Context. Context is everything.[Jessica] How do you approach making forms accessible? The dreaded form.[Ben] Yeah. It's something so simple, right? A form, it's one of the first things you're taught to do when you do any development course. Computer science, go on, make this form, putting something in, and press Enter. And like, I think that was the first PHP project I did was just a login form.[Jessica] Yep.[Ben] So, yeah. How do you make them accessible? Well, I personally suggest not including any optional fields. If you need the data, make it mandatory. And if you don't need the data, don't include it on the form. That will make it less taxing, less cognitive load for people, and then make it easier to complete and faster to complete. I also think that Gov.UK does this quite well. Sometimes I think they go a little overboard. Sorry to GDS, but when you have a piece of information, such as if you are needing a driver's license number and you've got a form asking for that, tell your users where you will find that piece of information, because the user doesn't know the data as well as you know the data. So say it's number... Is it number nine on the back of your driving license? Or it's on the paper about two thirds of the way down if you still have the the paper copy. Or you can find it if you log on to X, Y, Z. And pre-populate things as well. I'm sure when— There’s so many times I've completed forms where I'm pretty certain that the company that I'm filling this form out for already has that data from me, and they're asking from me as a user again. So don't collect too much data that you don't need. Explain to users where to find information and pre-populate things as much as you can that removes even further questions. I think there's some really obvious ones. I have a favorite WCAG criterion 4.1.2 Name, Role, Value. And I think that is the key to understanding how to make something accessible for screen readers at the very least. And automated tools can really help you work out 4.1.2 compliance. If you make sure that every form field in your form has a correct name, a correct role, and allows the user to put the value in that they want, you're on to a winner, there. That kind of leads me on to another thought. One of my bugbears when completing a form, let's say you are completing a post code. So an example, OX2 2AA I think is one of our office buildings. And if I put that in without a space, the form says, oh, I'm sorry, we can’t accept that, it doesn't have a space in it, and there's some logic going on in the back there where it's saying, this doesn't have a space, but otherwise it's the right length. If the form can use a service like the Royal Mail postcode checker, like as an API behind the scenes to detect that if the website added that space for you and then saved it and it corresponded with the first line of address that you've put in the form, accept it. Let the user put in the data in a way that works for them, and then you do the heavy lifting, the processing. Take cognitive load, take that burden away from them.[Jessica] That was a really good point.[Ben] Thanks. Phone numbers as well. Sorry, it needs to start with +44? Add +44 to my number! You told me the answer![Jessica] And the phone numbers. What's one myth about digital accessibility that you'd like to debunk?[Ben] Oh, can I have two? [Jessica] Yeah. Why not.[Ben] Awesome. So the first one is that WCAG compliance doesn't equal accessible. You can have a completely accessible site for the vast majority of your user base, including those with disabilities, and not come close to meeting WCAG. And the same is true as well; you can meet WCAG and not be accessible. You have to consider the usability of your site as well. The second one is that you don't have any disabled customers. So this is something that I've heard multiple times. Oh, I don't need to be WCAG compliant because nobody using my app or website has a disability. With the greatest of respect, Mr. Product Manager, you don't know. And even if you did, even if you looked at your entire user base and you asked them personally whether they had a disability or not, there are two things. Firstly, they might not want to tell you they have a disability. And so that means any figures that you have collected in your organization or customer base may be and probably will be wrong. The numbers will be less than the actual number. And secondly, that might be the case today, but it might not be the case later this afternoon, tomorrow, next week. The vast majority of disabilities are acquired during working ages between 18 and, you know, 65. Yeah, the vast majority, I can't remember the stat, something like 70-80% of all disabilities are acquired during working age. So your customer base might not have a disability today, but, they might have a disability tomorrow. Yeah. The working age thing is really...[Jessica] It's really important. The other thing I always think about is the Click-Away survey. Where like, the fundamental reality is, is that as gross as it sounds to say this, if you're someone who uses a screen reader and you find a website that works, you know, even if you say, hey, you can get your groceries cheaper over here, if they've got one that works, they're not going to give you another chance, right? You might have been cheaper, but if you weren't functional, then you said you don't care about them, and their needs. So like when people tell me I don't have any disabled customers, I don't need to care about screen readers. I'm like, well, I can guarantee you don't now. And you never know. [Ben] That's a really good point. You might not have disabled customers because your product is inaccessible.[Jessica] Yeah. So you're alienating a surprising amount of the population, actually.