GitHub’s ‘General-Purpose Accessibility Agent’

GitHub’s Eric Bailey wrote last week in a blog post about the company’s efforts to build what’s called “a general-purpose accessibility agent.” The program is deemed “experimental,” with the pilot reinforcing GitHub’s commitment to accessibility.

“It is an understatement to say agents have become a popular way of working with code,” Bailey said on Friday. “GitHub has adopted agent-based code creation and editing for many of its initiatives, including piloting an agent to help with our commitment to accessibility.”

Apropos of GitHub’s ethos, Bailey’s post is pretty technical in nature. He cogently states up front, though, the accessibility agent has two primary goals: (1) provide engineers with reliable, timely answers to accessibility-centric questions; and (2) identifying and remediating issues prior to code going into production. The agent has reviewed 3,535 pull requests with a 68% resolution rate thus far, according to Bailey.

Notably, the “Mindset” section of Bailey’s post is well worth a shoutout.

“The social model of disability teaches us that access barriers—and consequently impairment—can be created because of how an environment is built. The same thinking applies to digital experiences,” he wrote. “With the accessibility agent, we are not attempting to ‘solve’ accessibility in isolation. We are instead attempting to augment our peers’ efforts, to better help them remove the barriers that may be created as a result of how we construct GitHub’s user interfaces.”

Bailey continues to say the accessibility is no silver bullet that can magically fix everything, adding “understanding, honoring, and socializing this better helps set the agent’s scope of responsibility.” This approach, he noted, has helped a lot to “[speed] up the experiment’s launch, leading to more buy-in for the effort.”

More poignancy from Bailey: “To say it plainly: Organizations will be at a disadvantage if they have not already invested in manually identifying and remediating accessibility issues. There are many reasons for this, including building an accessibility agent. To that point, GitHub has a mature system in place for logging accessibility issues, as well as verifying fixes to issues are working as intended.”

For all the handwringing in some design spaces about the rise in prominence of artificial intelligence in software development leading to perhaps trading craft for expediency, there can be no denying the impressiveness of the technological feat in things like, say, Claude Code to aid programmers in their work. By the same token, this rise is also beneficial for coders with disabilities who, for instance, may be unable to endure long work sessions that involve producing hundreds of lines of code. To have the ability to give an agent a few-sentence prompt of what they need or what and have it virtually instantaneously generated is a remarkable assistive technology. And to GitHub’s efforts, the idea that leveraging AI to help track accessibility when writing code is to lean on a computer’s greatest strength: automation. Humans can do only so much, so the one-two punch of the proverbial man and machine can only help to make computers and the internet more accessible to disabled people. As I’ve written in the recent past, ChatGPT did a lot to assist me, the decided non-web developer, generate CSS/JavaScript code for Curb Cuts’ visual design. The chatbot gave me good code, too!

Relatedly: My 2023 interview with GitHub accessibility boss, Ed Summers.