01/03/2022
The AIs we make are only as good as the society we make them in. So what do we do when that society is ableist or racist?
When an AI trains on data, it implicitly picks up on the biases of humans. Controlling for the ability to pay rent, African Americans in the United States experience disparate treatment in the housing market 50% of the time compared to white Americans. So the solution is AI algorithms that can assign housing fairly right? Well, not so fast.
During all parts of an AI’s development, from construction to deployment, there are implicit biases at work. AI algorithms have been found to have overcharged minority buyers by millions of dollars.
As we move forward into the future, we have to ask ourselves how to better construct AI for a kinder, fairer world.
Image credit: Pedro Saleiro