[Ben] Exactly. And and I'll let you into a little bit of a secret here. People with disabilities have money to spend, and they might want to spend it on your digital e-commerce platform, you know, but if it's inaccessible, they can't. And that's why you might not have customers with disabilities. So maybe there is a business case for making accessible.[Jessica] What are your thoughts on the impact AI could have on digital accessibility? I'm going to make a note. While I say this question, it is currently February of 2025. So if you're watching this in 2027 and they've actually got, you know, like the general AI that we're all waiting for, you need to know it was 2025. Okay. We didn't have that yet. [Ben] Yes. Very good point. So as of February 2025, I don't know the impact AI is going to have on digital accessibility, but I know that whatever happens, it's not going to be completely good. So I know there are some positive improvements. I've seen new products come out, a product called Otiter. It was released onto the Chrome Web Store just this week, which includes testing where, after the accessibility automated tests are happened, I think it uses AI in large language models to help interpret the results. That's my understanding of it. Again, sorry, David, if that's completely, mischaracterizing your product. And I also know that accessibility consultancies are trying to add, like, ChatGPT access into screen readers. So I think it's really exciting that that's happening. But yeah, I am worried about the negative impacts as well. So I'm guilty of using Google Gemini and Microsoft Copilot to help me write code. I am a hobbyist Next.js developer in my spare time. And sometimes I don't know the answer because I haven't had any formal training on Next.js, and I make use of large language models to write my code for me, and I do a cursory check of it. And one of the things I check the code for is the accessibility of it. And most of the time the large language models have not provided me with WCAG-compliant code. I have to go in and make changes myself to the code and then remember next time to add, and make sure it is WCAG 2.2 compliant into my prompt that I'm feeding into the, large language model. And I'm pretty confident to say that the majority of hobbyist developers like me taking shortcuts with LLMs won't be adding that to their prompts. If they're not familiar with accessibility already. So it worries me that people are using AI for shortcuts and, forgetting or they're at risk of forgetting about accessibility when they give the heavy lifting to these incredible tools.[Jessica] They are incredible. But I would not feel guilty that you were leaning on it. The reality is, like there are stats out today, that like 80% of coding problems are being solved by AI right? Today, like, right now in February. Right?[Ben] Absolutely. So I'm on, I'm on a beta version of Gemini, I think it's called Flash 2? I think it's called and it helped me write a Next.js app in three days, and I've got a working version like an MVP of this project.[Jessica] When we say MVP, we don't mean Most Valuable Player. We mean Minimum Viable Product, as in just enough that it does what you need it to.[Ben] And I think if I was doing this without Gemini, would have taken me weeks of study into APIs, into, like, headless browsers, all sorts. And I didn't need to do that. And I it worries me that when people are going through education now, if they're not learning about digital accessibility at university today, if universities are showing them how to use LLMs to write code for them, which I think they should, you just hope that it isn't forgotten about. Like the whole joy of writing semantic code and proper code isn't isn't lost because, because someone else is writing the code for you.[Jessica] Me personally, I worry that what we're going to experience is that it's not the shortcut so much as that people won't care, right? They're going to say, oh, well, I'll just have this thing build it for me. And they won't even learn about accessibility. Because most people don't even realize it's a thing. Yeah. And if we don't have things that do understand it, and I'm just going to say this out loud, using DeepSeek really kind of tweaked my brain because it understands WCAG and like, digital accessibility, in a way that ChatGPT just doesn't. Maybe the newer versions? Because like, I mostly hang out in 4o, because I want to talk to it. But if I'm left with the situation where, like, ChatGPT will actually explain it improperly. And be like, this is what this guideline is. And I'm like, no, it isn't. You know. And I find myself thinking, this is so dangerous because anyone who didn't know WCAG well, you know, would have just accepted that answer, right? It sounded good. [Ben] Exactly.[Jessica] Whereas like, DeepSeek is busy being like, you know, oh, let's get into the nitty gritty. And, I called it a WCAG zealot the other day, because it was prepared to throw down about whether or not we could use emojis in a CAPTCHA and whether that would be acceptable. And I was like, I don't think it is. And it was like, oh no, it is. It's like just backed itself up with like, see? It's covered in like four different success criteria. And I was just like, who are you? Right? But I've never had a conversation with ChatGPT where I felt like it actually knew it well enough to fight with me. Right? And I was like, here we go. But as you said, like, if you don't tell it: I need this to be accessible, you've got a problem, because, like, we built Fake University, by hand, before AI. Right. And because they built it in a framework, that I'm blanking on. But because we use the framework— Bootstrap, we used Bootstrap to make Fake University, which meant that breaking the website was actually this huge undertaking because it had put accessibility into it. And we had to go and like, rip it back out. And I was so excited then because I thought, if they're making it so you can just not fall foul of certain rules, then there's a way to go forward where you can make the framework where people won't be able to make as many mistakes, and it's like it just fizzled out, right. It got so far and then they just stopped. And you know, if you tell someone like, you know, here, this thing is just going to solve all your problems, that's what I'm worried about is that they'll be like, I don't need to understand this. And I don't even mean like understanding WCAG. I mean, understanding how people interact with the web so that you can at least have a better understanding of, like, what's out there. You know, it might be people, you know, right? Yeah. They're just not talking about it. [Ben] Yeah. And I think this comes down to the fact that large language models are biased.