Artificial Intelligence’s White Guy Problem
nytimes.com
nytimes.com
There was no suggestion that these systems were built to intentionally produce certain outcomes for particular races or sexes, which leads me to believe that these outcomes are "pure" effects of the ML algorithms that use race or gender as inputs.
Assuming this, is the author suggesting we -should- be coding racism and sexism into these systems to achieve "fair" outcomes? She didn't come out and say it, but I'm not sure how else to read the article. This sounds absurd to me, and would be acting against what machine learning is supposed to give us.
I'm assuming women are shown less ads for highly paid jobs because other ads are, in general, more profitable to show (and by extension more relevant on average). One thing I like about AI is that it doesn't suffer from biases in the same way we do. Why would we want to make it do that?
> A very serious example was revealed in an investigation published last month by ProPublica. It found that widely used software that assessed the risk of recidivism in criminals was twice as likely to mistakenly flag black defendants as being at a higher risk of committing future crimes. It was also twice as likely to incorrectly flag white defendants as low risk.
Is skin color even something that is fed into the algorithm? It could just as well be other factors like number of convictions or socioeconomic background. Just because it produces an output that worse for black people does not make it racist.
Aside from that, do we want it to be fair or do we want it to be right?
Let me see some code that alludes to racism or sexism and not just an analytical algorithm that says yes, this guy that has a long history of being arrested has a good chance of being arrested again.
I highly doubt that biometrics are being used to assess these things when proven data is exponentially more reliable.