That's roughly right in a moral sense, but in a Bayesian sense I don't think it's unreasonable. If race (or background or country of origin, etc.) and some attr

That's roughly right in a moral sense, but in a Bayesian sense I don't think it's unreasonable. If race (or background or country of origin, etc.) and some attribute (education, wealth, etc.) are correlated, and names and race are correlated, it's intuitive that names and that attribute are likely correlated. It's a huge problem when you reverse the process and say, "Despite evidence of having X attribute, I'm going to treat you as though you don't because you belong to a set where X attribute is less common." So overlooking someone's resume because of their race/class/background/etc. is absolutely racist and unreasonable. Having a biased expectation a priori, however, is not necessarily unreasonable. However it may be morally wrong because having and acting on those expectations reduces the opportunities given to members of those groups who are outside the norm, which in turn reduces the group's ability to change those correlations and expectations.