I don't think it was possible before today's microtarget-based internet advertising to use demographic correlates to substitute for banned targeting of protected groups. That it's now much more feasible than it's ever been in the past to target protected groups without facially making protected group membership part of the filtering process puts anti-discrimination legislation in a new position. The past 50 years of case law isn't particularly relevant to this new situation: in 1968, it wasn't possible to filter ads based on a suite of seemingly-irrelevant characteristics so as to accurately re-create banned targeting.

There is some precedent (e.g., Griggs) for applying a "disparate impact" standard to anti-discrimination measures. My argument is that this reasoning, if followed rigorously, will essentially prohibit ad targeting altogether, since every meaningful ad targeting criterion will end up targeting different arms of protected groups differently.

I've seen no rebuttal of this point, either from you or from anyone else.

> I find it scary that we’re having a discussion

Now that's a terrifying thought. A discussion!