6 ms·

Could a competitor have its employees flag a submission? Does HN have automatic vote manipulation detection? Does HN have automatic flag manipulation detection?

dang 5y ago
Yes, we've worked a lot on detecting such manipulation. I looked into whether it was employees doing the flagging, but I didn't include that in the GP because I can't tell for sure who was or wasn't an employee. What I did find is that of the flaggers, most haven't ever voted for a stripe.com submission (before or since). There were two flaggers who have voted for quite a few stripe.com submissions, but not enough to trigger any manipulation detectors. They both had voted for many thousands of submissions over many years, so I didn't find any evidence of abuse. In the case of one of those two users, there is a bio in the profile which makes it seem unlikely that they worked for Stripe. In the case of the other one, I don't have any info. FWIW, looking at the flagging data on that post pattern-matched to 'normal' in my brain, not to 'suspicious'. I realize that won't (and shouldn't) carry any weight for users who aren't already inclined to trust what I'm saying, but (a) we do have a ton of experience looking at suspicious voting and flagging patterns, and (b) it's true, so I figure I might as well say it. Edit: I just thought of another way to investigate this, which I overlooked earlier: I looked at everything else that those users had flagged. That was easy enough, because none of them has flagged more than a few dozen submissions over the years. There were no other cases of people flagging Stripe-related submissions, except for this one, which was a pro-Stripe story: Stripe has democratised online payments and reshaped the digital economy - https://news.ycombinator.com/item?id=18153909 https://news.ycombinator.com/item?id=18153909 - Oct 2018 (43 comments)
- emerongi 5y ago
  Out of curiosity, how often do you detect vote and flag manipulation events? I assume the accounts in this flagging event were otherwise active and "healthy"? On Reddit, accounts are created years in advance so they seem legitimate when put to use for nefarious purposes later on.
  - dang 5y ago
    Alas, I can't answer that because HN is under constant attack by people trying to game it and I can't satisfy your curiosity without satisfying them too.
- rattray 5y ago
  Thank you for this! Can you share whether the people who flagged the article were YC-affiliated (eg; funded by YC or YC employees)?
  - dang 5y ago
    None of the flaggers were YC employees or YC-funded founders. Does that answer your question?
    - rattray 5y ago
      Yes, thank you!
- skinkestek 5y ago
  I'd almost suggest reducing the power of flagging on posts even more, but I won't for two reasons: - This is your job and from everything I see you already put a whole lot of thought into it and I don't think I can outsmart you here. - IMO things have become better the last decade so it seems to me some adjustments (either technical or in the user base) have already happened. [1] That said, maybe we as a community could become even better at not flagging things that others enjoy? [1]: alternative explanation, I've learned to look at /active instead
  - skinkestek 5y ago
    Ok, one perfect (I think) example that this still isn't perfect: If I hadn't checked /active today I would have missed the story about Google classifying 1 and 0 as copyrighted. Is there a reason why flagging that story shouldn't count as abusing flags?
    - dang 5y ago
      Do you mean https://news.ycombinator.com/item?id=30060405 https://news.ycombinator.com/item?id=30060405? It wasn't flagged. (More precisely, it was, but the flags were dropped by HN's anti-abuse system, and wouldn't have affected the story's rank in any case because the upvotes were massively dominant.) The post spent 13 hours on HN's front page. This seems like things working the way they should—am I missing something?
      - skinkestek 5y ago
        OK, you are probably right again. It is just that I spend quite a lot of time on HN and I missed it until I checked /active. So either I didn't notice it at first or I had a 13 hour break from HN ;-) That said, doesn't some stories with a 5-600 hundred upvotes linger on the front page for days? Am I mistaken? (And if not - do you have time to explain why I can find a <700 points story here the next day but not this one?)
        dang 5y ago
- ksec 5y ago
  This is strange and sort of answer some of the question I have had about HN. Personally I dont flag stories much, ( I wish I could look back at the history of post I flagged LOL ) I just dont vote on it. I am surprised so many people decided to flag a post just because it is Enterprisy. It may also be the case some people create account for submission, voting and flagging purposes. And only participate in minor discussions to gain / wash karma points.
  - dang 5y ago
    That wasn't the case here - at least the flaggers looked legit to me, and they had long pre-histories as normal users. The thing to understand about flagging is that people mostly flag stuff when they see it on the front page and they don't feel it's good enough to be there. Different users have opposing opinions about what content counts as 'good' on HN—so some will upvote an article while others will indignantly flag it ("WTF is this doing on HN?") once they see it. The front page is a composite of this struggle between upvotes and flags, based on different tastes. It's actually a rather intense struggle because frontpage space is so scarce (https://hn.algolia.com/?dateRange=all&page=0&prefix=true&query=by%3Adang%20scarce&sort=byDate&type=comment https://hn.algolia.com/?dateRange=all&page=0&prefix=true&que...). But it's mostly not manipulation—it's just that to most people, their own taste about what's 'good' is so obvious to them that for someone else to have a different taste feels impossible—it simply has to be disingenuous! I have a theory which I didn't post yesterday because it's speculation—but it's speculation that fits my intuition about how HN works, so maybe I'll share it in case anyone's interested. I speculate that some users upvoted Bolt's post because they thought it would be funny to see "How Bolt does fraud detection better" in a cage match with Stripe's "Improved fraud prevention" on the front page at the same time, as briefly happened—see #5 and #6 here: https://web.archive.org/web/20180418212112/https://news.ycombinator.com/ https://web.archive.org/web/20180418212112/https://news.ycom.... In other words, classic internet drama. The two titles are so similar that I wonder if Bolt had seen Stripe's post on the front page and rushed their article out to challenge it with that baity title. (If you look at https://web.archive.org/web/20180418211511/https://blog.bolt.com/better-fraud-detection-with-bolt/ https://web.archive.org/web/20180418211511/https://blog.bolt..., you'll see that the article itself had a completely different title.) That would certainly have been clever! Either that or it was just a coincidence, but either way, some users would have seen both, would have found the collision hilarious, and would have upvoted for that reason. Since that has nothing to do with article quality, you would expect other users to then see the article on the front page, go "WTF is this doing on HN", and flag it. This is standard HN dynamics: baity title -> upvotes -> annoyed users -> flags. That's how the front page operates all day. If this is what happened, then Bolt only made HN's front page in the first place because Stripe's article was already there—which would add considerable irony to yesterday's antics.
aliswe 5y ago
I have brought up the problem of flagging (or downvoting) in a HN post, but it got flagged. :)
- Kye 5y ago
  Some threads/subthreads seem to have a gravity for flagging and downvoting (and upvoting!) that's wildy off average and hard to pin down to a cause. Did you check other posts around yours to see if they were getting flagged too?
  - aliswe 5y ago
    it was the actual post, not a comment to a post though!
neya 5y ago
Yes, it happened even a few years back. I have seen this first hand with a startup (not going to name). At the time, the startup asked its employees to flag comments against them on HN, and vote only what their founder said. Today, they're a big multi-million dollar business, the founder gives many useful pieces of advice and interviews and what not, but to me, ethics is everything and I stopped recommending them to people in my circle.
- ilrwbwrkhv 5y ago
  Without a name is hard to trust your story

