A new way to attack Facebook is to fraudulently report a news story as false in hopes of reducing its visibility, either because someone wants to censor it or just doesn’t agree with it. Sometimes known as “brigading,” a concerted effort by trolls to flag a piece of content can reduce its visibility. Facebook now sends stories reported as false to third-party fact checkers, and these purposefully inaccurate reports can clog the already-overcrowded queues that fact checkers struggle to worth through.
That’s why Facebook gives users a trustworthiness score ranging from 0 to 1 depend on the reliability of their flags of false news, The Washington Postreports. If they flag something as false news but fact checkers verify it as true, that could hurt their score and reduce how heavily Facebook factors in their future flagging. If users consistently report false news that’s indeed proven to be false, their score improves and Facebook will trust their future flagging more.
Facebook’s News Feed product manager Tessa Lyons confirmed the scoring system exists. There’s currently no way to see your own or someone else’s trustworthiness score. And other signals are used to compute the score as well, though Facebook won’t reveal them for fear of trolls gaming the system.
This isn’t the only way Facebook ranks users, though. It assigns you a shifting score of affinity toward each of your friends that determines how frequently you see them in the News Feed. This “friend-ranking” score is essentially a measure of graph distance from you to someone else.
If you like a ton of someone’s posts, get tagged in photos with them, search for them, view their profile, communicate with them, have lots of mutual friends, are in the same Groups and have similar biographical characteristics like location and age, your score toward them is lower and you’ll see more of them in your feed. However, they have a different score for you depending on their behavior, so constantly viewing someone else’s profile won’t make you show up in their feed more if they don’t reciprocate the interest.
The need for this score highlights the difficulties of Facebook’s battle against fake news. Between subjectivity and purposeful trolling, there’s a lot of noise coming in with the signal about what should be removed. Anyone saying Facebook should have easily solved the fake news problems is likely underappreciating the nuance required and the intelligent human adversaries Facebook must defeat.
Facebook has a huge array of signals it can use to calculate Friend Rankings or trustworthiness scores. The question will be whether it can intelligently sort those signals to make coherent inferences about what to show us and when to believe us.