[Jessica] Yeah.[Ben] If you look at the training, like all large language models are fed training data and the data that is available on the web is not often accessible. I think if you looked at the WebAIM survey, the, 1 million websites that they look at every year. So if you look at their survey from 2024, 95.9% of the million home pages they looked at had accessibility failures that could be automatically detected on them. So, I mean, there's not much data there for these large language models to look at and correctly piece together what an accessible website must be, must look like, because all of their training data is wrong. So unless the developers of these models are feeding in very specific information to ignore the vast amount of incorrect HTML markup, and then just look at the nought point something percent of correct websites, which they're not doing. Yeah, that's that's a problem. Bias is a problem. [Jessica] Yeah. And to me, I’m like how do you get around that when, I'm just going to say it, accessibility is quite gray. It's like well you could do it this way or you could do it this way, or you could do it this way. And each of those is valid. So like when you're the large language model and they go, I need to make a directory of all the staff at the company, what's it going to tell you the best answer is? You know. And like that's the thing. It's a conversation I had yesterday where I'm like, well, you could do it this way or... It depends. Right? I just found myself being like, if I— I should have asked it just to see what it said—but like, it's the same question of like, you know, what kind of information am I going to get back? Like, if we can't even come to consensus opinions ourselves, what do we think the LLM is going to do based on what it can read, that we've all written where we all disagree, right? Talking about WCAG, if you could add something to WCAG, that's the Web Content Accessibility Guidelines, what would it be?[Ben] I think it would be a bit of leniency because at the moment you either pass or you fail, and it doesn't take into account the amount of good work you and your company may have done to get towards compliance. There is, in terms of what it says in the spec, there is no distinction in a conformance report between one defect and 5000 defects. If you fail on one page with one defect, even if it's single-A, you can't claim to be compliant and I think that puts some people off aiming for it because they do so much, they spend so much time, money, effort into making something as good as it can be. And then they get an audit from, an accessibility consultancy and they say, sorry, you are not compliant in terms of WCAG. And that is disheartening. The W3C is working on the next iteration. It's a complete rewrite of WCAG 3, it used to be called Project Silver, and that is something that, they are taking into account, but it's not going to be released anytime soon. So, I wish— I wasn't around when, 2.0 was being, thought up. I wasn't in the industry then or I wasn't, you know, ready to contribute to the WCAG. And I think I was just getting to grips with how Jaws worked. And so that's what I would add, I would add some sort of recognition for people trying to do the right thing, because that's what's missed, I think a lot of conversations that, between an accessibility consultant and the rest of the business, the IT team in charge of delivering an IT project, are so negative, they are often brought in to go through a list of defects and prioritize them and triage them. And yes, that's important. But as an— I think as an industry, we need to be more positive. And it doesn't help that our holy book, our guidelines doesn't allow for that at the moment. So that's what I would want to add.[Jessica] I think that's really interesting. It is kind of a purity test, isn't it? You know, you're doing it or you're not.

And we don't really allow for:

look how far you've come. Progress over perfection...[Ben] Exactly, progress over perfection. Yeah. Continuous improvement is really important in accessibility as it is with with every other industry.[Jessica] And I think you bring up an interesting point. It can be a quite adversarial relationship when you're like, you need to do these things. And they're like, I don't have time, I don't have resource, I don't have... reason, you know? So we need to find a way to bring them with us rather than alienating them [Ben] 100%. I've heard of many organizations who, just stopped bringing in accessibility consultants because they were always bad news, and they wanted some good news, but that's probably their choice of supplier rather than anything else. But...[Jessica] I'd like to believe that I am super friendly.[Ben] So am I, so am I, Jessica.[Jessica] Okay. On a similar note, if you could remove something from WCAG, what would it be?