Tweeted By @rctatman
debiasing & similar approaches are about *mitigating* the impact of specific latent variables that are relevant to the current application and that's genuinely the best you can hope for
— Rachael Tatman (@rctatman) August 25, 2021