Exploring AI Opportunities for Digital Accessibility

This article presents a balanced perspective on the intersection of artificial intelligence and digital accessibility, serving as a complement to Joe Dolson’s skeptical view of AI. Written by a Microsoft accessibility innovation strategist, the piece acknowledges significant risks and current limitations of AI, such as poor context awareness in image analysis and the potential for hallucinations. However, it highlights constructive opportunities where AI can aid people with disabilities. Key areas discussed include using AI to generate initial drafts for alternative text, thereby supporting human-in-the-loop authoring, and training models to distinguish between decorative and contextual images to improve efficiency. The author also envisions a future where onboard AI models allow users to interactively query complex data visualizations, such as charts and graphs, to extract specific insights. While admitting that current computer-vision models are imperfect, the article argues that with proper training and human oversight, AI can transition from a potentially harmful tool to one that creates meaningful, inclusive differences for users with disabilities.