dang 5y ago

Yes, we've worked a lot on detecting such manipulation. I looked into whether it was employees doing the flagging, but I didn't include that in the GP because I can't tell for sure who was or wasn't an employee. What I did find is that of the flaggers, most haven't ever voted for a stripe.com submission (before or since). There were two flaggers who have voted for quite a few stripe.com submissions, but not enough to trigger any manipulation detectors. They both had voted for many thousands of submissions over many years, so I didn't find any evidence of abuse. In the case of one of those two users, there is a bio in the profile which makes it seem unlikely that they worked for Stripe. In the case of the other one, I don't have any info. FWIW, looking at the flagging data on that post pattern-matched to 'normal' in my brain, not to 'suspicious'. I realize that won't (and shouldn't) carry any weight for users who aren't already inclined to trust what I'm saying, but (a) we do have a ton of experience looking at suspicious voting and flagging patterns, and (b) it's true, so I figure I might as well say it. Edit: I just thought of another way to investigate this, which I overlooked earlier: I looked at everything else that those users had flagged. That was easy enough, because none of them has flagged more than a few dozen submissions over the years. There were no other cases of people flagging Stripe-related submissions, except for this one, which was a pro-Stripe story: Stripe has democratised online payments and reshaped the digital economy - https://news.ycombinator.com/item?id=18153909 https://news.ycombinator.com/item?id=18153909 - Oct 2018 (43 comments)

