“Don't generalize” amounts to statistically illiterate virtue-signaling.
It’s rational to lean on base rates in low-information settings and then update a posteriori.
To avoid tribalism, we must then apply the exact same standard when the in-group has the ugly priors.



