4 ms·

I've anecdotally noticed the same thing, but I have also noticed moderators asking users to refrain from conspiracy comments like this. I guess we have to trus

I've anecdotally noticed the same thing, but I have also noticed moderators asking users to refrain from conspiracy comments like this. I guess we have to trust that the administrators of this site, with full access to account and posting data, would notice if a concerted network of accounts was flagging certain submissions. I'd hope that their findings would be discussed publicly, but until then I don't think there's anything you or I can do to prove that this effect isn't due simply to many HN users disliking China politics stories.

potatofarmer45 7y ago
I've met quite a few of the YC team before and I know they don't have a person with this sort of speciality. There is a reason the top fraud detection analysts get paid so much on ecommerce companies - you have to be good at data science, behavioral analytics, and know enough tech to be able to spot this sort of patterns. I wish YC would have something similar to FOI, where people can publicly request for specific anonymized data. This way, we can a clearer idea of how various actors interact with a forum... in this case HN. If I could get anonymized data, I'd 1) filter first for all posts with keyword "China" that have made it any given point to the HN front page over last 3 months 2) Categorize articles as "positive towards", "negative towards", and "neutral" in content and tone 3) view upvote/flag rate before reaching the front page vs post exposure on the front page. 4) Compare across the 3. If people simply don't like China articles, then the pattern should be uniform. If it's not the same pattern, then you have drill down to specifically where the downvotes are coming from and when they start. 5) Since you can only flag but not downvote stories, is there an automated surge of upvotes for all surrounding posts excluding the targeted "negative" post? There's a lot to look at, but based on my anecdotal experience, over the last few months, negative Chinese articles are simply flagged/downvoted rapidly which contrasts to just a months earlier where there were robust discussions on each article. Real people predictably comment/interact consistently over time- that's proven. The fact that counter-posts to China-negative posts have largely disappeared suggests a large degree of coordination in the first place. Almost like rather than trying to win the argument online, the default is now to bury all negativity in the first place by promoting everything else on HN rapidly. If this is the case, there should be patterns of suspicious accounts. Key questions to ask: how did these accounts that downvoted get to the 400? karma threshold to vote. Was it a bot net where they upvote each others random posts? Or is it a manual process like the paid online trolls where accounts are created and provide interactions to reach the threshold, and then kept on ice storage until needed to downvote/flag? Just looking at the metadata and interaction patterns will give you a clear idea if this is typical post interaction behavior, or something more sinister.

potatofarmer45 7y ago

I've met quite a few of the YC team before and I know they don't have a person with this sort of speciality. There is a reason the top fraud detection analysts get paid so much on ecommerce companies - you have to be good at data science, behavioral analytics, and know enough tech to be able to spot this sort of patterns. I wish YC would have something similar to FOI, where people can publicly request for specific anonymized data. This way, we can a clearer idea of how various actors interact with a forum... in this case HN. If I could get anonymized data, I'd 1) filter first for all posts with keyword "China" that have made it any given point to the HN front page over last 3 months 2) Categorize articles as "positive towards", "negative towards", and "neutral" in content and tone 3) view upvote/flag rate before reaching the front page vs post exposure on the front page. 4) Compare across the 3. If people simply don't like China articles, then the pattern should be uniform. If it's not the same pattern, then you have drill down to specifically where the downvotes are coming from and when they start. 5) Since you can only flag but not downvote stories, is there an automated surge of upvotes for all surrounding posts excluding the targeted "negative" post? There's a lot to look at, but based on my anecdotal experience, over the last few months, negative Chinese articles are simply flagged/downvoted rapidly which contrasts to just a months earlier where there were robust discussions on each article. Real people predictably comment/interact consistently over time- that's proven. The fact that counter-posts to China-negative posts have largely disappeared suggests a large degree of coordination in the first place. Almost like rather than trying to win the argument online, the default is now to bury all negativity in the first place by promoting everything else on HN rapidly. If this is the case, there should be patterns of suspicious accounts. Key questions to ask: how did these accounts that downvoted get to the 400? karma threshold to vote. Was it a bot net where they upvote each others random posts? Or is it a manual process like the paid online trolls where accounts are created and provide interactions to reach the threshold, and then kept on ice storage until needed to downvote/flag? Just looking at the metadata and interaction patterns will give you a clear idea if this is typical post interaction behavior, or something more sinister.