> I would consider humans manually reporting tweets to be much more reliable than AI flagging, especially given the bizarre false positive results we've seen from Twitter AI flagging.
Sure. However, in this case Twitter is the one with the AI and the one deciding whether to trust a source. The case for child porn is virtually binary and even where it isn't, they're going to lean on the side of not allowing something. The case for harassment is reported by unknown humans who are probably personally involved, and whether something is or is not harassment is not always black and white. From Twitter's point of view, it probably makes more sense to trust their own AI.
Here, the author hasn't provided specific examples of alleged harassment, so readers can't even make a judgment call.
Sure. However, in this case Twitter is the one with the AI and the one deciding whether to trust a source. The case for child porn is virtually binary and even where it isn't, they're going to lean on the side of not allowing something. The case for harassment is reported by unknown humans who are probably personally involved, and whether something is or is not harassment is not always black and white. From Twitter's point of view, it probably makes more sense to trust their own AI.
Here, the author hasn't provided specific examples of alleged harassment, so readers can't even make a judgment call.