> When you optimize for one metric, other metrics fall. This is very true. People often assume that things like machine learning or AI won't have racial biases

> When you optimize for one metric, other metrics fall. This is very true. People often assume that things like machine learning or AI won't have racial biases, but then we find out time and time again that our algorithms are in fact biased. It's not that the bias is intentional (even human bias is rarely intentional), it's that by focusing on metrics, the metric of racial equality often suffers. I'm not sure why you were down voted for pointing this out. This is a key concept to understand if we care about creating a more equal society in an age of big data and machine learning. Google did some research in this area: https://research.google.com/bigpicture/attacking-discrimination-in-ml/ https://research.google.com/bigpicture/attacking-discriminat... The TL;DR version is that in order to eliminate bias, you have to be aware of biases as metrics and optimize to eliminate them.