Thinking about timestamps w.r.t. social graph PageRank:
Namely, let's say a new account joins, gets a few followers based on social referral, waits a few hours, and then immediately starts baitposting/spamming to gain followers. Surely, the accounts who followed the account before the spam are not culpable, but the accounts who follow *after* the baitposting are culpable, because choosing to follow is an implicit endorsement of the spam.
So, if spam scores are attributed onto the posts rather than the account, the negative score should be backpropagated onto to the followers if-and-only-if they followed after the posts occurred. Punishing users for not unfollowing is probably a step too far, but unfollowing could also be taken as a signal of negative endorsement, with some sort of exponential decay function over recent posts. Same could be done for following, with a follow signal being an endorsement of recent posts with an exponential decay. 3 replies
1 recast
5 reactions
1 reply
0 recast
1 reaction
0 reply
0 recast
1 reaction
