Bias in algorithm

Aidan

Women's Studies

Bias in algorithm


Algorithms help with so much in the age of the internet. Almost everything has algorithms working to enhance it: Google, Youtube, social media and more. But it is possible for algorithms to become bias. Essentially this is caused by a lack of variation in trials. So for example, many facial recognition algorithms struggle with women, especially with black women. This is because of a lack of trails with black women: meaning the algorithm isn't used to black women so it doesn't recognise them. One example I've experienced personally is the face filters on Snapchat. It works much more often when I take off my glasses, with my glasses it's much less accurate.

Now, of course, the filters on snapchat are no big deal. But these flaws in algorithms can have big repercussions. So much of today's world relies on the internet if algorithms can become bias against, say, black women it can very seriously throw a wrench in their lives.

It can affect things like applying to college and sending in a resume to a workplace. That workplace might never even see the resume if the algorithm deems it not worthwhile. What factors it might consider deciding whether or not the resume or college application is seen is relatively unknown.

The solution to this is having a broader variation of sex and skin tone represented in the world of programming these algorithms.