Recently, I viewed a talk on AI and accessibility at Columbia University given by my former colleague Gigi Etienne. The presentation focused on creating AI products that everyone can use and gave examples of how these tools affect people with disabilities.
She gives four helpful tips to make sure the AI product you are developing is accessible:
- Use Diverse Data
AI relies on data, so using diverse and inclusive datasets is crucial. Make sure the data you use doesn't discriminate against people with disabilities. For instance, ensure your speech recognition software can understand the voices of people with Down Syndrome.
- Include People with Disabilities
This is a big one. Involve people with disabilities in every stage of product development—from planning to testing. Hire people with various disabilities and test your products with them to get real, valuable feedback. And yes, pay your testers!
- Stay on Top of Regulations
AI regulations are just starting to take shape. The European Union’s AI Act, for example, is the first major regulation and it emphasizes non-discrimination. If you're developing AI, you have a unique insight into potential issues and can help shape these regulations.
- Ensure Accessible User Interfaces
Even if your AI is inclusive, it's all for nothing if the user interface isn't accessible. With 96% of websites still not accessible, we need to keep this front and center.
Gigi reminds us not to repeat the mistakes of the past as we move forward in developing new technology. She encourages development teams and product owners to ask themselves these questions when they are developing a new AI product:
- Is our data inclusive?
- Are we involving people with disabilities in the process?
- Is our final product accessible to everyone?
This is how true progress happens—by ensuring AI benefits everyone. The full video presentation is below.