I’ve read Accessibility Has Failed: Try Generative UI = Individualized UX from Jakob Nielsen (opens in new window) and immediately felt the need to express my thoughts about some parts in the article that I can’t agree with and some parts that may, maybe, help to improve accessibility. Perhaps.
Jakob’s experience and contributions in the field of web usability are really on the level of a guru, but I can’t really agree with his claim that traditional methods for accessibility failed to substantially improve usability for disabled users. Maybe that was just a click-bait and marketing, but I still can’t agree with the thesis…
In my experience we are still far from those methods actually being put into products and I think that is the main problem. When we look at Web Content Accessibility Guidelines (WCAG) we know that almost no website or mobile app really can achieve conformance to all success criteria. That is the first problem. The second problem is that, yes, WCAG is far from perfect and doesn’t guarantee accessibility even when technically conforming, but chances are way better than when WCAG is not considered at all.
Saying that WCAG (I guess that he meant that by traditional methods) failed is not really fair. WCAG didn’t even get the chance when we consider the current level of it’s implementation in reality. I do agree that it seems like it has failed, but it is only because it was (and often still is) ignored for all these years.
I feel that a metaphor is in place for this – it’s like claiming that medication is useless and not helping the patient, but at the same time knowing that patient didn’t even take the medication.
As mentioned in “Don’t care about accessibility? Resistance is futile!” – I was relatively late when it comes to awareness of WCAG. And when I look around (websites, applications, repositories) I see that people are still not aware of (or just ignore) WCAG. So claiming that WCAG is not working is not really fair – WCAG needs to get a chance first. Sure, it’s not ideal. But I know now, that considering WCAG early on when making digital products make them more accessible. Sometimes we need to reach out of WCAG, for sure, especially when WCAG is a bit stuck, considering how fast our technologies evolve, but if we will continue to use HTML for web content, then we can still rely on WCAG to improve it’s accessibility.
And even if we totally abandon HTML and move to some user interface that we can just discuss with (like a chat) – large portions of WCAG will still apply (WCAG really tries hard to be technologically agnostic), so even if we get personalized interfaces we will still need to respect WCAG, at least some of it.
I also can’t agree that computers are difficult, slow and unpleasant for disabled users. That is not necessary true, at least. Again, if WCAG is totally ignored, then yes, difficult and unpleasant for sure. And slow also. But on the other hand I have seen some people with disabilities enjoying accessible websites and applications, and some of them were even way faster using them than I was, with my sight following the mouse pointer instead of just jumping right to the form, as a simple example.
Accessibility may be expensive, but I still see it as a systematic problem that will improve when more people will be aware and understand. So again, having standards that are not perfect, but still helping a lot doesn’t help when those standards are ignored. Ignored by people who make the systems. Ignored by people who teach people that make the systems. Ignored by people who pay for the systems. It’s not strange that ignorance cost a lot on the end when somebody finds out that the whole system is inaccessible. And then people need to go back, check the whole system and gradually fix it (or rewrite it). That is certainly expensive. If awareness and knowledge were there from the start – accessibility would not need to be expensive at all. In my post “Mind the accessibility gaps – most of accessibility issues originate in design and how to fix that” I refer to a study that established where inaccessibility probably originated in the first place. Preventing this would certainly make the whole effort much less expensive. It’s like security – either you implement it all the time and it will not cost you a lot, or you implement it at the end and it will really be expensive…
Most blind users don’t really have problems with two-dimensional interfaces if they are accessible. So claiming that two-dimensional graphical user interfaces are substandard for blind users is not always true. Sure, there are parts where it is a bit more difficult, but those parts are usually also difficult for sighted users. A complex table can be difficult for everybody, so we need to consider how to make it understandable and simplify it. Some diagrams, complex images and maps are certainly more difficult to understand and need proper text alternatives, but this will remain a problem for any alternative user interface as well.
Generative User Interface as the answer – perhaps yes, perhaps even more accessibility and other issues
At first I kind of liked the idea of a dynamic interface that would adapt to the user based on user preferences. That is, kind of, already happening actually. We have personalized interface possibilities – like high contrast mode and similar – in the operating system for years, we have assistive technologies that we can personalize, we also get more and more personalization with new CSS features.
Our tools to offer more personalization are getting better and I guess there are large potentials with generative user interfaces that are really tailored based on single persons preferences (prompts). This would be perhaps ideal in theory, even if that goes directly against universal design in a way. Perhaps generative user interfaces can work for some use cases, like for example talking with a website or document instead of classical ways.
I just have some issues with buying into this idea of generative user interfaces:
- AI is expensive – not only in terms of money but also environment and speed. So using generative user interface to “talk” with a website costs extremely more than if we use existing (assistive) technology to access it (if it is accessible, of course).
- AI adds another layer between us and information and this layer can be dangerous when we need to absolutely trust the information. It can reveal training biases, it can hallucinate, it can be lazy (save tokens) and it can also be manipulated, technically. So while we need to trust the source in the first place we will also need to trust additional layer between us and the source.
- We need to reveal a lot about ourselves to the AI if we want it to generate personalized interfaces for us. This information can be collected and abused later (reminds me also about accessibility overlays where we need to reveal our disabilities to the tool that collects this information in their analytics). While currently our assistive technologies actually protect us from tracking (at least for most parts, looking at you – mobile apps).
To conclude – I can’t predict the future, but…
I really can’t predict the future, but I am certain that it’s too soon to claim accessibility has failed. We haven’t really allow it to grow into our society. With current legislation trends I am quite sure we will soon see more positive effects. Maybe not a lot of WCAG conformance yet, but nevertheless improvements. And with current legislation improvements we should also see better awareness, more schools teaching accessibility, more producers hiring accessibility specialists and people with disabilities to help them make more accessible products and yes, also AI cooperating with people to make accessibility better.
It’s not fair to claim accessibility has failed when we haven’t really started with it properly.
My reflection on Jakob’s thesis that accessibility has failed…
It’s not fair to claim accessibility has failed when we haven’t really started with it properly. And there are more and more products that are quite accessible already. Otherwise I would not be able to know, personally, people with and without disabilities that can live and work mostly independently, provided their environment and tools are made in accessible ways.
I love the idea of AI collaboration, but I am afraid that we can’t rely on it too much. Even when it will be trained with “perfect” data it will still need human responsibility…