After We Get In: What Accessibility Already Knows.
GPT generated image. The prompt was my written blog. “Inclusive Future Through the Doorway”
There is something lovely about the word accessibility. At its root is the idea that something should be possible to access.
And when something is inaccessible, it raises a question. Why? What is it about the environment, the information, the technology, or the way something was designed that makes access difficult for some people?
I like that accessibility begins there. It does not begin by locating a problem in the person. It asks us to look at the relationship between a person and whatever they are trying to reach.
We can see this in buildings. A ramp is not a favor extended to some visitors. Accessibility is built into the standards governing how public spaces are designed and altered. The expectation is that people with disabilities should be able to approach, enter, and use those spaces (U.S. Department of Justice, 2010).
The same idea moved into technology. The Web Content Accessibility Guidelines, or WCAG, established shared standards for making digital content perceivable, operable, understandable, and usable by people with disabilities (World Wide Web Consortium, 2023). We build captions because not everyone hears in the same way. We build screen readers because not everyone sees in the same way. We create different ways to communicate, navigate, and interact because people move through the world differently.
To build for accessibility is already to acknowledge human variability. That part gets lost.
Accessibility becomes a standard, a feature, a requirement, a box to check. Those things matter, but underneath them is a much larger idea. We know people are different, and we have already agreed that sometimes the technology should adapt to the person rather than expecting the person to adapt to the technology.
Accessibility is a frame we build inside, not a feature we add at the end…and it has a way of proving itself.
Curb cuts emerged from disability activism and the need for wheelchair access, but the resulting change also made movement easier for people pushing strollers, pulling luggage, making deliveries, and moving carts (Blackwell, 2017). Captions make audiovisual content accessible to Deaf and hard-of-hearing people, but research has also found benefits for comprehension, attention, memory, language learning, and people consuming media in environments where listening is difficult or impossible (Gernsbacher, 2015). Speech recognition likewise has a long history as a form of alternative computer access for people with disabilities and is now embedded throughout mainstream consumer technology (Noyes et al., 1989). It’s not enough, but it’s something.
Technology keeps moving, though. And now it is difficult to talk about technology without talking about artificial intelligence.
AI does more than help us access something. It increasingly interprets what we say, summarizes what we know, identifies patterns in our behavior, makes recommendations, assigns categories, and shapes decisions. People who study disability and technology have been asking for a while what happens when those systems carry forward existing definitions of normality, ability, and desirable behavior (Whittaker et al., 2019; El Morr et al., 2024).
To illustrate this, picture a system that transcribes someone's speech flawlessly, accepts their account in whatever order it arrives, and then summarizes it as "unclear." Happens all the time with any transcription AI from Otter to Fireflies, to speech to text in our phones.
An AI system can be accessible and still misunderstand the person using it. It can accept different ways of communicating and still judge those communications against a narrow idea of what is normal. It can recognize human variability when someone is trying to access the system and lose sight of that variability once it begins interpreting or evaluating them.
We have a standard for whether a person can get in. I am not sure we have one for what a system is allowed to conclude about a person once they are there.
That is what interests me. Not what comes instead of accessibility. What accessibility already knows, and what we might miss when that knowledge does not travel with the technology.
The question I keep returning to is why that recognition so often stops at the door. That is where I begin thinking about something I have been calling cognitive dignity. I’m still cooking on the concept and further depth but, dignity is what’s important in these spaces.
If accessibility recognizes that people should not have to interact with technology in one prescribed way, what would it mean to carry that same recognition into the way artificial intelligence understands people?
What happens when access is no longer the end of the question, but the beginning of another one?
ReferencesBlackwell, A. G. (2017). The curb-cut effect. Stanford Social Innovation Review, 15(1), 28–33.El Morr, C., Kundi, B., Mobeen, F., Taleghani, S., El-Lahib, Y., & Gorman, R. (2024). AI and disability: A systematic scoping review. Health Informatics Journal, 30(3). https://doi.org/10.1177/14604582241285743Gernsbacher, M. A. (2015). Video captions benefit everyone. Policy Insights from the Behavioral and Brain Sciences, 2(1), 195–202.Noyes, J. M., Haigh, R., & Starr, A. F. (1989). Automatic speech recognition for disabled people. Applied Ergonomics, 20(4), 293–298. https://doi.org/10.1016/0003-6870(89)90193-2 U.S. Department of Justice. (2010). 2010 ADA standards for accessible design. Civil Rights Division.Whittaker, M., Alper, M., Bennett, C. L., Hendren, S., Kaziunas, E., Mills, M., Morris, M. R., Rankin, J. L., Rogers, E., Salas, M., & West, S. M. (2019). Disability, bias & AI. AI Now Institute.World Wide Web Consortium. (2023). Web Content Accessibility Guidelines (WCAG) 2.2. W3C Recommendation. https://www.w3.org/TR/WCAG22/