My question is: does HN actively attempt to counteract government actors from influencing the site? I think it’s been proven that China among other countries employs folks to try to influence social media sites. Not necessarily by influencing staff, but by creating user accounts who do things like downvote unfavorable comments or flag stories they don’t like.
This seems like it would be a prime target for that behavior.
I guess there is also a "flagging brigade" detector. [If not, I upgrade this comment to a feature request.]
Quote from Satya Nadella Q1 2019 Earnings Conference Call "...In fact, this morning, I was reading a news article in Hacker News, which is a community where we have been working hard to make sure that Azure is growing in popularity and I was pleasantly surprised to see that we have made a lot of progress in some sense that at least basically said that we are neck to neck with Amazon when it comes to even lead developers as represented in that community..."
Mentioned here before: https://news.ycombinator.com/item?id=27293480
However, if any executive is getting graded against this metric, Goodhart’s law applies, and there’s a good chance astroturfing would happen. Satya probably wouldn’t know about it.
If a Hollywood CEO says that they are trying to raise the audience Cinemascore ratings of their movies, we’d interpret that to mean that they are trying to make audience-friendly movies, not that they are trying to astroturf Cinemascore. And similarly, if someone at the studio were astroturfing Cinemascore, the CEO wouldn’t talk about it on the earnings call.
The better HN gets, the more people want to suck its juices for their own purposes. Most haven't figured out that the above-board way to do that is simply to make interesting contributions, so they do other things, and there's probably a power law of how sinister those things are. The majority are relatively innocuous, but lame. (Think startups getting their friends to upvote their blog post, or posting booster comments in their thread.)
Users are good at spotting these innocuous/lame forms of abuse, but when it comes to $BigCo manipulation (or alleged manipulation), user perceptions get wildly inaccurate—far below 0.1%—and when it comes to $NationState manipulation (or alleged manipulation), user perceptions get so inaccurate that...trying to measure how inaccurate they are is not possible with classical physics. Almost everything that people think they're seeing about this is merely imagination and projection, determined by the strong feelings that dominate politics.
How do I know that? Because when we dig into the data of the actual cases, we find is that it's basically all garden-variety internet user behavior.
It's like this: imagine you were digging in your garden for underground surveillance devices. Why? Well, a lot of people are worried about them. So you dig and what do you find? Dirt, roots, and worms. The next time you dig, you find more dirt and more roots and more worms. And so for the next thousand places you dig. Now suppose someone comes along and insists that you dig in this-other-place-over-here because they've convinced themselves—I mean absolutely convinced themselves, to the point that they send distraught emails saying "my continued use of HN depends on how you answer this email"—that here is where the underground device surely must be. You've learned how important it is to be willing to dig; even just somebody-being-worried is a valid reason to dig. So you pick up your shovel and dig in that spot, and you find dirt, roots, and worms.
Still with me? Ok. Now: what are the odds that this thing that looks like a root or a worm is actually a surveillance device? Here my analogy breaks down a bit because we can't actually cut them open to see what's inside—we don't have that data. We do, however, have lots of history about what the "worms" have been doing over the years. And when you look at that, what do you find that they've been up to? They've been commenting about (say) the latest Julia release or parser combinators in Elixir, and they've been on HN for years and some old comment talks about, say, some diner in Wisconsin that used to make the best burgers. And in 2020 they maybe got mad on one side or the other of a flamewar about BLM. (Nobody please get mad that I'm using worms to represent HN users. It's just an analogy, and I like worms.)
Or, maybe the history shows that the person gets involved in arguments about China a lot. Aha! Now we have our Chinese spy! How much are they paying you? Is it still 50 cents? I guess the CCP says inflation doesn't exist in China—is that it, shill? If @dang doesn't ban you, that proves he's a CCP agent too!
But then you look and you see that they've been in other threads too, and a previous comment talks about being a grad student in ML, or about having married someone of Chinese background—obvious human stuff which fully explains why they're commenting the way they are and why they get triggered by what they get triggered by.
This ordinary, garden-variety stuff—dirt, roots, and worms in the analogy—is what essentially all of the data reduces to. And here's the thing: you, or anyone, can check most of this yourself, simply by following the public history of the HN accounts you encounter in the threads. The people jumping to sinister conclusions and angrily accusing others don't tend to do that, because that state of mind doesn't want to look for countervailing information. But if you actually look, what you're going to find in most cases is enough countervailing information to make the accusations appear absurd...and then you'd feel pretty sheepish about making them.
I'm not saying the public record is the entire record; of course it isn't. We can look at voting histories, flagging histories, site access patterns, and plenty of other things that aren't public. What I'm saying is that, with rare exceptions [1], what we find after investigation of the private data is...dirt, roots, and worms. It looks exactly like the public data.
