Affordances for Me, but Not for Thee (werd.io)

A recent discussion highlights the paradoxical development of web accessibility for artificial intelligence (AI) compared to that for disabled users. While the modern web has been primarily designed with sighted users in mind, recent advancements prioritize accessibility features that benefit large language models (LLMs), a shift that many in the tech community are embracing with urgency. This contrasts sharply with the historical neglect of accessibility for blind and low-vision users, who have long advocated for structured, concise content to improve their web experience. Interestingly, 95% of websites still possess accessibility flaws, and efforts to enhance these features for humans have often been sidelined. The significance of this shift lies in the stark realization that companies are now allocating resources to develop clear specifications and documentation for LLMs—tasks that have previously been treated as afterthoughts. This prioritization raises ethical concerns, as AI affordances appear to cater to those in power rather than addressing the needs of vulnerable populations. Ultimately, while the adaptations for LLMs could improve productivity and ease of use for organizations, they risk perpetuating a dystopian narrative where the needs of disabled individuals are overlooked in favor of optimizing systems for AI. This trend underscores a critical need for the AI/ML community to reconcile its advancements with genuine accessibility efforts for all users.