3 ms·

Well, now everyone can see the massive network of "bot" accounts that HN has. It's a growing trend on posts; you'll see them interacting with each other; they l

by gzer0 6y ago

Well, now everyone can see the massive network of "bot" accounts that HN has. It's a growing trend on posts; you'll see them interacting with each other; they like to work in nodes to boost the comment count and therefore somehow increase the chances of making front page. Even if they don't hit front page, the comment itself adds to the longer term plan of these accounts. I don't doubt for a second that someone has, or is currently manipulating a lot of the traffic and comments for certain posts. This is easily established via networks of rotating residentials; or better yet, real humans in Ukraine [1]: something that cannot be defeated unless you start nodal map chaining. High likelihood that these troll farm/serious commenters algos are not equipped to deal with such complexities such as human satire yet, which is a reasonable conclusion as to why there are so many "serious" posts [1] https://www.occrp.org/en/investigations/inside-a-ukrainian-troll-farm https://www.occrp.org/en/investigations/inside-a-ukrainian-t... [2] https://snap.stanford.edu/data/ego-Twitter.html https://snap.stanford.edu/data/ego-Twitter.html

leafmeal 6y ago
This comment would be really funny if it was a joke.. but I just can't tell what's real anymore.
- gzer0 6y ago
  [redacted]
  - dang 6y ago
    That's an example of a ring that was trying (rather hamfistedly) to manipulate HN, which was caught by our systems and neutralized long before the thread you're linking to ever happened. As I said at the time, if anyone can find an example of a story from those accounts making the HN front page, I'd like to hear about it at hn@ycombinator.com; examples of cases we missed are hugely helpful in plugging holes. But certainly the vast majority of their efforts went nowhere. People try every day to plant stuff on HN's front page, and every day our anti-abuse systems catch such efforts and apply different types of counter-action depending on the case. I'm not saying no one gets away with it (obviously we can't know that) but I can say for sure that anyone getting away with it would have to be way smarter than that group. In case anyone's interested, it happens that I've been working for the last few weeks on a new type of anti-abuse system, which is designed to more tightly enforce this site guideline: "Please don't use HN primarily for promotion. It's ok to post your own stuff occasionally, but the primary use of the site should be for curiosity.". https://news.ycombinator.com/newsguidelines.html https://news.ycombinator.com/newsguidelines.html
    - gzer0 6y ago
      Thank you for all you do and the thorough explanations.
      - dang 6y ago
        Appreciated! You didn't have to redact your comments and I'm sorry if my replies were more testy than they should have been.
driverdan 6y ago
Please provide evidence of this "massive network of bot accounts". I'm sure there are some but to this outside observer there doesn't seem to be any obvious large scale bot behavior.
- gzer0 6y ago
  [redacted]
  - dang 6y ago
    (I guess since you posted this three times I need to reply three times. Can you please stop? this is getting repetitive and wasting time that should be going into doing other things.) That's not evidence of what you claimed upthread; it was an inept bunch of content marketers who were easily caught by HN's anti-abuse systems. The thread you're linking to contains over 20 comments that I posted explaining various aspects of this. And if you read Troy's original report you'll notice that the examples he shows of them succeeding come from other places than HN: https://twitter.com/troyd/status/1315441811658219520 https://twitter.com/troyd/status/1315441811658219520 https://twitter.com/troyd/status/1316038695535833089 https://twitter.com/troyd/status/1316038695535833089
dang 6y ago
Please don't post insinuations like this unless you have specific evidence. This is in the site guidelines: https://news.ycombinator.com/newsguidelines.html https://news.ycombinator.com/newsguidelines.html. We have this rule because internet users are too quick to imagine they're seeing these patterns when really they're just reading them into what they're looking at, much like a Rohrschach diagram. It's not that it never happens for real, of course. Both things happen: real manipulation and imaginary accusations of manipulation. Both are bad, and we need some sort of evidence or data or $objective-thing to look at in order to have a hope of distinguishing the one from the other. We put a ton of effort into protecting HN from manipulation and have many years of experience with this. I can't say we're not being pwned but we certainly catch many attempts to game the system here. There's plenty of past explanation here if anyone wants more: https://hn.algolia.com/?sort=byDate&dateRange=all&type=comment&storyText=false&prefix=true&page=0&query=astroturf%20by:dang https://hn.algolia.com/?sort=byDate&dateRange=all&type=comme...
- gzer0 6y ago
  [redacted]
  - dang 6y ago
    I'm not sure whether you're asking "does this count as breaking that rule?" or "does this count as evidence that HN is overrun by bots?" Answer to "does this count as breaking that rule?": no, Troy emailed us long in advance and he had lots of specific data that we could look at. Answer to "does this count as evidence that HN is overrun by bots?": no, our systems had caught most of that activity long before. I posted over 20 comments in the thread you link to, covering a lot of these points; see e.g. https://news.ycombinator.com/item?id=24785165 https://news.ycombinator.com/item?id=24785165. And if you read Troy's original report you'll notice that the examples he shows of them succeeding come from other places than HN: https://twitter.com/troyd/status/1315441811658219520 https://twitter.com/troyd/status/1315441811658219520 https://twitter.com/troyd/status/1316038695535833089 https://twitter.com/troyd/status/1316038695535833089 If there are botnets or rings escaping our attention and succeeding on HN, they would need to be several levels more sophisticated than that one. One thing to realize is that HN's anti-abuse systems don't always reveal publicly that accounts or sites have been caught. It can often be more effective to preserve information asymmetry and fight the abuse in subtler ways. Re Rohrschach: the fact that it's considered dodgy is part of why I use the analogy. Most internet speculations about bots, astroturfers, shills, spies, brigaders, foreign agents and all the rest of it, are way dodgier than that.