[Ben] So the one thing that I would remove are the levels, single-A, double-A, triple-A is what I mean by levels. I don't think they're helpful when project managers or business analysts, whoever's in charge of triaging results says, oh, let's go for the single-As first. Then we'll attack the double-As, then we'll attack the triple-As... It works sometimes, but sometimes there are really important things that are found in a triple-A or double-A criterion that you should be fixing so far before the single-A, some of your single-A results. A single-A result might only impact one person, but a triple-A result might impact 20 people. You know, it could be 20 times more important to fix. What people are often shocked at is that there's not really a definition for what these levels are, anyway. If you look at, the guidelines, if you do the full one page version and you, you read it top to bottom, or ask a large language model to find the definition in there for you, there isn't one. I was lucky enough to speak to one of the people who helped write WCAG 1 and WCAG 2.0, and he said that double-A was the things that they couldn't find consensus to fit in single-A, and triple-A were the things that they couldn't find a consensus to fit in double-A. And I'm just thinking to myself like, that just goes around in circles. What are you doing? But they— I mean, credit to them. Credit to the people who who put WCAG 1 and 2.0 together. They did a lot of hard work and a lot of good research, and they came up with something that, look, it might not be perfect, but it works and it's been adopted as law into so many like countries and territories now. But that was one thing that I just think maybe they should have left out. So yeah, sorry. Sorry guys and gals.[Jessica] Thank you for that. That is a radical approach. But if that's how they got to A, double-A and triple-A, so much makes so much more sense now where I'm like, why is this triple-A? Why did you just remove this entirely? There's one in 2.2 that they've removed about like getting the page numbers to be universal no matter what format you were using, which I think is really important. And I was actually really excited about and it was listed originally as triple-A, which made sense to me because... At Silktide, we say A is must do, double-A is should do, and triple-A is aspirational reaching for the stars stuff. Right. So I kind of felt like triple-A is where it belonged, because maybe technology isn't there yet, right? But if you don't put it in the list, no one is going to talk about it. They're not looking at whether or not they can make your Kindle go, this is the same page as the printed version, so that when a professor says, read pages 78 to 84, or look at the diagram on page 72, like you could actually know which one they mean. And if you need larger print fonts, or you need it to be in an e-reader. So you could make the whole thing substantially bigger, right? It shouldn't impact your ability to still get the same information as your classmates, but I feel like it's it's one of those things where I hadn't even thought about it until I read it in the 2.2 draft, and I thought, oh, that. That's really good. Like, we should totally do that. And I was like, man, that's going to be a wild tech thing though, isn't it? And then they just removed it entirely and I was really sad and did a bunch of webinars where I talked about 2.2 and I would bring it up because even though they removed it, I felt we needed to talk about it because I wanted them to be able to know that it was something people were considering. Right.[Ben] Absolutely.[Jessica] It's one of those ones where like, it actually just made me really frustrated. Because I was like, why didn't you just leave it as triple-A? No one, no one in the world makes triple-A a requirement, right? No government thinks that's a requirement. And there's a reason, you know, but I feel like— How do you get the conversation started? How do we even get the companies who make e-readers and who set the specs, like, you know, get them to a point where they're considering that, so in the following years, maybe they find a way to do it, you know? But if no one's talking about it, then I don't see how they'll even know about it. I didn't.[Ben] Did you know that before it was triple-A, it was single-A?[Jessica] That hurts me in a new, new way.[Ben] Absolutely, absolutely.[Jessica] What changed my opinion on WCAG permanently? It was what happened with Focus Appearance, where it was like, you're going to do it. Except maybe not.[Ben] And I was wearing two hats at the time, right. So at the time I was working at TPGI— it used to be called the Paciello Group—and we took a lot of time to review the first working draft, public working draft, that W3C released back in 2021. It's something I wrote back in June. I collected all the feedback from the engineers at TPG. And if you if you scroll down to see Page Break Navigation, it was 2.4.13 and it was Single-A, at the time. So we were generally supportive of it. When I say we, I'm not with them many more, but TPGI as a company was very supportive of the page break navigation. Had a few queries... But I had another hat. So I was representing TPG on the Accessibility Guidelines Working Group at the same time. So I was kind of reviewing it and writing at the same time, which, which gave me a fun insight. And what I found was how respectful the working group are to the concerns and criticisms of work that the public have. You don't have to be an accessibility specialist. You can be just a consumer, right? You can be someone with a disability who relies on this, who doesn't necessarily know the ins and outs of, you know, digital accessibility. Or you could just be a developer or project manager, you know, of a big corporation who you're going to have to make a million websites. Work with this new criterion, and you have to consider disproportionate burden as well as the needs of people with disabilities. And it's balancing that, in some international guidelines, is really important. So that's— I have huge respect for the chairs of AG, the Accessibility Guidelines Working Group. And I don't envy what they do whatsoever. So, yeah, just a little insider look into this. There's a lot that goes on that's not just looking out for the— Obviously, the digital requirements of people with disabilities is at the heart of what they do and what they, as chairs, look at. But there's so much more, and it's a very difficult job. And that's why it takes so long for these new versions of the guidelines to come out.