[1] Birdwatch: Crowd Wisdom and Bridging Algorithms can Inform Understanding and Reduce the Spread of Misinformation, https://arxiv.org/abs/2210.15723
1. What % misleading/false posts are flagged
2. What % of those flagged are given meaningful context/corrections that are accurate.
It seems there's circular logic of first determining truth with 1, and then maybe something to do with a "trust"/quality poll with 2. I suspect a good measurement would be very similar to the actual community notes implementation, since both of those are the goal of the system [1].
The end-user experience of Facebook's moderation is that amidst a sea of advertisements, AI slop, the rare update from a distant acquaintance, and other engagement-bait, you get sporadic warnings that Facebook is about to show you something that it thinks you shouldn't see. It's like they're going out of their way to make the user experience worse.
A lot of us here probably have the experience of reporting posts to Facebook for violating this or that clearly-stated rule. By contrast, I think very few of us have the experience of Facebook actually taking any of them down. But they'll still flash weird fact-checker posts. It's all very silly.
Yes, but are they true?
Fact-checker warning labels are effective even for those who distrust fact-checkers
https://www.nature.com/articles/s41562-024-01973-x
They also cited this paper for the proposition that Community Notes doesn’t work well because it takes too long for the notes to appear (though I don’t know whether centralized fact checks are any better on this front, and they might easily be worse):
Future Challenges for Online, Crowdsourced Content Moderation: Evidence from Twitter’s Community Notes
https://tsjournal.org/index.php/jots/article/view/139/57