The problem I have with this kind of bias evaluation is fairly straightforward: what if a group of people aligned in a particular way engages in more "untrustworthy" speech than their counterparts? Follows then that a filtering that aimed to curb untrustworthy content would now be disproportionately filtering against their content in general, so now they can point at the filtering and call it biased, and even pass the smell test as far as the numbers are concerned.

Very curious things start happening when we look through lenses like these. It's very difficult to effectively discuss unfortunately too, without getting into any catfights.