December 16, 2019
Introduction – What is Artificial Intelligence and How Does it Relate to Accessibility?
Artificial Intelligence (AI) has been at or near the center of virtually every discussion surrounding innovative technologies for the better part of the last decade. The term AI loosely refers to “intelligence” demonstrated by technology comparable to the natural intelligence of humans. It is closely related with the concept of machine learning, wherein computer devices run algorithms, or sequences of instructions, to manipulate sets of data, and gain insights from that data. Although AI has not fully matured or reached its potential, it is already upending the way that humans use computers and other technology. AI either will impact or is already impacting everything from the way we work, socialize, travel, shop, and consume media.
Some of the most promising aspects of the artificial intelligence revolution are the ways that it will improve accessibility in information and communication technology (ICT) for people with disabilities (PWD). Digital accessibility awareness has risen in recent years, but most organizations are still struggling to strategize and implement accessibility changes to their websites, videos and documents, even if they are aware of the requirements. AI has the power to augment accessibility on websites, apps and connected devices in a way that has never been done before.
Below, we have summarized several of the most interesting AI-based features and tools for web accessibility and provided insight into the implications and limitations of these AI-powered technological advancements.
Some of the most visible ways that AI is affecting web accessibility today is through features both on connected devices and within software, websites and web applications. Examples of such features, and how they augment accessibility on the web, include:
- Image Recognition and Automatic Alt Text (e.g. Facebook Photos), which dynamically creates alternative text for images by analyzing previous data and images and recognizing similar images and tagged words.
- Voice Recognition, automatic captions and translations (e.g. Youtube auto-caption), which analyze video and audio content to transcribe the audio through natural language processing AI.
- Text Processing & Adaptation (eg. adjusting text based on reading level) which adapts text content based on an individual user’s needs through AI that changes the structure of the content based on the parameters of the individual reader.
- Voice Control (eg. Amazon Alexa) which allows users to interact with both devices and the web content they browse on them utilizing voice as an input. This is also done through natural language processing AI.
- Affective Computing (eg. Microsoft Emotion API) which is an emerging branch of AI-powered features that recognize the emotional state of a user through facial recognition and other inputs to adapt user interfaces to meet the needs of the individual.
In addition to the features above which impact end user accessibility within devices, websites and applications through artificial intelligence, there are innovative tools powered by AI that are impacting the process of evaluating accessibility on websites, video content and web documents, and fixing the issues found. Some examples include:
- Automated Accessibility Scanners (eg. Microsoft Accessibility Checker) which automatically scan through web pages or web documents and note any accessibility issues found. AI has drastically improved these scanners through machine learning to identify issues quicker by recognizing specific elements on a page (eg. button) or content structure issues.
- Automated Accessibility Remediation (eg. AccessiBe) which automatically remediate issues on a website to comply with WCAG success criteria by automatically monitoring and dynamically updating the site for accessibility based on any changes made. The AI-powered systems learn from individual users, other website data and the templates and pages on a website to make the accessibility fixes quicker and with higher accuracy through code-level changes.
The various features and tools that exist today utilizing Artificial Intelligence to improve accessibility outcomes are already making a big impact in the lives of web users with disabilities, and in the work of businesses trying to make their own web content accessible. Here are some of the largest implications for users and businesses being driven by these AI-powered innovations:
- More accessibility is being baked into devices and sites, which is creating more control for both users with and without disabilities. Features like autocaption and autotranslate are helping users who speak different native languages and/or who are hard-of-hearing or deaf to communicate. Tools like automated accessibility overlays are building in accessibility to websites as well, helping both people with disabilities and the aging population through contrast, button size and placement, and navigation controls.
- Both the level of effort that companies need to put into making their websites accessible and the cost to undertake such initiatives are being driven down. AI-powered accessibility remediation tools reduce the time, effort and cost to Implement scaled accessibility changes, and are consequently removing or reducing a large barrier for most organizations to comply.
- Web authoring too, is becoming more accessible for people with disabilities, and Content Management Systems and web design tools are becoming usable alongside business-and-customer-facing websites. This could empower a whole generation of entrepreneurs and digital marketers with disabilities to own and operate websites that are both accessible to create and use.
One exciting thing to note is that these implications are only the ones that can be seen TODAY. The advances in both AI and Accessibility on the horizon will likely create even more impact for users with disabilities and organizations trying to make their web content accessible. The growth (as with all things technological) should be exponential.
While we must look eagerly at the advances and subsequent impact that AI-powered tools and features on websites are creating for accessibility, we must also look at where the technology currently falls short . At present, AI has structural and functional limitations that need to be evaluated, understood and addressed, including:
- AI is not as contextually intelligent as humans, and thus cannot yet replace human data input or description. Current tools and features for auto-captioning for example, are not yet nearly as accurate as human transcribers. Auto-generated alt text is often not described perfectly accurately. These things will get better, but for now, they do not replace the human-centered accessibility tactics that have been traditionally relied on.
- There is deep bias within systems, algorithms and data used in AI-powered innovations, which skews the results of the algorithms, at times in harmful ways. The algorithms that power the “learning” element of AI draw from existing data that might be flawed. For example, if an AI-powered job board is rejecting candidates based on traits of current employees, and the hiring company has few employees with disabilities, the data leading to new hires might be biased against applicants with disabilities.
- The sensitivity and security of data, especially user data, has to be accounted for and AI doesn’t yet fully address this. If an AI-powered tool for making a website accessible has access to a company’s sensitive medical or financial records, and the tool running that data isn’t as secure, it could create a very real risk for data breach and exposure.
We are living in the most exciting time for technological advancement since the advent of the Internet, and Artificial Intelligence is a big driver of the innovations of the day. The various tools and features both that already exist today and are being worked on for the future have the power to transform the landscape of digital accessibility, making it both more effective for users with disabilities and easier to accomplish for companies. The current and future implications of these advancements are wide-reaching and important, but we must also carefully examine the risk factors and limitations of these technological breakthroughs.
- Abou-Zahra, Shadi, Judy Brewer, and Michael Cooper. “Artificial Intelligence (AI) for Web Accessibility: Is Conformance Evaluation a Way Forward?” Web4All 2018, 23-25 April, 2018, Lyon, France, ACM, 2018. © 2018 Authors
Chief Information Officer, The Viscardi Center
Michael oversees all aspects of technology at The Viscardi Center, where he implements and innovates accessibility for students, staff, and faculty.