> Moderators work 12-hour shifts, Frazier explained, and only get one hour off for lunch and two 15-minute breaks.
> Despite the long shifts, moderators often have to watch multiple videos at once, ranging from three to 10, and only review 25 seconds of each video, she said. ByteDance keeps a close eye on the moderators’ performance, the Verge reported, and “heavily punishes any time taken away from watching graphic videos.”
This is awful for the users and the moderators. I really hope there is a better way.
People hate on AI moderation and user identification. But I think a solution is a well-trained AI, with some form of “redress” which requires identity or money and gives manual review, for people who are incorrectly flagged.
Seriously all these moderation jobs sound awful and the people who do them should have free counseling/therapy (that the company doesn’t have access to)
The moderators aren't responsible for filtering shitposts.
They're removing child abuse and violence (including rape, beheading videos, etc), in addition to racist, *-phobic, etc const. I'm guessing the trauma is more due to the former than the latter, but on the other hand there's a different between seeing the occasional bigoted comment and seeing them solidly for 8 hours a day, every day.
Based on what content I see getting taken down I have a hard time believing this.
Any data available showing the type of content and the frequency at which it’s seen?
I handled abuse reports for a ISP in the late 90s and the majority of it was shock/offensive stuff like lemonparty. The real severe things were few and far between even then.
9 comments
[ 0.23 ms ] story [ 34.7 ms ] thread> Despite the long shifts, moderators often have to watch multiple videos at once, ranging from three to 10, and only review 25 seconds of each video, she said. ByteDance keeps a close eye on the moderators’ performance, the Verge reported, and “heavily punishes any time taken away from watching graphic videos.”
This is awful for the users and the moderators. I really hope there is a better way.
People hate on AI moderation and user identification. But I think a solution is a well-trained AI, with some form of “redress” which requires identity or money and gives manual review, for people who are incorrectly flagged.
They're removing child abuse and violence (including rape, beheading videos, etc), in addition to racist, *-phobic, etc const. I'm guessing the trauma is more due to the former than the latter, but on the other hand there's a different between seeing the occasional bigoted comment and seeing them solidly for 8 hours a day, every day.
Any data available showing the type of content and the frequency at which it’s seen?
I handled abuse reports for a ISP in the late 90s and the majority of it was shock/offensive stuff like lemonparty. The real severe things were few and far between even then.