Hacker News Reader

Comment by marcinzm | Hacker News Reader

The issue is also one of false-positives. Even if you could spot fake accounts 100% of the time if you also flagged 1% of legitimate accounts as fake by that approach then it's non-viable. Remember that at Facebooks scale a 1% false-positive rate means 20 MILLION accounts. That's too much to even try sending for manual review.

View on HN

Trust networks seem to work here: for example I've reported accounts like "Blacksmith Arms" (not a real example) for being not a real person. Blacksmith Arms is a common pub name, it's a business account masquerading as a user which is (or was) against Facebook rules.

Facebook responded on a couple of such reports but the accounts weren't altered.

If my account had sufficient standing then the action could have been automated. My accounts actions could then be meta-moderated by other accounts that are more trusted.

You make decisions using reports from trusted accounts, and verify them with multiple more-trusted accounts. Verified actions improve an accounts trust level.

Surely Facebook already use such a system? I imagine Google Maps are using such a trust network too?

Use the scale of the system to manage itself.

20 million reviews per year * 1/(2000 hours per personyear * 10 reviews per personhour) = 1000 people

Facebook could afford a thousand fulltime account reviewers, they just don't want to.

You're implicitly assuming that manual reviews would have a 1% or better error rate. I sincerely doubt that's realistic based on what I've seen of assembly line style processes.

A manual review would cost what, a dollar at most?

That's about 1/2000th of FB's yearly revenue. They should be able to handle that.

Reply on news.ycombinator.com