Online hate is often discussed as if it were only a content-moderation problem: remove the worst posts, suspend the worst users, publish a transparency rep