AI and Disability: Progress or Exclusion?
summary
The talk examines how current AI systems often amplify biases against disabled people rather than solve accessibility problems, citing image generators that repeatedly show wheelchair-only scenarios and stereotyped depictions (e.g., many Midjourney images of depressed white men). It reviews practical accessibility tools like Seeing AI and Be My Eyes that deliver real benefits but have mixed reliability, warns that errors (such as from counselling bots or automatic captions that fail on atypical voices and technical jargon) can harm users, and notes differing model behaviours—Mistral gives neutral explanations while ChatGPT tends to infantilize disability.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 3:12Disability statistics in France
- 5:06AI bias in representation
- 9:03AI development and data bias
- 12:13Production and maintenance
- 16:00AI impact on daily life
- 23:44Ethical risks and dangers
- 27:18AI and web accessibility
- 35:12Inclusivity in Tech
- 39:23Conclusion and call to action
- 40:07Q&A
related talks
43mAI and Disability: Progress or Exclusion?
34mAI and Accessibility: the Good, the Bad, and the Bollocks
34mArtificial Intelligence and Disability: A Brief Analysis of Some Concrete Cases
30mAI and Accessibility: the Good, the Bad, and the Bollocks
32mAI in digital accessibility: challenges, benefits and risks
17mWhat AI Means for the Future of Accessibility and Disability Aides
52mPreparing Accessibility for the Future and the Present
55mAm AI Accessible? Present and Future Intersections of Artificial Intelligence and Accessibility