3 ms·
So what I don't understand is that its perfectly able to categorise this content really well, but is perfectly happy to allow it on the network, even though it
by KaiserPro 2y ago
So what I don't understand is that its perfectly able to categorise this content really well, but is perfectly happy to allow it on the network, even though it clearly violates the community guidelines.
Yes moderation at scale is hard but most of the battle is getting decent categorisation.
- r0fl 2y agoThe extreme content gets the most user interaction and leads to higher daily active users Which is a very important metric. As long as the content is not illegal, if it positively affects time on site/app, it stays for as long as it can on the platforms
- KaiserPro 2y ago> The extreme content gets the most user interaction and leads to higher daily active users For some people yeah, but for a lot, it causes the opposite. That's why its quite heavily policed. Its not like thirst traps which are liberally sprinkled in the first 5 minutes of creating an account.
- jon-wood 2y agoThe recommendation algorithm isn't necessarily identifying this as shock content or gore, all it's doing is identifying it as having the hallmarks of something that a certain circle of people engage with, followed by it deciding that everyone engages with it and it should be the default. It's possible that internally it's tagging it as gore, it's also possible that it's simply categorising it as "fast moving video with an abundance of red".
- soco 2y agoI would assume after such a successful AI based categorization, one single human moderator could do wonders deciding whether it's an abundance of red or something else. But yeah, why would we censor the money making?
- AlecSchueler 2y agoThey moderated it up until January of this year. The decision was politically motivated, not technical.
- KaiserPro 2y agothird party fact checking is related but not the same as content categorisation.