26 Comments

As a long-time reader of your work (dating back to the earliest days of useit.com), I’m genuinely shocked that you’d write this piece, Jakob. It simultaneously diminishes the significant gains in accessibility that various hardware, software, and ux design advances have achieved; flatters the leadership of organizations that refused to invest in accessibility by suggesting they’ve “tried and failed” rather than “gestured and forgotten;” and ignores the significant (yet unproven) work that is necessary to achieve consistency and control over specific outcomes with generative AI. Individually they’re serious oversights, but together I’m left scratching my head.

Solving these problems with AI is *more* work, not less, than simply applying established a11y practices; it makes sense when the resulting systems can be applied at an industrial/enterprise scale, but still relies on the foundational work that large organizations have resisted for cost reasons.

Expand full comment
Mar 1·edited Mar 2

Also shocked. Parameters would still need to be set for AI outputs and the results in different scenarios would need to be tested. And I am at a university that is setting up testing with different users, so Nielsen is wrong that this will not happen. Only if we stop trying to make the world better for everyone will change not happen.

Expand full comment
Mar 5·edited Mar 5

Also shocked. I was at a UX conference today and Jacob's general well being came up. We were all worried something is actually wrong with him. This type of delusional, sloppy writing just doesn't feel right.

Expand full comment
Feb 29·edited Feb 29

This assumes the system has enough context to know the current state and capabilities of the user, all other applications and devices in use/conjunction with it, that there is only one user / not a social setting with diverse user needs. These are not trivial issues.

It also introduces complex problems of conventions and standards in UIs and interactions when every person has a completely custom interface how do we work with one another, further deceasing usability. In the same approach as focusing on usability rather than accessibility, doesn't Universal Design approach solve this problem? @Jakob, I'm genuinely curious how you would approach these implications.

Expand full comment

For all the good Jakob Nielsen has done for product and service design, this is a both/and situation.

You can BOTH recognize his initial contributions to this space AND understand people you hold in high esteem can be problematic and sometimes need to take a seat and listen.

When leaders with intense followings (or not, in some cases), use their platform to produce content like Jakob’s article without regard for the disability community or hard work of accessibility professionals everywhere, it sends a message. And in this world of “sound bites” and quick hit “truths”, this can be very dangerous. One cannot assume people read this type of article with the ability to critically think about the implications for disability and accessibility - especially executive leaders in large organizations who have little to no experience with the disability community or the nuanced work of accessibility.

Accessibility is not debatable. It’s not a place for a “devil’s advocate” argument. Accessibility is helpful for many and a NECESSITY for many more.

If you’re debating or “arguing” access, it’s your time to sit down, recognize your privilege, and think about where your energy can be better spent.

I’m tired.

It’s not accessibility that has failed, Jakob - it’s people like you who continue to create barriers for moving accessibility forward with problematic “thought pieces” that completely disregard the people actually working (and LIVING) in the disability and accessibility space.

Everyone should expect more from you.

To hear opinions from actual people doing hard work in this space, start listening to disabled voices and accessibility professionals:

https://www.linkedin.com/posts/sheribyrnehaber_userexperience-artificialintelligence-accessibility-activity-7169454946420011008-Ucee?utm_source=share&utm_medium=member_ios

https://www.linkedin.com/posts/kileyjaneschmidt_accessibility-genai-activity-7169997317146656769--tRu?utm_source=share&utm_medium=member_ios

https://www.linkedin.com/posts/khallconsultant_in-light-of-jakob-nielsens-fallacy-riddled-activity-7169685050794459137-X0Ns?utm_source=share&utm_medium=member_ios

Expand full comment

https://www.briandeconinck.com/jakob-nielsens-bad-ideas-about-accessibility/

In case anyone wants to understand more about accessibility and have a critical think on Jakob’s problematic and pitiful article.

Expand full comment

Why would anyone suggest that just because it is expensive, you should not do it at all? Didn't all designers take a vow to make interfaces simple for 'all' users?

Secondly, the concept of generating UI runtime seems like a far fetched and dangerous concept, where the second generation generative UI will need a lot of data, mostly sensetive information about the user disabilities.

This is not encouraging for the people who are trying really hard to make the content universally available, especially coming from an icon in usability studies.

Expand full comment
Mar 12·edited Mar 12

I'm skeptical of the trust that's put into generative UI to be able to adapt to differently-abled users. On one hand, adaptive, ability-based design for accessibility is a concept that's been around for a while, and the concept of second-generation generative UI the way you suggest could be one way of trying to carry out those aims. Second-generation generative UI is an interesting idea in theory, but rapidly generating new UIs afresh for unique users completely excludes testing and validation of design interventions, inevitably leading to faulty designs. This is even the case if the AI progressively learns from users throughout usage, because at the end of the day, the AI is still constrained to the specified parameters or data inputs that are part of the language model, which can't adapt to what the machine learns about users. It's a flawed approach in practice, I believe.

