I'd love to see a study with real data on this topic too, but at the the most basic level, with up/down voting you can figure out several metrics: - quality (o

I'd love to see a study with real data on this topic too, but at the the most basic level, with up/down voting you can figure out several metrics: - quality (or rather, agreeableness): up - down; - engagement: up + down; - controversy: min(up, down). First one is the most common (HN, reddit, ...). Second famously is (was?) one of the core metrics for youtube recommendation algorithm, and also something pretty close to the only metric upvote-only forums end up with (twitter); same goes for forums where "reactions" are intentionally orthogonal/ambivalent/confusing (facebook). Third feels a bit dangerous for direct use, but might be a good signal to flag for a human moderator to review.