[Jessica] That it does. When you're like, eight years for 2.1, huh?[Ben] Yeah, absolutely. Another myth, digital accessibility myth. Just because someone has a disability doesn't make them a digital accessibility expert. They might understand how their lived experience affects the way they use the web. It may give them an insight into how they best use their assistive technology and their screen. But just because there's someone in your business who has a disability, they might not know everything to do with digital accessibility. And even if they do, they might not want to be the digital accessibility specialist in your organization just because they have a disability. So just bear that in mind.[Jessica] That is why we say you should do automated testing, manual testing, and then user testing.[Ben] Exactly.[Jessica] Like I'm very big on like, task based testing where it's like go to the website and try to do this thing that we would want people to come to the website to do.[Ben] Yeah. 100%.[Jessica] Because as you said, like it can be difficult. Also as you said, they might just enjoy their job, right? They’re with you. They're good at it. Don't rock the boat. [Ben] 100%. In a previous job, there was a business analyst. So they were employed as a business analyst, and they were really, really good at what they did. They happened to be blind as well. And because of that, they were bombarded with emails from people around the organization asking them to test something for accessibility because they were blind and because they could do it in their 9 to 5 and not— it wouldn't be a cost. They wouldn't have to go to, an accessibility consultancy. It's what someone thought because they just, you know, they asked the the business analyst who was blind. That might be okay, some of the times. But just like, like you said, don't rock the boat. They were not hired as an accessibility tester. They might have been. If you have a blind, quality assurance professional, fantastic. If they're not, don't rely on that. Yeah. Don't rely on them. Let them do their job.[Jessica] Remember, we're all people.[Ben] Correct.[Jessica] What are your thoughts on overlays? Dun-dun-dunnnn![Ben] So for those who don't know, a gentleman named Karl Groves. He made a talk on Tenon. He put together a website called OverlayFactSheet.com. If you haven't seen that, please read it. I was the 51st person to sign the overlay fact sheet, and, I haven't removed my name from that. I believe that everything it says on it is correct. And what it goes into, well, it goes into the pros and cons of overlays and automated accessibility repair and remediation and how these tools are negatively impacting people with disability when they use the web. What I find really good about the overlay fact sheet is, one, that it’s open source. So if there's any corrections that need to be made or any more news articles or testimonials from people with disabilities need to be added, anyone can add it. There's a, you know, it needs to be accepted by a certain number of people, but it's not going to be, you know, filtered, if it's correct. And I won't spoil too much about it, but it goes yeah, it has a conclusion after speaking to so many people in the community and so many people that make these tools. I think that these tools were created with the best of intentions. But maybe these companies bit off more than they could chew, to use, you know, to use a common phrase, I don't know if that's a British phrase— They took on too big a task. And I think these companies that had masses of funding from venture capital firms to try and achieve this, they tried to... they tried to do too big of a job. And I think that at the start, some of them had marketing that might not have been completely, 100% correct when they were selling their tools and widgets to the public and to big businesses. Their best of intentions came across as being quite malicious, and I think that they have got better now. Some of the big tools have joined organizations like the W3C, and I have been to talks that have been very well moderated, very constructive between overlay technologists and people from the accessibility community and the standards community. They've started to come together to find some middle ground, and they're working together to work out how their technology and their expertise can best help people with disabilities and large corporations, where it is way too complex and way too time consuming to do a manual review on, you know, their entire digital state. There are not enough hours in the day to do, you know, a manual review of every single web page that Amazon, for example, has. Sorry, Amazon. I think your site is great. It's no criticism, but just an example. It's a huge site, like tens of millions, hundreds of millions of web pages. You can't automate all of that. Sorry. You have to automate all of that. You can't manually test all of that. So maybe just do more research into tools that you are going to use to do that automated testing, rather than necessarily believing these firms that say, with one line of code, all of your issues can be solved, because I don't think that in the foreseeable future, unless AI, you know, grows a conscience and, consciousness, we're not going to get there. So I tried to be as diplomatic as I could with that answer. I am not a fan, in summary, of overlays as they currently stand, but I think they have a place in the future. Just, they're not ready yet. As of February 2025.