Ask HN: What kind of steps could Twitter do to reduce harassment?

In broad strokes, what could they do? What technological solutions are there? How much of it would involve human review?