> Why would you need to do that? You can just have humans review the reports that algorithms generate... At some point, in designing the algorithms, in instruc

> Why would you need to do that? You can just have humans review the reports that algorithms generate... At some point, in designing the algorithms, in instructing the human reviewers, and so on, you still need to agree on what abusive means. So there's no way around that. As to the volume, I agree, one can deploy algorithms as a first filtering step. In this case, I don't usually trust humans to be smarter than algorithms, but I do trust humans to be better at abuse than algorithms are at recognising it—unless the algorithms have so many false positives that they're barely filtering at all.