emerongi 5y ago
Out of curiosity, how often do you detect vote and flag manipulation events? I assume the accounts in this flagging event were otherwise active and "healthy"? On Reddit, accounts are created years in advance so they seem legitimate when put to use for nefarious purposes later on.
- dang 5y ago
  Alas, I can't answer that because HN is under constant attack by people trying to game it and I can't satisfy your curiosity without satisfying them too.
rattray 5y ago
Thank you for this! Can you share whether the people who flagged the article were YC-affiliated (eg; funded by YC or YC employees)?
- dang 5y ago
  None of the flaggers were YC employees or YC-funded founders. Does that answer your question?
  - rattray 5y ago
    Yes, thank you!
skinkestek 5y ago
I'd almost suggest reducing the power of flagging on posts even more, but I won't for two reasons: - This is your job and from everything I see you already put a whole lot of thought into it and I don't think I can outsmart you here. - IMO things have become better the last decade so it seems to me some adjustments (either technical or in the user base) have already happened. [1] That said, maybe we as a community could become even better at not flagging things that others enjoy? [1]: alternative explanation, I've learned to look at /active instead
- skinkestek 5y ago
  Ok, one perfect (I think) example that this still isn't perfect: If I hadn't checked /active today I would have missed the story about Google classifying 1 and 0 as copyrighted. Is there a reason why flagging that story shouldn't count as abusing flags?
  - dang 5y ago
    Do you mean https://news.ycombinator.com/item?id=30060405 https://news.ycombinator.com/item?id=30060405? It wasn't flagged. (More precisely, it was, but the flags were dropped by HN's anti-abuse system, and wouldn't have affected the story's rank in any case because the upvotes were massively dominant.) The post spent 13 hours on HN's front page. This seems like things working the way they should—am I missing something?
    - skinkestek 5y ago
      OK, you are probably right again. It is just that I spend quite a lot of time on HN and I missed it until I checked /active. So either I didn't notice it at first or I had a 13 hour break from HN ;-) That said, doesn't some stories with a 5-600 hundred upvotes linger on the front page for days? Am I mistaken? (And if not - do you have time to explain why I can find a <700 points story here the next day but not this one?)
      - dang 5y ago
ksec 5y ago
This is strange and sort of answer some of the question I have had about HN. Personally I dont flag stories much, ( I wish I could look back at the history of post I flagged LOL ) I just dont vote on it. I am surprised so many people decided to flag a post just because it is Enterprisy. It may also be the case some people create account for submission, voting and flagging purposes. And only participate in minor discussions to gain / wash karma points.
- dang 5y ago
  That wasn't the case here - at least the flaggers looked legit to me, and they had long pre-histories as normal users. The thing to understand about flagging is that people mostly flag stuff when they see it on the front page and they don't feel it's good enough to be there. Different users have opposing opinions about what content counts as 'good' on HN—so some will upvote an article while others will indignantly flag it ("WTF is this doing on HN?") once they see it. The front page is a composite of this struggle between upvotes and flags, based on different tastes. It's actually a rather intense struggle because frontpage space is so scarce (https://hn.algolia.com/?dateRange=all&page=0&prefix=true&query=by%3Adang%20scarce&sort=byDate&type=comment https://hn.algolia.com/?dateRange=all&page=0&prefix=true&que...). But it's mostly not manipulation—it's just that to most people, their own taste about what's 'good' is so obvious to them that for someone else to have a different taste feels impossible—it simply has to be disingenuous! I have a theory which I didn't post yesterday because it's speculation—but it's speculation that fits my intuition about how HN works, so maybe I'll share it in case anyone's interested. I speculate that some users upvoted Bolt's post because they thought it would be funny to see "How Bolt does fraud detection better" in a cage match with Stripe's "Improved fraud prevention" on the front page at the same time, as briefly happened—see #5 and #6 here: https://web.archive.org/web/20180418212112/https://news.ycombinator.com/ https://web.archive.org/web/20180418212112/https://news.ycom.... In other words, classic internet drama. The two titles are so similar that I wonder if Bolt had seen Stripe's post on the front page and rushed their article out to challenge it with that baity title. (If you look at https://web.archive.org/web/20180418211511/https://blog.bolt.com/better-fraud-detection-with-bolt/ https://web.archive.org/web/20180418211511/https://blog.bolt..., you'll see that the article itself had a completely different title.) That would certainly have been clever! Either that or it was just a coincidence, but either way, some users would have seen both, would have found the collision hilarious, and would have upvoted for that reason. Since that has nothing to do with article quality, you would expect other users to then see the article on the front page, go "WTF is this doing on HN", and flag it. This is standard HN dynamics: baity title -> upvotes -> annoyed users -> flags. That's how the front page operates all day. If this is what happened, then Bolt only made HN's front page in the first place because Stripe's article was already there—which would add considerable irony to yesterday's antics.

aliswe 5y ago

I have brought up the problem of flagging (or downvoting) in a HN post, but it got flagged. :)

Kye 5y ago
Some threads/subthreads seem to have a gravity for flagging and downvoting (and upvoting!) that's wildy off average and hard to pin down to a cause. Did you check other posts around yours to see if they were getting flagged too?
- aliswe 5y ago
  it was the actual post, not a comment to a post though!

neya 5y ago

Yes, it happened even a few years back. I have seen this first hand with a startup (not going to name). At the time, the startup asked its employees to flag comments against them on HN, and vote only what their founder said. Today, they're a big multi-million dollar business, the founder gives many useful pieces of advice and interviews and what not, but to me, ethics is everything and I stopped recommending them to people in my circle.

ilrwbwrkhv 5y ago
Without a name is hard to trust your story