I have a question: what is the most effective way for a social platform to prevent harassment? Every relevant social platform has a harassment problem. People are mean; relative anonymity & tribalism make it worse.<p>This article offers no new solutions. Ramp up reporting and moderation? Instagram services a billion people. Even if Facebook multiplied their moderation team by a 100, they still would be drowning in information.<p>Considering the scale of this issue, I can see why these engineers are deferring to their idée fixe - machine learning. I guess I'm just skeptical. What are they going to do? Have a team (arbitrarily?) identify harassment by eye and use it as training data? That sounds like a draconian recipe for disaster. At the very least, its insufficient. Identifying harassment is not the same as identifying a street sign; it often requires the context of complex, offline interactions or shared info between two people. A friend asking "When can I see you again?" is fine. A stalker asking the same thing? Not so much.<p>I really want harassment to be solved. I hate that the internet is like this. But this problem would exist even without platform negligence.