[Jessica] You may note we are both not using names. Naming names when it comes to these particular software companies, they are notoriously litigious. There's something called a strategic lawsuit called a SLAPP suit, which essentially tries to get you to stop talking and shut you up. This has been applied a number of times, and since it is public, I can say that AudioEye, they sued Adrian Roselli. Fingers crossed, we'll get him on here. But the fundamentals is, this is a man who's never done anything, like, untoward. He just started blogging about his experiences with their software, and they wanted him to stop. And they tied him up with a legal battle. And no one knows how much that cost. But an undisclosed amount of money at the end was donated to a charity of his choice. But I don't have the resources to fight them. So when I went to CSUN and put on a presentation called One Line of Code Won't Fix Your Website— you can watch it on YouTube—I didn't name a single name because it's too dangerous. That said, the one thing I will say to everyone is if you're going to use any kind of overlay and I mean like the little box that just has some toggles that'll like, make the colors different or enlarge your text, right? There's some really big ones, popular ones that do this. Please test it. You are responsible for what you put on your website and what that experience is for users. And if you have put a widget on your website that enlarges the font, and what it does is cause the text to overlap or run off the sides so you can't even get to it, that's actually really bad, and there's no guarantees that it will work the way you expect. So please just test it before you push it out to the public. I always wished for the librarian from Snow Crash. And now I can actually sit and talk to ChatGPT. Is there tech you've been waiting for?[Ben] I have not read Snow Crash. I know about it. I know, I'm sorry. However, one thing I do know about the author, Neil Stephenson is that he is credited in some ways of coining the term metaverse. And I think that blows my mind because Snow Crush was, what, mid 90s when it was released? So the fact that his view of what the future was going to be has basically come to fruition is incredible. And I just wish he'd have written a book on what he thought the accessible web would look like, but that's, you know, for another day. What tech am I waiting for? Oh, I would cherish a day that I don't need to use a computer anymore. You know, whatever comes next, after the web and the internet, the internet and computers, bring it on. That's what I want. I don't know what that is or what it looks like. And I'm not going to coin a phrase, but, yeah, whatever comes next, please get me away from sitting at my desk.[Jessica] You need to read Snow Crash. The gargoyles, like, I knew I would be one. They implant everything they can. Because that way they're just walking tech hubs. They don't need to have computers or whatever because they're just built into their bodies. But in Snow Crash, because they start early, they're like, quite noticeable and bulky and the whole nine so like, they're frowned on. But let's face it, some of us would be that person anyway.[Ben] So there's a lot going on in the world of politics with Elon Musk. However, putting that to one side for a second, Neuralink, his company that is pioneering brain computer interfaces, is really exciting to me. It's not exciting what they're doing with animal testing. I don't agree with it. I think it's cruel. However, that aside, if they can do it without animal testing, if they can come up with a brain computer interface that allows humanity to not risk of their life getting it implanted, and then have access to accurate data would push society so far ahead. If that came to fruition. I know that I think like one human has had it successfully implanted now, and it's scary, but I think that's the future. I think, brain computer interfaces would get me away from my screen, would get me away from my speakers that I can talk to. It would get me off of my phone, and I think it would allow humanity to get back to, you know, the outside world and, put an end to the 9 to 5 office grind. So that's what I'm excited about.[Jessica] I am really big into, you know, like wearable tech and implantable tech, right? You can tell I read Snow Crash when I was— It was 1992, by the way, for Snow Crash. Snow Crash, 1992. So I first read it in my teenage years. That's embarrassing. But, it really shaped my views on technology and where things might be going. However, and I can't stress this however enough, did you read about the bionic eyes?[Ben] No. Not recently.[Jessica] It was called the Argus 2. They implanted this thing into hundreds of people, and, literally what I called nightmare fuel was reading about the lady who was— I think she said she was in the subway —and there was a faint beeping, and her world went dark. Now, the thing is, the Argus 2, from what I can tell, was not giving— It wasn’t like they gave people their full sight back, right? But fundamentally, they had given people extra that they didn't have otherwise. And it helped them lead better and fuller lives. And because of money, they turned it off. And I have been thinking about this since 2022 when they turned it off, because it really upset me. Like I said, nightmare fuel. Flash forward to last November-ish. There was an AI toy for children called Moxie. He's like a little doll that talks to you. This thing was not cheap, for the record, like it was something like£800, you know, for Moxie, right? And, while I think they made just a huge, huge mistake by not doing some sort of subscription model to continue