by gzer0 6y ago

leafmeal 6y ago

This comment would be really funny if it was a joke.. but I just can't tell what's real anymore.

gzer0 6y ago
[redacted]
- dang 6y ago
  That's an example of a ring that was trying (rather hamfistedly) to manipulate HN, which was caught by our systems and neutralized long before the thread you're linking to ever happened. As I said at the time, if anyone can find an example of a story from those accounts making the HN front page, I'd like to hear about it at hn@ycombinator.com; examples of cases we missed are hugely helpful in plugging holes. But certainly the vast majority of their efforts went nowhere. People try every day to plant stuff on HN's front page, and every day our anti-abuse systems catch such efforts and apply different types of counter-action depending on the case. I'm not saying no one gets away with it (obviously we can't know that) but I can say for sure that anyone getting away with it would have to be way smarter than that group. In case anyone's interested, it happens that I've been working for the last few weeks on a new type of anti-abuse system, which is designed to more tightly enforce this site guideline: "Please don't use HN primarily for promotion. It's ok to post your own stuff occasionally, but the primary use of the site should be for curiosity.". https://news.ycombinator.com/newsguidelines.html https://news.ycombinator.com/newsguidelines.html
  - gzer0 6y ago
    Thank you for all you do and the thorough explanations.
    - dang 6y ago
      Appreciated! You didn't have to redact your comments and I'm sorry if my replies were more testy than they should have been.

driverdan 6y ago

Please provide evidence of this "massive network of bot accounts". I'm sure there are some but to this outside observer there doesn't seem to be any obvious large scale bot behavior.

gzer0 6y ago
[redacted]
- dang 6y ago
  (I guess since you posted this three times I need to reply three times. Can you please stop? this is getting repetitive and wasting time that should be going into doing other things.) That's not evidence of what you claimed upthread; it was an inept bunch of content marketers who were easily caught by HN's anti-abuse systems. The thread you're linking to contains over 20 comments that I posted explaining various aspects of this. And if you read Troy's original report you'll notice that the examples he shows of them succeeding come from other places than HN: https://twitter.com/troyd/status/1315441811658219520 https://twitter.com/troyd/status/1315441811658219520 https://twitter.com/troyd/status/1316038695535833089 https://twitter.com/troyd/status/1316038695535833089

dang 6y ago

Please don't post insinuations like this unless you have specific evidence. This is in the site guidelines: https://news.ycombinator.com/newsguidelines.html https://news.ycombinator.com/newsguidelines.html. We have this rule because internet users are too quick to imagine they're seeing these patterns when really they're just reading them into what they're looking at, much like a Rohrschach diagram. It's not that it never happens for real, of course. Both things happen: real manipulation and imaginary accusations of manipulation. Both are bad, and we need some sort of evidence or data or $objective-thing to look at in order to have a hope of distinguishing the one from the other. We put a ton of effort into protecting HN from manipulation and have many years of experience with this. I can't say we're not being pwned but we certainly catch many attempts to game the system here. There's plenty of past explanation here if anyone wants more: https://hn.algolia.com/?sort=byDate&dateRange=all&type=comment&storyText=false&prefix=true&page=0&query=astroturf%20by:dang https://hn.algolia.com/?sort=byDate&dateRange=all&type=comme...

gzer0 6y ago
[redacted]
- dang 6y ago
  I'm not sure whether you're asking "does this count as breaking that rule?" or "does this count as evidence that HN is overrun by bots?" Answer to "does this count as breaking that rule?": no, Troy emailed us long in advance and he had lots of specific data that we could look at. Answer to "does this count as evidence that HN is overrun by bots?": no, our systems had caught most of that activity long before. I posted over 20 comments in the thread you link to, covering a lot of these points; see e.g. https://news.ycombinator.com/item?id=24785165 https://news.ycombinator.com/item?id=24785165. And if you read Troy's original report you'll notice that the examples he shows of them succeeding come from other places than HN: https://twitter.com/troyd/status/1315441811658219520 https://twitter.com/troyd/status/1315441811658219520 https://twitter.com/troyd/status/1316038695535833089 https://twitter.com/troyd/status/1316038695535833089 If there are botnets or rings escaping our attention and succeeding on HN, they would need to be several levels more sophisticated than that one. One thing to realize is that HN's anti-abuse systems don't always reveal publicly that accounts or sites have been caught. It can often be more effective to preserve information asymmetry and fight the abuse in subtler ways. Re Rohrschach: the fact that it's considered dodgy is part of why I use the analogy. Most internet speculations about bots, astroturfers, shills, spies, brigaders, foreign agents and all the rest of it, are way dodgier than that.