Artificial Intelligence: a boon for accessibility?

The accessibility of digital services is a topic as large as it is interdisciplinary. It brings together very different fields, ranging from design to paramedic, legal to technological… A social theme in a nutshell. The development of artificial intelligence solutions could be a boon for the emergence of more accessible digital services. However, the equation is not so simple, the democratization coefficient not so obvious. Because behind artificial intelligence, there is its shaping by humans. However, by now, people have proven that accessibility is not necessarily their priority. Only so far?

A contribution by Mario Arias, Microsoft Business Applications & AI MVP and Power Platform Architect at Prodware

From theory to practice

Artificial intelligence can intervene in all fields, in all areas, for all subjects. Its impact zone is too large to be summed up as an acceleration effect of a single element. The correct question is therefore not theoretical (CAN AI make digital solutions more accessible?), but practical (Will AI solutions have real benefit?). And immediately, it's much less obvious.

First, because many tools that promote accessibility already exist and not all of them are, far from it, permanently installed. How many images posted on the Internet, conveying information, have alternative text? How many pages containing text remain readable when the text size is increased by 200%? How many emoji in a mobile app benefit from a text alternative? However, all these criteria appear in the RGAA technical method, the European reference standard for the analysis of the accessibility of a digital service. This means that despite the 12 million people with disabilities in France, the market opportunity is not that necessary.

Then, because to really enforce a solution on all digital players, the law needs to get involved. Here, as in many other issues, arises the second great dilemma of accessibility in digital services: that which opposes its progress and universality, the struggle for innovation and the respect of a legal and regulatory framework that regulates it and therefore slows it down. below. Legislating an emerging issue takes time and is not always well received by actors in the field, who sometimes experience the law as a useless barrier rather than a necessary framework for the expression of freedom. This was true for the Internet, it is true for artificial intelligence, so it is likely to be true for the accessibility of digital services as well. The problem is therefore less technological than decidedly human and social.

From identification to anonymization

The issue of accessibility of digital services goes hand in hand with another contemporary digital issue: data protection. A critical issue of today's web, personal data is often viewed as private stock, an asset to be protected. However, if many digital services depend on the (consensual) use of personal data, so does accessibility. To know that a customer should receive a measure of enhanced access when using its services, a business needs to know that person. Consent in a stated way of course, but doesn't that run counter to the oft-stated desire to make personal data a quasi-protected issue?

The challenge of service optimization is closely related to that of data protection. The famous formula “if it's free, you are the product” has survived, but another concept succeeds it: if I don't give anything of myself, how can I financially justify a company that offers me a personalized service? The solution could then be to return to a paid system which would, in effect, limit the concept of accessibility as it would then be filtered by financial ability. We are already seeing this on some websites which, following GDPR regulations, offer their visitors the option to pay to remain anonymous while browsing. And it's a safe bet that among the millions of daily visits to a site that offers this option, few visitors choose the exact solution. The “free” habit is quickly formed and takes its toll.

From data to synthesis

As with all issues raised by the development of artificial intelligence, its ability to improve the accessibility of digital services is linked to an equation that constitutes artificial intelligence algorithms. The solutions it proposes are by definition based on accumulated knowledge. ChatGPT version 3.5 specifies, for example, that its last update dates back to January 2022. In the digital world, two years is an eternity. If the machine learning process is autonomous and powered by the data that exists (where services are often not all inclusive and accessible), it can never automatically become comprehensive and accessible.

Therefore, automation is both an advantage and a limitation of artificial intelligence in terms of the accessibility of digital services. The issue is larger, more complex and involves more factors than the simple criteria of technical progress. Because better accessibility requires regulation, standardization, the will of a few that will surely spread, but also simplification of processes and technological solutions that allow this accessibility. Technology is a fundamental part of the solution here, but people are probably the main cause of the problem. Therefore, through its intervention AI will become a more favorable tool for more accessible digital services.

At present, artificial intelligence is therefore not essentially a tool that facilitates the accessibility of digital services. On the other hand, depending on its use, its evolution, its implementation, it has the potential to change all internet navigation and all digital services, making them more adaptable, more comprehensive, more accessible. Accessibility is thus one of the frontiers of artificial intelligence COULD redefine. The parole still stands, for now.

Read also : Artificial intelligence, the human resource

Leave a Reply

Your email address will not be published. Required fields are marked *