How do you tackle bias in tackling bias? This is not even a machine problem. Not referring to this incident in particular, but rather, thinking ahead. In a hyp

How do you tackle bias in tackling bias? This is not even a machine problem. Not referring to this incident in particular, but rather, thinking ahead. In a hypothetical scenario, if it turns out that immigrants from a certain country are in fact high-risk for certain kinds of fraud, and the AI automatically picks this up, is a human intervention countering its bias an erroneous one? It's telling that so much of racism and stereotyping stems from a human desire for convenience and conformity - it is easier to shape your worldview when you've identified enemies and how to classify them. When you add algorithms with the goal of increasing efficiency into the mix, you create a real possibility that it could profile certain populations out of convenience. Of course, humans do that anyway. And when these humans train AI, they may pass on the bias. But assuming the objective truth does lead to some form of selection discrimination, how does one counter this?