GAAD Founders Talk Accessibility in new Interview

In late January, I ran an interview featuring Global Accessibility Awareness Day (GAAD) co-founder Joe Devon and ServiceNow vice president and global head of digital accessibility and globalization, Eamon McErlean, in which the men talked about the state of digital accessibility and the AI Model Accessibility Checker. As I reported, the API is described as “[evaluating and comparing] how well coding-focused large language models (LLMs) generate accessible code [by providing] benchmarks for companies to test and demonstrate the accessibility of their models’ output.”

Fast-forward several months later, and we’re a few weeks removed from GAAD’s 15th anniversary on Thursday, May 21. I reconnected with Devon, as well as his co-founder Jennison Asuncion, to further discuss the current state of digital accessibility. Devon said the current state is in “flux,” pointing to legislation like the European Accessibility Act and its effect to “strengthen” accessibility across the continent. Meanwhile, in North America, the United States feels decidedly “murkier.” Much of the attention here, he added, is paid to artificial intelligence—something he acknowledged is “transforming software development and accessibility along with it.” For his part, Devon fancies himself an “AI optimist,” but told me there’s both good and bad in it.

“We will have to wait and see what the next two years bring,” he said.

To Devon, a big yet “underestimated” challenge for the disability community vis-a-vis access is a general lack of knowledge on the subject of accessibility. People can’t institute what they don’t know or understand. The solutions that do exist, he added, are “unevenly distributed” throughout the tech industry and pointed to learning about advanced iPhone features while attending a workshop. They were so revelatory to him that he now uses them on an everyday basis. Such an approach would do wonders for accessibility and the community writ large, he said, calling it “a world of good.”

Asuncion concurred, telling me “there is a call to action here in which organizations that serve people with disabilities need to prioritize AI literacy and training on how to use the different tools, such as Copilot, Gemini, and ChatGPT.” AI is not going away, probably ever, he added, and noted things are moving so quickly—so fast, in fact, he believes “there is a real danger that people with disabilities will be left behind.”

“We need to have a workforce within [research and development] that includes engineers, designers, product managers, and others who have the lived experience of disability, actually helping build digital products and technology,” Asuncion said. “That requires intentional actions, such as mentorship, targeted recruiting of new grads with disabilities graduating from relevant disciplines, and so on. Separately, it goes without saying that having product and technical leadership explicitly mandating that accessibility be a core requirement, prior to launch, would go a long way.”

Devon agreed, saying companies must hire disabled people to test products, as well as imploring said makers to “bring them into every phase of product development.”

When asked whether digital inaccessibility is reflective of society’s collective view on disability, Devon demurred and said it’s hard to answer concretely. Instead, he told me the “greatest problem” with exclusion is software complexity. A web developer himself, Devon explained web standards and frameworks which prove popular are things invented by giant tech companies to address scalability and other problems aimed at the enterprise. Small businesses or individual pursuits are irrelevant. These big corporations, Devon went on to say, need complex systems—and developers know full well complex is where the money is. They drop what he characterized as “the easier, more semantic path”—which ultimately means accessibility is a “casualty.”

Asuncion piggybacked on Devon’s sentiments, telling me he believes there remains “a wide gap” amongst many companies in understanding people with disabilities use technology and the internet like anyone else. We visit the same websites and use the same apps. The reality, according to Asuncion, is sobering: “disability is still an unknown [and] uncomfortable topic to many people,” he said. It’s unfortunate, he told me, but the disability community is mostly viewed as “a niche group who are largely at the margins with bigger problems” than wanting to use technology like everyone else.

“For some reason, disability is treated as not a profit center,” Devon said. “Remember the frenzy by executives clamoring to get a foot in the door of the enormous Chinese market of 1.4 billion people? Aside from India, the market of people with disabilities is almost as large… 1.3 billion people and growing (whereas China’s [disabled populace] is shrinking). The finance people should be excited by serving this enormous market, but it’s up to us to understand the demographic story, and teach it to executives.”

Back to the AI Model Accessibility Checker. AIMAC, as it’s known, is “going quite well,” according to Devon. He lauded Open AI for “killing it” with its latest models, also noting they achieved “an almost perfect score on our benchmark.” Moreover, Devon praised Anthropic for its embrace of accessibility and pointed to a new job listing for a staff software engineer for accessibility based in San Francisco, Seattle, and New York City.

Anthropic’s job posting is “a great sign,” Devon said.

As to AIMAC’s future, Devon told me work is ongoing for “developing a more scientific treatment” for the next version of the API. To that end, Devon and McErlean have tapped Seattle-based Ph.D student Yumeng Ma to assist with building the project.