funding because it's running off AI, that's what Moxie is, Moxie was particularly good for autistic children. She brought them out of their shells. She was a safe person to talk to. And there's plenty of videos of parents talking about the impact it had on their children. And they shut down Moxie because they didn't have any money. They actually released the code so that you could potentially host it yourself, because they wanted to not have to sever the relationship they made with lots of children. And in their defense, they made a really lovely letter that made me cry. You know, it was like from Moxie to all the kids. It scares me that we're in in these kind of situations, you know, like, I would have been, like, totally prepared to sign up for brain implants. I’m all over that. But this idea that they can just go, well, we can't support this anymore because it's not financially feasible. That's... that's scary. Like, I'm not sure what we do about people who are— Literally have obsolete tech embedded in their bodies or in their brains. And yeah, Neuralink, watching someone write their name with a robotic arm they are controlling with their brain, giving someone that piece of their life back is... It's that, sufficiently advanced technology is indistinguishable from magic.[Ben] Exactly. And the reason Neuralink sprung to my mind is because of the positive impact it will have on people with disabilities more than the general public, because you... It's the same brain, right? If you have a physical disability, you have the same brain as the rest of the population. And if it will allow you to like literally open doors that you were not able to open before, you know, it will provide you with so much more opportunity. And yeah, it will positively impact people with disabilities more. But yeah, the risks are huge. And it's the same with more common assistive technologies as well. So I think of things like, like braille displays and screen readers, you have to keep updating them. Otherwise you lose functionality. And that is sad because of the rate of change in the online world and the new the new frameworks, the new standards that come out to better technology for the world, when those are released, it means that all of the assistive technology has to update too, and that takes time and costs money, and they're going to have to put that on their customers. Which is why companies like Freedom Scientific charge like annual fees or upgrade fees, if you need to do it because the time and money isn't free. But yeah, that's the society we're in, which we need to do something about. Jessica. We need to do something about.[Jessica] Freedom Scientific totally knows my opinions on this, since I tracked down a bunch of them to complain, when I was in CSUN. In America— this is a detail I think everyone should actually know —in America, access to Jaws, which is literally one of the best screen reader software platforms in the world. Right. And it runs on Windows, which means if you don't have a lot of money, Windows is where you're going to be. You're not going to buy a Mac, right? But the problem for me is that in the US, it's $100 a year. It's software as a service, right? You pay your hundred bucks a year and you're good to go. And while that's not free, it's pretty cheap, right? The company that licenses it and sells it in the UK charges £1,000 for a license, and it does not come with the yearly upgrades. Those cost extra money, it’s something like £250 when you need to upgrade. Because, as Ben points out, the software actually is being worked on by large teams who are making it better and are looking at the new things that are coming with HTML5 and how to interact with different things. And like... God bless them. It's amazing. Right? But depending on where you are in the world, it's either easier to access or it isn't. And particularly as accessibility professionals, it can be really difficult when you're like, you're going to need a suite of tools to do all the things you need to check. And, you know, honestly, like mad power to Apple, the guy who got it through eventually to make VoiceOver said he had to wait for people to die, because the old guard were just not interested in doing it. So he literally waited for people to retire and pass away so that they would be gone and stop blocking him. And now we have VoiceOver. And that's the thing, VoiceOver is astounding, you know, like this phone is, is like one of the most accessible things in the world. It's amazing. Who knew you would get to this place, right? But like, it's expensive, right?[Ben] Yeah, absolutely. So, yeah, pricing is is really interesting. I think the $100 US license is for non profit. It's for home use, for noncommercial use.[Jessica] Yes. It's for regular people. [Ben] Yeah. But again if you use it outside of work in the UK it’s £750 I think something like that. It's a lot of money and I do not have a Jaws license as an accessibility professional. I'd like a personal license. I have a professional license. I've always had a professional license for doing my job, but I would love have a home license. But I just can’t— I can't justify the cost. And I'm someone who is in full time employment. If you are a person with a disability, you are much less likely to be in full time employment in the UK. So you are much less likely to have an expendable income. When people talk about reasonable adjustments, it's normally reasonable adjustments in the workplace. So to to have, you know, the luxury of being offered a screen reader, Jaws, you need to be in a job first. And, yeah, I don't know, it's chicken and egg situation, isn't it? You need a— you might need a professional screen reader to get a job, but you can only afford it once you're in a job. And, yeah, it's a difficult situation.