Friday, March 18, 2016

Computer Bias or Human Bias?

NPR has a fascinating story on, Can Computer Programs Be Racist And Sexist?  The concern is that even though learning algorithms, such as used by recommendation engines (like Amazon's) or search engines (like Google's) start off as completely unbiased they may evolve to reflect the bias of the people using them.

A Harvard study cited found, "found that when someone searched in Google for a name normally associated with a person of African-American descent, an ad for a company that finds criminal records was more likely to turn up." Why? "Because people tended to click on the ad topic that suggested that that person had been arrested when the name was African-American, the algorithm learned the racism of the search users and then reinforced it by showing that more often."

Regarding gender bias the article states that studies have show that women are shown more lower-paying job ads then men. This is due to the fact that women click on the lower-paying jobs less than the higher paying ones. Why would woman do this? You might say that our sexist society teaches woman that only lower paying jobs are open to them. Or you could say that out sexist society teaches women that men will work at the more dangerous and stressful jobs in order to protect and make money for women. No doubt both are partly true.

I have some background in technology and know that it would be nearly impossible to eliminate such bias in learning algorithms because you would have to purposefully introduce a counter-bias to correct for it. And, of course, everyone would have a different opinion on exactly what that counter-bias should be.

So if we cannot programmatically correct for bias what can we do? We can think. We can use knowledge to understand. If you know these systems may be biased you can take that into account. If you are looking for a job in a profession dominated by members of a different gender, know that you might have to look a bit harder.

Bias is endemic. Learn to recognize it. For example, this week I finished a novel that was recently written and authored by a woman. In it, a couple has a hole in the wall of their house. The wife complains that the husband hasn't fixed it. The implicit bias is that somehow the wife is unable to fix it herself. That is sexism. Recognizing and being aware of bias is all its forms is the key to overcoming it.

No comments:

Post a Comment