We need to talk about Jakob
by Matt May
Jakob Nielsen woke up one morning last week and chose violence.
If you haven’t read his recent post, “Accessibility Has Failed: Try Generative UI = Individualized UX”… well, you can, I guess? I had to do it, though, so hopefully some of you won’t have to. It’s… thoughtless. Hopeless. Soulless. Nielsen built his reputation on sharing his hot takes with a nascent blogosphere in the 1990s and early aughts, and if he hasn’t spent it all by now, I hope this finishes the job. What concerns me, though, is that he’s selling a class of executives hostile to disabled access a convenient fiction that will end up putting accessibility work on the back burner for a future which may never arrive.
Nielsen kicks off with some characteristic grandstanding. “I started promoting accessibility in 1996 when I worked at Sun Microsystems, but by no means claim to have been the first accessibility advocate,” he says. Well, that’s good, because he wasn’t. The first web accessibility guidelines date back to 1994, and the first OS-level accessibility API was released in 1995, after real actual advocates sued Microsoft to make it happen. Screen readers themselves date back at least to the early 1980s, and foundational research dates back to the mainframe era. You’ll forgive us, Jakob, if we aren’t carving your face into the Mount Rushmore of accessibility.
If Nielsen was trying to pretend that this was an act of advocacy toward disabled users, albeit at the hands of the ineffectual “accessibility community,” the mask drops quickly. The first bullet point explaining why accessibility is a failure begins (direct quote): “Accessibility is too expensive.” He goes on to sound like the CEO who really wants to try but ugh, so many costs and complications. Sorry, folks, we’d really love to care about you, but that just fell below the cut line. Later, he doubles down by citing the amount of money on the table if one were to condescend to support a disabled user. The subtext, of course, being that one is only worthwhile if one is a good consumer. The ableism and victim-blaming here reek to high heaven.
But wait, there’s more! The other bullet, in full:
Accessibility is doomed to create a substandard user experience, no matter how much a company invests, particularly for blind users who are given a linear (one-dimensional) auditory user interface to represent the two-dimensional graphical user interface (GUI) designed for most users.
Seems like one could find actual sources for this, no?
Nielsen sounds like a dilettante on these issues, because he is. This part of the post isn’t so much an argument on the merits of disabled access as it is a projection of himself in the shoes of a blind user, and how utterly miserable he thinks it would be. At no point in any of this—again, classic Jakob Nielsen style—does he cite an actual blind user, much less any blind assistive technology researchers or developers, like Mick Curran and Jamie Teh, the creators of the NVDA screen reader. Or Chris Hofstader, who helped build JAWS and has written for over a decade about blind UX. Or Chancey Fleet, who teaches 3D modeling using screen readers at the New York Public Library. Or Chieko Asakawa, one of IBM’s most decorated computer scientists. Or TV Raman, who among numerous other things built a screen reader inside the Emacs text editor. Or Josh Miele, a literal MacArthur Fellow for his work in the field, including tactile maps and object identification.
It makes my blood boil that there are numerous actual experts, who are not hard to find, and could talk about their own experiences in great detail. (Isn’t that what usability people usually look for?) But instead, Jakob ignores them in order to talk about his feelings as facts. I’d love to hear how they’d speak to their own lived experience with assistive technology, but it’s the weekend as I write this, and I don’t want to ruin theirs by making them read this garbage.
If Nielsen’s post were to be the only thing a UX practitioner reads about accessibility—and sadly, for some it will be—they could be forgiven for taking away that the only disabilities that need addressing are blindness and low vision. For example, there’s literally no mention of automated captions for deaf and hard of hearing users, an AI-driven feature that’s dramatically increased basic access to videoconferencing, games and face-to-face interactions (with the caveat that it’s also squeezing out human captioners from scripted TV and film work, where humans can do more polished work).
He says that “old users and low-literacy users” are actually well-served by the status quo (again, without sources), but support for other disabled users “will never happen” (emphasis his), and therefore, we must abandon all hope for equitably-designed software products and wait for the machines to save us.
At this point, I have to tell you what’s about to happen. Nielsen has given everyone who sees accessibility as an unnecessary cost center a way out. Listen, they will say, the experts are telling us that AI is going to solve this, so let’s deprioritize the work now so we don’t duplicate our efforts.
I’ve been through this hype cycle before. In 2004, Jeff Veen, then at Adaptive Path, told a crowd of fawning web designers at SXSW, “I don’t care about accessibility.” It was a rhetorical flourish meant to show how far one would get by designing with web standards. I met Jeff at that session, and worked with him later on. He cared. But here’s the thing. People heard that one sentence, and they fell in love with it, because it was cheaper to do what they wanted that way. And it served to undermine a lot of the work that accessibility people were doing, even though that wasn’t his intent. What’s troubling to me in this case is not just that Nielsen gives lazy execs a big, fat get-out-of-accessibility-debt pass, he seems to actually believe what he’s saying.
There is one area where Nielsen is dangerously close to having a point. I think some form of generative AI is going to end up being beneficial for us, as individuals, over the next several years, and some of it is useful in narrow domains now. For example, I absolutely hated doing literature reviews in school, something I had to do this week—ironically, on the very subject of UX and cognitive disabilities. This time, I used Consensus in ChatGPT and Perplexity’s built-in academic assistant in place of the clunky search tools I used to use. They helped a lot in the search-refine-skim phase, and provided (real) links to the source papers, some dating back to the 1950s. They made me at least twice as effective with my time, such that working with them for a couple hours easily justified paying the annual subscription. Would I say everyone in every academic domain should jump on board? No. All I know is that it made my work better.
(Sidebar: Someday I’ll document my ethics relating to gen-AI tools, but for now: if my name is on it, I wrote it. If I cite it, I read the source in full, not a generated summary. Where there are exceptions, I’ll call them out.)
Where he goes off the rails on it is another matter. The thesis of his post is made of “those people” energy. It would be one thing if Nielsen were to say that generative AI or generative UX had potential for every user, because if anybody can manage to get it right with any reliability, it does. But instead, he looks at the whole landscape and decides, hey, it’s going great for abled users. It’s only the disabled ones who “the accessibility community” has failed, and they need user experiences poofed into existence for them by the magic box. Never mind that it’ll be a different experience each time, based on what model is used and how, contravening decades of UX theory. What matters is we can say we solved it! Then we can wash our hands of them once and for all.
Disabled people are tired of being the canaries in each technical coal mine. What we know from the history of accessibility as a bleeding-edge technology is that not nearly enough vetting is done to ensure that what is created is in fact beneficial, much less generalizable. If that’s led to the kind of morass Nielsen paints, then what in the absolute hell do you think is going to happen with “generative UX”? So much of the field of gen AI is driven by overpromises. We’ve heard those for decades in tech, and over a century when you add things like wheelchair access to the pile. (Did you know: there are physical disabilities in addition to sensory ones? You wouldn’t know from that article!)
If this idea of a generative user experience is such an unvarnished good, so universally applicable and feasible, and so obviously better than what we have now, then why would you limit the benefits of such a thing solely to disabled users? Isn’t this something that everyone should want? Or have you just decided for the global audience of users that a curated experience, one that’s tested, secure and respectful of the user’s privacy, just isn’t worth doing well if your user happens to be disabled?
Don’t worry. We know the answer.