Could AI compromise digital access for people with disabilities?

In light of the ongoing Paralympic Games, which showcase the incredible achievements of athletes with disabilities from around the globe, it seems like the perfect moment to delve into the impact of AI technologies on digital accessibility - enabling people with disabilities to navigate websites and applications comfortably and efficiently. While Generative AI introduces groundbreaking advancements, it also raises critical questions regarding its effect on people with disabilities. Here, I’ll explore the challenges and opportunities this technology presents and suggest how we can ensure continuous improvement in digital accessibility.

Yossi Synett, Chief Scientist at Evinced

The Development of AI and Its Impact on Digital Accessibility

Generative AI is advancing rapidly and offers numerous benefits across various domains, significantly enhancing the quality of life for individuals with physical and cognitive impairments. For instance, advanced speech recognition systems empower people with physical disabilities to control devices using voice commands. Generative AI also assists those with hearing impairments by generating automatic subtitles for videos, making important content more accessible. Additionally, personalized machine learning tools can offer tailored recommendations and help individuals with cognitive impairments navigate complex information systems more simply. These tools not only improve accessibility but also foster independence, enabling a more active and equal participation in the digital world.

The Potential Risks of AI in Digital Accessibility

However, despite its many contributions, AI also presents accessibility challenges for some individuals. One major concern is the risk of inherent biases in AI models, stemming from the data they are trained on and the biases of their developers. These biases can exacerbate existing discrimination. For example, AI models trained on historical recruitment data might favor candidates without physical disabilities, potentially harming the success rates of individuals with disabilities in the hiring process. Additionally, Generative AI tools for creating code might lead to interfaces that aren’t fully accessible, as current user interfaces (which these tools are trained on) often lack comprehensive accessibility features. Consequently, code generated by large language models might contain accessibility issues that require human intervention to identify and resolve.

Limitations of Generative AI Solutions for Accessibility

Since the introduction of ChatGPT, public awareness of the potential of large language models has surged, generating excitement about their use in accessibility. Experiments are underway to use generative models that combine images and text to create alternative text for images. While these models offer promising solutions, creating effective text descriptions requires an understanding of the context and functionality of the image—a complex challenge that has yet to be fully addressed.

How Gen AI Can Improve Digital Accessibility if Implemented Correctly

Despite the risks and challenges, Gen AI can be a powerful tool for enhancing digital accessibility for individuals with disabilities if applied correctly. Integrating AI with assistive technologies such as screen readers, text magnifiers, and eye-tracking can significantly improve user experience and facilitate easier access to digital information. Moreover, AI can help developers and designers better understand the unique needs of people with disabilities and integrate accessibility solutions into their development processes. The key lies in controlled use and adherence to ethical standards to ensure AI tools serve all users fairly and inclusively.

Embedding digital accessibility in the early stages of website or app development is not only the socially responsible thing to do but also a smart business move that can lead to significant savings. There is a growing trend of enforcing accessibility in websites and apps, and addressing accessibility issues post-development can be costly. Moreover, code written with accessibility in mind is often more readable and organized, making it easier for developers to understand and work with quickly.

While AI poses substantial challenges in the realm of digital accessibility, it also has tremendous potential to improve the current situation if used wisely. Technology developers, regulators, and communities of people with disabilities must collaborate to ensure that these innovative tools enhance accessibility rather than hinder it. Only by doing so can we harness the potential of new technologies to create a more accessible and equitable digital environment.

Yossi Synett, Chief Scientist at Evinced, a company developing tools for developers to implement accessibility at scale in web and mobile applications.