[Jessica] How do automated testing platforms like Silktide help organizations like the University of Oxford? So yeah. Great question. So the University of Oxford has a number of different content management systems that allow different departments, different academics, different students to make their own web pages. And what that means is we have a lot of web pages. There is a website crawling platform that looks at the number of web pages under the top, like 10,000 domains that exist in the world and it runs that every month and publishes the data freely. And on that site, we’re in the millions, of individual web pages that belong to the University of Oxford. And... [Jessica] Wow.[Ben] You cannot, just as I mentioned Amazon, just like other big corporations, it is extremely difficult to manually test all of that effort. Some accessibility consultancies charge £1,000 plus a day to do a manual audit of a website, and sometimes these companies only get through five web pages a day. So if you divide the millions of websites by £1,000 per five web pages a day, you can see how the cost adds up. And having to do that every month, every, you know, every six months, whatever your testing cadence is, gets very expensive very quickly. So balancing automated testing with manual testing is the only way that owners of large digital estates can can get this right. So I mean, Jessica knows that we we work together, right? We, we did a presentation to a select group of people at the university on what the state of the university's digital estate was like, and pockets of bad, pockets of good, and bits of in between is exactly what I think we expected, right? So we had the data to to back that up, thanks to Silktide. So I think what the tool did for us was show us the categories of what was wrong. So rather than having to attack each issue on each page one by one, we were able to see like the macro level, we were able to see what the most common issues were. We were able to see whether they were caused by the same line of code on different pages, and then that helped us work out whether it was a platform related bug or whether it was a bad piece of training, which ended up causing content authors to consistently make the same mistake across the different web pages. And we were only able to do that because we use Silktide. So yeah, that's what I've learned. I've learned that you can run a scan on a massive amount of web pages and find patterns. Then from the patterns, we were able to apply a fix once and see the benefit of that across so many sites, like uncountable numbers of sites, just because we fix something at the source, and yeah, that's not something that we could do alone. So, yeah. Thanks. Thanks Silktide.[Jessica] So, Ben, you've been the head of digital accessibility at Oxford for, like, eight or so months now. What's your biggest win?[Ben] So the biggest win is due in part to how engaged the university have been to digital accessibility. We have just started as a joint project between the accessibility team and the AI and Machine Learning Competency Center have teamed up to look at how machine learning can be used to solve some of the testing success criteria in WCAG that need to know the human context. So we've spoken a lot about alt text. Can AI and machine learning be used to solve that problem? So we are not looking to solve any particular issue, but we are trying to look at what AI might be able to solve. And I think the University of Oxford is one of the few places in the country that has, you know, the bandwidth, the budget, the expertise, the knowledge, the enthusiasm for learning and pushing the frontiers of technology to to try and work that out. So I think... I know I haven't got anything to deliver and show it to you yet, but I feel that it's a conversation that I've just not had anywhere else I have worked over the past ten years, and I think just having that conversation with people outside of the accessibility space is a huge win in itself. And not only that, the biggest part of the win is that we were approached by them. They came to us to say that it's something that they wanted to explore. So I think the win was being in the right place at the right time and being able to be part of those initial conversations. Whether they'll go anywhere and how long this will take, I can't say, but that's... It's exciting, and it's just something that shows that the work at Oxford that I'm doing is different to everywhere else I've worked and, yeah, that excites me.[Jessica] That's amazing. Taking some of the biggest and brightest minds on the planet and having them... [Ben] Exactly. So I'm not an I'm not an academic, and I've mentioned before many times that I'm a university dropout, so I have a different skill set. Right? I can, you know, read WCAG out loud, you know, in my sleep. But that doesn't help with the world of AI and machine learning and the fact that collaboration is a part of life at the university. Yeah, it's going to make my life a lot easier. And, hopefully eventually better the lives of people with disabilities.[Jessica] That is wonderful. I'm so pleased for you.[Ben] Honestly. Honestly, I'm so happy.[Jessica] And so pleased for us, honestly. I'm so excited to know that that's what they're working on. I'm like, yes, please! Go forth and do, and change the world. That's a wrap for this episode of The Outer Circle. We called it that because accessibility can sometimes feel like a closed-off world, like there’s an inner circle of experts and everyone else is stuck on the outside. But that’s not how it should be. We talk to the people doing the work, from seasoned pros to unsung heroes, because their stories help us all see accessibility from a different angle. Sometimes surprising, always insightful. If this resonated with you, follow the show, and share it with someone who needs to hear it. Until next time, keep breaking barriers.