It's not common usage, but we're supposedly talking about unspoken and unconscious bias, so why aren't any vague linguistic or metaphorical associations fair game? The argument is precisely that you _don't_ directly think of red as related to Native Americans, so it doesn't make much sense for you to say that people don't think about it like that in response. Because well, yeah, the question is whether they _unthinkingly_ make the association.

Let me paraphrase this argument, but correct me if you have a different understanding:

  - 'black' and 'white' have some racial associative strength, x, which is sufficient to cause bias in other contexts.  I.e. x > r for some threshold r.
  - 'red' has racial associative strength kx, for some k < 1.    I think we agree that k < 1, since the association is less strong.  Where we differ is I am (hypothetically) saying kx > r, still, whereas you (seemingly) think kx < r.

This is a strange argument because we've never actually established the relative values of x or r. Even if we assume the first point is true, it tells us nothing about the second, because kx might still be below or above the threshold.

In fact, we _haven't_ demonstrated the first point anyway, so it's just compounding an already hand-wavey explanation of how things work. If someone can assume x > r with little evidence, why can't I assume kx > r? You might have priors on the size of k because you think 'red' is less strongly associated with race, but we know nothing about x or r, so it's pretty irrelevant. If you can hand wave the first point, you can hand wave the second. As I did.

I'd rather there was no hand-waving. But if that's the game we're playing...