Expand full comment

I keep checking the date of the article thinking this might be their annual April Fools' day post... very uncharacteristic and bizarre recommendations.

Expand full comment

If, as you say, you "consider users with disabilities to be simply users", then the usability field has failed, not accessibility.

Expand full comment
Mar 4·edited Mar 4

Lots of comments felt offended by Jakob's 'discredit of the accessibility industry' in this article. Yet so far none of them addressed this one thing: the lack of accessibility in general is a natural result of product-market-fit under capitalism. You can argue that how important and morally just it is to include everyone but as long as there is more able-bodied users than users that require accessibility measures, companies don't care.

In a world chasing ever-higher profit margins, any efforts made towards accessibility will always have diminishing returns in contrast to i.e. making short term flashy new features for most users, even if there isn't any real value to it.

The whole premise of the article is presented under this sad reality, and devices that understand specific users, with or without AI, could be a new vector in improving accessibility in general.

Of course we are still gonna make those users pay for the devices, who am I kidding - like we already do under the current model, but AI agents could put right the context for a million types and combinations of accessibility requirements - without the traditional cost of doing so. I think in general thats a pretty strong argument.

Expand full comment

I can see how this could work:

1. The disabled person pulls on his brain-computer interface.

2. Then he attaches sensors to various body parts.

3. In real-time, the Generative UI adjusts the interface to accommodate the disabled person's capabilities at that moment (they change throughout the day). If cloud-based, each user will have their own privacy-protected cloud, guaranteed to never be hacked or exploited by the companies who insist accessibility pay for itself, at a minimum.

In the decades it will take to develop such automatic personalization, people with disabilities, who will no longer be served by the valiant efforts of the "Accessibility Community" because the AI fix is coming soon and it will be cheap and easy, can take a holiday.

Who (besides me) said, "Good design is accessible design; accessible design is good design"? The emerging cohort of designers and engineers are embracing this message, some even learning it at school (thanks Teach Access). I feel confident that they will build better, with all the tools at their command, including AI, long before people with disabilities will be required to be implanted into their computers in order to use them. This next generation has resilience, persistence, empathy and smarts and isn't ready to throw in the towel as Jakob encourages.

Expand full comment
Mar 4·edited Mar 4

The paradigm of user-centered design has gaps, especially when considering neurodivergent individuals. Many design approaches and texts were conceived before fully recognizing the relevance of neurodivergent diagnoses.

It is true that the assertion that there are too many types of disabilities for companies to conduct usability tests with representative customers is valid. However, it is overlooked that the needs of a color-blind person differ significantly from those of an autistic person. During my studies, I faced discrimination by receiving lower grades for using low contrast in color and typography, without initially understanding what they meant. Later, I understood that I was expected to ensure "accessibility to the other group of disabled people," even though their needs were different. We are literally on opposite ends, and it is impossible to design for both at the same time.

I deserve access to a website without fear of it causing me pain. I share the belief that there have been failures, and that we, the disabled individuals, should be the ones leading the change.

Expand full comment

False: “Accessibility has failed as a way to make computers usable for disabled users.”, if true, why are so many viably employed and successful? What data do you have to make such a broad statement? Sure, it could be easier and less expensive, but it is making progress, most say no worse or better the usability.

2. “Assessed this way…” is like saying that the radio experience will never be as good as the TV experience or the in-person experience. Computers are still difficult, slow, and unpleasant for everyone BECAUSE the last great experience is now the expected experience everywhere. Don’t keep moving the goal posts or change the game and expect the measurement of progress to improve.

I worked on the original “un-pack and set-up” evaluation with potential users before the IBM PC was launched in 1981. I’ve been working on accessibility since 1996 when WAI was launched at W3C.

4. I’ve witnessed some major changes in usability and accessibility over the decades. I’ve witnessed many users who have survived a traumatic brain injury or simply age that now easily accomplish many tasks on a smart phone that they could never do with a Windows 95 machine.

False: “Accessibility is too expensive”, same as good design and good usability. But true when NOT done correctly. Things such as shifting left into design systems, automation, and not trying to test-in-accessibility at the end, is several magnitudes less expensive.

True: “There are too many different types of disabilities to consider for most companies to be able to conduct usability testing with representative customers with every kind of disability.” And if they are trying to do it that way, no wonder it is too expensive. That’s where meeting accessibility standards that were created by conducting and gathering the outlier requirements makes sense.

False: “Accessibility is doomed to create a substandard user experience, no matter…”, see previous points on radio vs TV vs in-person, especially when the “experience” is based on multiple sensory inputs, and disabilities will limit one or more modalities and/or mute them all. How to measure an experience is as unique as the individual, but there are still opportunities for standards, testable criteria, and design systems. And, they will need to change and evolve as AI becomes more pervasive, for example, just as the smart phone changed the landscape over a decade ago.

