You reminded me of a failed experiment I did a while back with music tags. I still purchase music monthly (Bandcamp when possible), and I always tag music. I have a bunch of multi-value tags including 'instruments' and 'moods'. It works really well with a huge collection. I can quickly find songs similar to the one I'm listening to, or generate playlists.

Anyway, a few years ago I thought I could speed up the tagging process by pulling tags from Last FM and filtering out the shite by only including tags that matched an instrument name or mood. Luckily I backed up before running the script on my whole collection. The data was absolute garbage.

It reminded me of when I went through and ripped my CD collection. I'd estimate that almost a third of the album data was complete trash, and there were a shockingly bad number of spelling errors. I couldn't help but wonder if someone was deliberately poisoning data.

Slashdot tried to solve this problem by making mod points limited and requiring you to give a reason for moderating a comment. Then there was a metamoderation system where you decided if the moderation someone gave was fair or not. If someone consistently doesn't use the moderation system correctly (and the metamoderation system is successful in identifying this), they don't get mod points anymore.

It didn't work perfectly (obviously the site is still highly opinionated in favor of certain things, like Linux and free software), but it also wasn't the smooth-brained hivemind that Reddit comments sections are. I can't think of another site that's even tried to come up with a better moderation system than simple up/down votes. At best, there are some behind-the-scenes algorithms that decide how exactly to prioritize comments based on the voting, but that's derived from popularity as well. An immeasurable side effect of this is that many heterodox comments simply aren't posted at all since people know they're going to be downvoted to oblivion anyway.