Attempting to censor nodes in a classifier’s Bayesian network will strictly result in some combination of A) the classifier routing around the damage and B) the classifier producing worse results. Either way, we’ve accomplished nothing. The linked paper, on the other hand, still allows the classifier to be accurate without giving incorrect results across protected categories.
A vastly more reasonable standard, but easy to confuse without stats background, is e.g. http://ttic.uchicago.edu/~nati/Publications/HardtPriceSrebro...
Attempting to censor nodes in a classifier’s Bayesian network will strictly result in some combination of A) the classifier routing around the damage and B) the classifier producing worse results. Either way, we’ve accomplished nothing. The linked paper, on the other hand, still allows the classifier to be accurate without giving incorrect results across protected categories.