I believe the “Old users and low-literacy users” are not the exception of accessibility, but the beneficiaries of the limited space of the smart phone real estate that forced it all to happen. Such as reflow, zoom, and task specific UIs instead of the old large screen desktop that can and does everything, and even does more with 3 or more multiple displays that some power developers use when employing all their multiple sensors. In contrast, I’ve witnessed and there are many cases of disabled users' ability to out text and out find content because of their mastery of certain skills and tools that even those power developers can't match.

Instead of saying that “Making it easy” is hard work, or that “Making it accessible is too expensive”, maybe we should acknowledge that they are both evolutionary, and should correctly take advantage of revolutionary milestones, such as AI is bringing before us. Alexa and Hey Google were the start of the simplified conversational model (a magnitude easier/accessible that the old phone - press 1 for blah blah), but that is very different than trying to do taxes for a business without data tables and the such.

And lets not “force” blind users to only have one short version of the image’s alt-text, or perhaps better said, they want the choice to have both, the short alt text and the longer description of the bear. A description and an alternative do not have the same definition in English.

AI is already making strides in providing “simplified” reading levels and “summaries” of longer texts. But not to always replace the original or to only produce one version because most often both versions are still necessary, same as alt and descriptions are still necessary at times.

I feel it all depends on the task, the situation (environment and technology), and the users’ capability at the moment. For example, a good mobile experience can be great, but not when I need to explore a data table like I can on a desktop, or write code across 3 displays. My masters was on situational leadership and how the leaders style needed to change based on the capabilities of those being led (or managed) and the tasks needing to be done. Playing “country club” leader (or nuclear engineer) with my toddler did not work well when they went running into the street and they didn’t even understand what a sidewalk was when told not to cross the sidewalk. Explaining the 48" average width composed of reinforced concrete did not help either. Inspiring leaders and empathic designers are great at certain things at certain times, but not always in every situation.

Approaches to accessibility and usability also depend on the situation, user’s capabilities, and the task(s). The secret is to know when to use which approach. And the wisdom is to keep trying, or as I say: Ready, Fire, Aim and keep iterating. My designer friends say: Treat everything as a prototype.

I still admire your work and the inspiration you've provided me personally.

Expand full comment

Innovation in accessibility should indeed be pursued, but not at the expense of dismissing the valuable progress already made. Your article could have been a call to action, a beacon for the next steps. Instead, it is a misguided dismissal of decades of hard work and achievement.

I am also surprised and disappointed to see you say that "Accessibility Has Failed" without adequately crediting all the hard work people have put into making tech more accessible for everyone, especially those with disabilities. You skipped the critical motto: "Nothing about us, without us." That means any conversation or decision about accessibility should include the people affected.

To call all the progress made so far a "failure," is pretty harsh and, honestly, not true. Much progress has been made to make tech more accessible for everyone, thanks to the hard work of advocates and people who care about an equitable experience for all users. Sure, there's still a long way to go, but saying it's all been a failure isn't fair to those who've put in the effort.

Making tech more accessible has been a team effort involving many people with disabilities sharing their experiences to improve things. Not including their voices in your solution is like saying we can solve these problems without the community's input is not an option.

It's also disheartening to see how quickly we fall into the trap of picking sides. This divisive manner does little to advance our collective understanding or to foster solutions that benefit everyone. The complexity and importance of making technology accessible to all demand that we transcend this tendency to polarize. It's time we embrace the nuances of these challenges rather than retreating into opposing camps.

Instead of ranting about Jakob's opinion, I believe that advancing this conversation requires us to adopt a tone of respect and collaboration. It is crucial to recognize that the journey towards full accessibility is a collective effort.

This path is always better with diverse perspectives and experiences, reminding us that our — shared — goal is to create a more inclusive digital world for everyone.

Expand full comment

Your perspective in this article seems ignorant and not conclusive. I get where you're coming from with AI and generative UI being the next big things for accessibility. It's an exciting vision, for sure! But, you're ignoring the most important piece of the puzzle, if we don't also recognize the hard work and progress made in accessibility so far. Those efforts have paved the way for where we are today, and they've taught us a lot about what works and what doesn't. As we dive into AI, let's not throw out those lessons. Instead, let's use them to make sure the new tech really works for everyone. It's about building on what we've learned, not starting from scratch. Also, just know that YOU ARE NOT ALONE as an expert!

Expand full comment

Jakob, how will the GenAI know that a user is low-literacy or older without mining their data? (If their browser or system cannot indicate this.)

Expand full comment

Jakob, this is intriguing. I agree with the basic premises that the status quo poorly serves users, and that gen-ui such as you describe could present far more appropriate UI to users. However, could you explain how this will avoid the first problem mentioned—Accessibility is too expensive? What you describe still requires separate experiences to be designed, tested, built, and deployed. Thank you for addressing the topic, I look forward to more.

Expand full comment

Accessibility features such as multi modal interaction benefit everyone, not just users with disabilities. For example, users with perfect vision may prefer to read and listen. To categorize some interfaces and not others reduces opportunities for all users to benefit because they are limited to one way of interacting.

Expand full comment