Re:Facebook moderators break NDAs to expose working conditions
by p1esk on Wednesday June 19, 2019 @11:09AM
(#20224040)
How about training a classifier to detect and remove potentially bad stuff, then let the users who uploaded it argue that the classifier made a mistake? Only in those cases have human moderators look at it?
[ Parent
| on HN ]
Re:Facebook moderators break NDAs to expose working conditions
by Sharlin on Wednesday June 19, 2019 @11:23AM
(#20224151)
They already have classifiers, of course. It's not like the moderators go through everything posted to FB–you'd need millions of people to do that. What the mods actually do, presumably, is review content reported by users.
Re:Facebook moderators break NDAs to expose working conditions
by p1esk on Wednesday June 19, 2019 @11:30AM
(#20224206)
Then perhaps they need better classifiers? I don’t remember seeing any papers from FB about that, so it does not seem like a priority for them. Or just make the existing bad classifiers more aggressive and shift the moderator’s effort to reviewing good content.