True but as @KirkegaardEmil noted, group membership IS an individual characteristic http://emilkirkegaard.dk/en/?p=5788 https://twitter.com/grossmanjoshua/status/747015618579283968 …
-
The right recommendation to give is: use strong individualizing information if one wants group membership to be irrelevant.
And: if group membership predicts because it is associated with some trait, use individual measures of that trait to neutralize.
