> Let me re-phrase that then to the main cause of Black over-representation in prison being bias. Because yes, one can always find some bias, but does it contri

> Let me re-phrase that then to the main cause of Black over-representation in prison being bias. Because yes, one can always find some bias, but does it contribute to 1-10% of the outcome, or 80-90%? I think the answer is "depends on how deep you look". Black Americans largely being the disadvantaged underclass stems from slavery and segregation (not genetic differences) for instance, with identity/culture being formed around those circumstances too. > Although one could argue that the bias goes in the other direction - i.e. that Blacks are favored by the legal system, since, estimating based on fraction of homicide perpetrators (56%), they are under-represented in prison at 32%. Of course this is just that - an estimation, and it could be simply that they genuinely commit fewer other crimes. 1. As you hedged, it does indeed appear that it's higher for homicides than other crimes (and there are comparatively very few homicides). In 2023 the average for arrests across all crimes was 30%[0]. 2. I'm not sure whether or not you're seeing the main point I'm trying to convey. The "proportional" prisoner count you take as the neutral unbiased point would mean that a Black person would be by default more suspected of being a criminal, more likely to be searched, more likely to falsely arrested, receive harsher sentences for the same crime, etc. - which is what many experience as bias. It's a bit of an unintuitive statistical fact about maximizing predictive accuracy. > In your place I would choose a different example than college admissions, however, because, at least for the Ivy League, whites are the most under-represented group by a significant margin, despite doing comparatively well in tests: The purpose of the hypothetical was to show how something like "asking for parent's income on admission form, then treating low-income as a negative factor" would be obviously unjust, and how instead even discriminating in the opposite direction (with the neutral point being uniform/blind, not proportional) could arguably be justifiable. College admissions seems a good example for this because it's where affirmative action has been practiced and hotly debated - I don't deny its existence. [0]: https://cde.ucr.cjis.gov/LATEST/webapp/# https://cde.ucr.cjis.gov/LATEST/webapp/#