As the source article covers, they’ve been facing a number of complaints from human rights organizations that they are being insufficiently censorious regarding the Brazilian election, failing to swiftly take down some things which ethical standards and Brazilian law require. So that presumably influenced their policy. A lot of people seem to have an idea that Facebook’s election policy should quickly block all the bad ads with no risk of blocking any good ones, and that’s just not possible.
One paper that really hammers this point: Inherent Trade-Offs in Algorithmic Fairness [1]
The example they focus on is different but the general principles and takeaways are very powerful and applicable to all classification problems including content flagging.
