there's at least one major reason to be skeptical of that result, and they mention it in their limitations section. what they did was create a lexicon of hate

there's at least one major reason to be skeptical of that result, and they mention it in their limitations section. what they did was create a lexicon of hate speech terms from two large subs that were banned. they then counted the frequency of those terms in other subs after the ban. they found that usage of those terms dropped substantially, and concluded that the bans were effective at reducing hate speech. if you're familiar with the dynamics of these sorts of subs, the problem with this approach should be fairly obvious. these subs tend to develop their own set of specific terms/memes (hate-related or otherwise). it may be the case that the bans were effective at reducing hate speech across the whole site. but it's also possible that the same people are still posting the same stuff coded differently. this study is far from the final word on the matter.