In addition to that, if an AI is able to understand who their user is based on their race, gender, or disability, AI tends to treat that user differently and not always for the better. There's also huge errors within AI. While AI, we're trying to build it to be much more efficient, more accurate, right now in the industry, we're still trying to move in this direction of getting AI to understand how confident it is in its output.
The example earlier about Be My Eyes, I don't know how the application works, but there's a very different level of confidence you need the AI engine to have between writing alt text for an image on your website versus describing the street and whether you can cross it. Both of those options have very high different levels of risk, and if you're going to be building an application that's important for a user and has safety concerns, you need to ensure that AI is able to give accurate results.
And on top of all that, a lot of users can be overly trusting of AI results, even when the information is inaccurate or very harmful to users. So if AI isn't this perfect blanket answer, if we can't just hand over AI the keys to make everything accessible, what is the thing that we need to do here? Well, here are some things you can do as an engineer. First, you can clean up your HTML. Most of the issues that are on the web right now are a result of improper HTML using the wrong tags, wrong attributes, and if you're able to go through and clean up and write well-structured code, you're going to fix a lot of the problems that hurt our users.
Also, get an accessibility browser tool. I use Axe Core Browser Extension. I also use Accessibility Insights for the Web, which is a Microsoft product. I've even tried Lighthouse a little bit. Using these tools will allow you to scan websites that you're navigating and see what are the issues, especially when you're developing. And please, if there's one big takeaway, please add automated testing into your development pipeline. It's a really easy thing to catch accessibility bugs before you ship them out to production and cause a lot of those issues that you see all over your website.
There's so many applications you can plug into your pipeline. I included a few here like Axe Core, Lighthouse CLI, Selenium from Deque, but I've also heard even at this conference, people using Plywright and the accessibility extensions they have there. I encourage you to check it out. Now say you're not an engineer and you're here at this conference. Welcome. If you're an organization, I highly encourage you to understand your compliance level and identify what you need to understand what your compliance level is.
And then after that, dig into all the issues that your customers are seeing and experiencing on your website. And if you have some money within your budget to hire accessibility experts, please do hire accessibility experts. Hire them in engineering, project management, design, user research, and testing. These are all areas that if you have an accessibility expert on staff, they can give you the best advice on how to make really clean and compliant code and a really great experience for your customers. In addition, listen to your users. If you have the ability to gather feedback from your customers, do that. Or if best case scenario, run user studies for them to navigate through your site using whatever tools they need and get honest feedback on how accessible your site is, what's interesting, what's clear, and what's confusing and not working.
And overall, my main point in this presentation is just to use AI strategically.
Comments