Exploring AI-powered solutions to accessibility: The accessibility browser
Part one - The ultimate accessibility browser: reframing the problem
There are many challenges associated with accessible technology, but I wanted to focus on the specific yet complex problem of why website owners end up with sites and services that aren’t accessible.
First, let's look at the problem.
According to the UK’s Government Digital Services 2022-24 monitoring report, 30% of public sector websites still didn’t comply with WCAG 2.2 AA. While this is a significant improvement on the previous report, this only focuses on a subset of public sector websites and services and doesn’t include business websites.
The bigger picture is that there are in the region of 1 billion non-compliant websites, worldwide, based on an average compliance rate of 4%.
When websites don’t comply with WCAG, browsers (and other specific assistive tech) get stuck or perform unexpectedly which causes difficulties for the user with the disability. At best, this means people with a disability get frustrated and it takes them longer to complete processes online; at worst, it means complete exclusion from a service, purchase or in some cases, lifeline.
Examples of positive, proactive approaches are growing. The GOV.UK team does an amazing job of maintaining and advising on compliant, accessible websites. They can do this because they have set up their organisation around achieving accessibility and inclusion across their estate.
For smaller government divisions, schools, charities and SMEs, this can be hard to achieve. They may lack both the in-house skills and/or the budget to carry out accessibility maintenance or remediation work.
So, they may decide to get a specialist to undertake the following:
- Audit the website;
- Pay someone to fix all the tech issues;
- Find the time to review and change website content – images, readability etc.;
- Pay for a re-test;
- Pay to have an accessibility statement written;
- Commit to an annual re-test.
All of which is all very commendable, and expensive. Often prohibitively so.
This got me thinking about an – admittedly ‘out of the box’ - alternative.
The accessibility browser
To alleviate the seemingly never-ending task of enforcing code and content changes manually across billions of pages of content on millions of websites, what if it were possible for the browser to aggregate and adapt content from any website into a WCAG-compliant form?
In coding terms, this is known as ‘parsing’ – taking text or other data and converting it to another format.
An ‘accessibility browser’ is not a new idea, and most browsers - including those on phones - now have some great accessibility features such as VoiceOver on IoS devices or other read-aloud features.
However, most of those features, and those of accessibility hardware such as braille keyboards, and navigation aids like mouth controllers, still rely on page content being fully WCAG-compliant for best performance.
There are a number of existing browser extensions (for example Accessibly) that do a pretty good job of addressing some common accessibility issues. This is also good progress but due to the nature of people’s disabilities, ‘edge cases’ (the cases that are less often encountered) are not always catered for.
But what if a browser (potentially in conjunction with a hardware ‘standard’) could negate the need for sites and services to be WCAG compliant because the compliance could be assured within the browser itself? Further to that, the browser could evolve the experience to align and support the user’s specific disability via direct feedback (prompts) given verbally or in text form.
There are approximately one billion non-WCAG-compliant websites worldwide. An 'accessibility browser' could parse content from websites and convert it to be WCAG compliant using AI. In addition, by analysing user feedback, it could build in continuously improving experience, and take account of hardware interoperability, for a user's specific disability
How AI can assist
AI could theoretically support the development and personalisation of such a browser that could do two things:
- Create WCAG-compliant pages from any data or content source by parsing it into a WCAG-compliant shell.
- Build a consensual evolving browser profile, tailored to the real-world disability experience and hardware choices of the user.
AI can be trained on the WCAG criteria, and it can be adapted one step further in that it could be trained to ‘empathise’ with a user with disabilities. The browser could be able to respond to users’ preferences and usage, then create new adaptations and behaviours in response to user feedback.
Imagine a browser that learns about its user and presents a view of the world to that user which is easy to navigate and understand because it has considered the way that person perceives and uses technology.
AI is very good at performing specific, repetitive processing - it is already possible to do things like text rewrites to improve readability – this could have immediate benefits for people with cognitive impairments. Users with visual impairment could enjoy high contrast, large text design on every website as the content could be reformatted for their preference.
An AI-supported browser would also be able to offer more ‘assistance’ and predict probable responses, preferences and statements for users to review, saving time on complex form-filling.
There are quite a few challenges that might hamstring the development of a ‘WCAG browser’ though.
Talking to one of our PDMS developers (thanks Jonny) the use of JavaScript on websites could be a challenge. This language is used to manipulate content across many websites and may interfere with the ability to convert (parse) content into different formats.
Another problem is privacy and IP. ‘Scraping’ content from websites to process it is a somewhat dubious practice, although there could be legal and technical ways to approach this. Any cached content could be simply deleted at the end of the browser session for example.
Additionally, users with disabilities might not want to disclose sensitive personal information to their browser.
Problems aside though – an AI-powered accessibility browser could have enormous benefits to both parties, making life so much easier for people with a range of disabilities, and also supporting organisations that may struggle to meet ever-changing industry standards by providing another route to inclusivity.
By moving the onus of compliance away from website owners to the software used to visit them, millions of partially compliant websites and services could effectively become accessible overnight. This could be digital – and more far-reaching - equivalent to the use of hearing loop systems, which are already widely used in public services and banking.
My view on the undisputable power of AI is that it should only be used ‘for good’ – and what better place to start than reinstating the democratic and inclusive environment of the internet.
To learn more about how AI can support your digital services, check out our blog on AI and UX Design
Topics
Author
