TLDR: Accessibility expert Léonie Watson has delivered a critical yet balanced perspective on Artificial Intelligence, highlighting its potential benefits for accessibility while cautioning against its ‘bollocks’ or misleading aspects. She emphasized the need for careful implementation, verification of AI-generated code, and robust regulation to ensure AI truly serves inclusive design.
In a recent address, prominent accessibility expert Léonie Watson offered a nuanced examination of Artificial Intelligence’s role in fostering inclusive design, distinguishing between its genuine advantages, inherent challenges, and outright ‘bollocks‘ – or misleading claims. Watson’s insights aim to cut through the pervasive hype surrounding AI, focusing on factual applications and necessary precautions.
Watson underscored several ‘good‘ applications of AI in accessibility. She cited speech recognition technologies, such as Dragon Naturally Speaking, as early and effective examples of AI’s power to convert spoken language into text, significantly aiding individuals who find conventional input methods challenging. Furthermore, AI proves beneficial in generating concise summaries of lengthy technical documents, making complex information more accessible. For developers, AI can assist in rapid prototyping, jogging memory, and even producing code for unfamiliar concepts, thereby streamlining the development of accessible interfaces.
However, Watson also highlighted the ‘bad‘ and ‘bollocks‘ aspects of AI in the accessibility landscape. A significant concern she raised pertains to ‘overlays‘ – third-party JavaScript solutions marketed as magical fixes for website accessibility. Watson warned against these tools, citing issues with their marketing and their failure to genuinely make sites accessible, often leading to legal challenges for companies that rely on them. She specifically mentioned companies like AudioEye and AccessiBe being called out for such practices.
Regarding generative AI, particularly for code creation, Watson advised extreme caution, labeling this area as part of the ‘bollocks.’ She presented findings from tests where leading AI models, including ChatGPT, Gemini, and Fix My Code, failed to consistently produce truly accessible code. While these tools might include ARIA attributes, they often lacked crucial elements like proper keyboard support, rendering the code inaccessible for screen readers. Watson stressed the imperative to always verify AI-generated code with trusted, human-led sources, as AI ‘hallucinations‘ can lead to significant errors.
Also Read:
- AI Governance Imperative to Prevent Societal Trust Erosion, Warns AI for Change Founder
- Australian Shadow Minister Cautions Against Excessive AI Regulation Amidst Copyright Debate
Concluding her address, Watson issued a strong call to action, urging governments to establish robust policies, laws, and regulations. She argued that only through such frameworks can companies developing these powerful AI tools be held accountable and guided towards creating solutions that genuinely contribute to a more accessible digital world.
