I suspect that there is indeed no winning, and that under such circumstances every possible course of action is likely to be perceived as racist by some number

I suspect that there is indeed no winning, and that under such circumstances every possible course of action is likely to be perceived as racist by some number of people. This is because the population at large doesn't seem to be able to agree on what the word actually means (https://slatestarcodex.com/2017/06/21/against-murderism/ https://slatestarcodex.com/2017/06/21/against-murderism/). Regarding ML based predictive policing specifically, it seems like a spectacularly bad (but efficient and statistically effective!) idea to naively apply a machine learning based classification approach to such a problem.