As government agencies increasingly work to improve accessibility, in part due to the Department of Justice’s April mandate to ensure digital accessibility, some stakeholders are turning optimistically to AI. However, others are underlining potential risks.
BREAKING DOWN THE MOU
First, the partners will establish opportunities for relationship-building and information sharing across disability and technology communities. Second, they will identify and develop potential solutions to civil rights concerns, accessibility barriers and AI risks. Third, they will provide information, assistance and resources to enable both disability and technology communities to ensure safe, accessible and equitable creation and use of AI.
This MOU is historic for the Access Board as its first formal MOU with organizations outside of the federal government, according to Amy Nieves, public affairs specialist in the Access Board’s Office of Executive Director. And while its focus is at the federal level, the impacts will likely be far-reaching.
Implementing AI in a responsible and accessible way is a shared effort across the federal government, Nieves said, and the work is intended to benefit people with disabilities — but also to ensure “safe and equitable AI use for everyone.” To achieve this, the organizations will work together and with federal agencies’ chief AI officers throughout the process.
“We really wanted this to be a formalized partnership where we were taking this de-siloed and collaborative approach to how we would address AI in the disability community,” Nieves said.
The collaboration, Nieves said, teams the Access Board’s expertise on government with CDT’s tech savvy and AAPD’s expertise on people with disabilities, for a comprehensive approach. The three entities, she said, represent government, industry and community. Part of what led to this MOU, Nieves said, is the federal executive order on AI, which emphasizes accessibility best practices and ensuring people with disabilities benefit from AI, while being protected from risks.
As she explained, as AI is rapidly advancing, it is creating both opportunities and barriers for people with disabilities. The MOU’s ultimate goal is to have a lasting impact in positioning government and industry to keep up with the rapid rate of change; the Access Board, meanwhile, wants the MOU to benefit everyone, not just those with disabilities.
The endeavor’s focus is at the federal level but, AAPD Technology Policy Consultant Henry Claypool said, there will be implications for state government. Many programs, he said, are federally funded and state-administered.
“So yes, the federal agencies should be providing guidance to states,” he said, explaining that lessons learned by the federal government throughout the executive order implementation should help states be better stewards of federal funding. Claypool said state and local governments will have a lot to learn from the federal government as they work to improve accessibility. To address this, the MOU's initial phase involves scoping out the longer-term work ahead to enable issue-specific community engagement.
For those interested in learning more about this MOU and the AI-related work these groups are embarking on, the Access Board is hosting an informational session with CDT and AAPD on AI at 3:00 p.m. Eastern time on July 9.
AI RISKS AND OPPORTUNITIES FOR ACCESSIBILITY
A key part of the MOU is balancing AI risks with potential benefits.
Administrative tasks like scheduling meetings or correcting grammar provide some opportunities for the use of AI to benefit people with disabilities, Claypool said. The MOU, however, goes beyond, aiming to help inform organizations of AI-specific issues and risks that people with disabilities face, and which largely stem from systemic issues like bias or gaps in data collection practices.
“I think measured optimism is the right disposition here,” Claypool said. He noted that automated decision-making poses risks for people with disabilities, especially when used around benefits programs that people with disabilities rely on for their well-being.
Another risk it poses for people with disabilities is in the inequities created by biometric recognition technology, Ariana Aboulafia, CDT policy counsel for disability rights in technology policy, said. As she explained, there is an “inherent incompatibility” in using these systems to serve people with disabilities. This is because the technology is trained on pattern recognition, and people with disabilities may not adhere to a standard pattern.
As Aboulafia put it, the work to mitigate risks for this population is more complex than stating the benefits and harms of AI technologies. It’s about taking a more intentional approach to understand the reason those risks or benefits exist, and what measured steps can be taken to minimize those risks.
But there will not be a one-size-fits-all solution, she said: “People with different disabilities experience different systems in different ways, and they also experience technology in different ways.”
As such, she underlined the importance of approaching this work with the perspective of intersectionality, understanding that people with multiple disabilities or those who belong to multiple marginalized groups may face more barriers.
Inclusive design work regarding AI technology should leverage the same principles as inclusive design of physical spaces, which requires engagement and participation from those being served, Aboulafia emphasized.
“I really think that the more we know, the better we will do here,” she said. As such, she encouraged other individuals and organizations to participate in this effort to make AI accessible for all.