And here's the most important point: the accusations about spying, brigading, shilling, astroturfing, troll farms, and so on, are all exactly the same between the cases where the public data refutes them and the cases where the public data is inconclusive. I realize this is a subtle point, but if you stop and think about it, it's arguably the strongest evidence of all. It proves that whatever mechanism is generating these accusations doesn't vary with the actual data. Moreover, you don't need access to any private data to see this.
There are also trolls and single-purpose accounts that only comment in order to push some agenda. That's against the HN guidelines, of course, and such accounts are easy enough to ban. But even in such cases, it doesn't follow that the account is disingenuous, some sort of foreign agent, etc. It's far more likely that they're simply passionate on that topic. That's how people are.
[1] so rare that it's misleading to even mention them, and which also don't look anything like what people imagine
---
Still, power laws have long tails and one wonders what may lie at the end, beyond our ability to detect it. What if despite all of the above, there is still sinister manipulation happening, only it's clever enough to leave no traces in the data that we know of? You can't prove that's not happening, right? And if anyone is doing that it would probably be state actors, right?
You might think there's nothing much to be said about such cases because what can you say about something you by definition don't know and can't observe? It seems to get epistemological pretty quickly. Actually, though, there's a lot we can say, because the premise in the question is so strong that it implies a lot. The premise is that there's a sort of Cartesian evil genius among us, sowing sinister seeds for evil ends. I call this the Sufficiently Smart Manipulator (SSM): https://hn.algolia.com/?dateRange=all&page=0&prefix=false&so....
There are two interesting things about the SSM scenario. The first is that since, by definition, the SSM is immune to anti-abuse measures, you can't postulate any technical measures for dealing with it. It's beyond the end-of-the-road of technical cleverness.
The second interesting thing is that, if you go in for this way of thinking, then either there already exists an SSM or there eventually will be one. And there's not much difference between those two cases. Either way, we should be thinking about what to do.
What should we do in the presence of an SSM? I can think of two options: either (1) give up, roll over, and accept being manipulated; or (2) develop a robust culture of countering bad arguments with better ones and false claims with true information. Of those options, (2) is better.
If you have such a culture, then the SSM is mitigated because the immune system will dispose of the bad parts of what they're saying. If there are any true bits in what they're saying, well, we shouldn't be rejecting those, just because of who said them. We should be big enough to accommodate everything that's true, regardless of where it comes from—just as we should reject everything that's false, regardless of where it comes from. We might prefer to reject it a little more rudely if we knew that it was coming from an SSM, but that's not a must-have.
The nice thing is that such a culture is exactly what we want on HN anyway, whether an SSM exists or it doesn't. The way to deal with the SSM is to do exactly what we ought to be working at as a community already: rejecting what's false and discovering what's true. Anti-abuse measures won't work forever, but we don't need them to—we only need them to last long enough to develop the right habits as a community. If we can reach a sort of (dare I say it) herd immunity from the viruses of manipulation, we'll be fine. The answer to the Sufficiently Smart Manipulator is the Sufficiently Healthy Community. That's what the site guidelines and moderation here are trying to nurture.
Edit: I should add that I'm not 100% confident that this can work. But it's clear that it's the best we can do in that scenario, and the good part is that it's what we ought to be doing anyway.
In most cases it is the politics aspect or the unfair coverage aspect that leads users to flag a story, like say on lab leaks; but this story being flagged so easily was interesting. It is about a tech platform intentionally/mistakenly censoring things we will count as free speech.
> well, the reason that's not done yet is because moderation takes 90% of my time, answering emails takes the other 90% of my time, and counteracting abuse takes the other 90% of my time.
So much this. There just isn't enough time with a small staff.
> Most haven't figured out that the above-board way to do that is simply to make interesting contributions
So much this too. This is what we always told people on reddit -- brands would ask us "how do I get more popular on reddit" and we tell them, "make interesting content".
> Almost everything that people think they're seeing about this is pure imagination and projection, entirely determined by the strong feelings that dominate high politics.
Same with all social media. People assume governments have heavy handed control of all content on social media, when in most cases the government couldn't care less. They focus on using propaganda to control individuals and then let those people make a mess of social media.
Your whole post resonates with my experience on the inside of moderating a big social media site and meeting with moderators of other big sites.
I'll be honest, at first I wasn't too keen on you moderation style, as I found it too heavy handed. But I take that back. HN doesn't cover everything I want to talk about (I go to reddit for the rest), but what it does cover, it covers better than reddit does.
So thank you, and I hope you get some more help with one of those 90% jobs!
Maybe I was misreading it, but to me at the time it seemed like a flood of unreasonably positive people gushing about something they couldn't really have had any experience with.
Ah, but this is just proof, that the communist sleeper agents are entrenched even deeper among us, than we expected!
Unfortunately, it seems that we all do this. It's just easier to notice when other people are doing it!
Edit: I got to it! See the lower portion of https://news.ycombinator.com/item?id=27398725, after the "---".
If you or anyone notices something wrong with the argument, I'd like to hear what it is.
Plenty of orgs are surely trying to do that actively for all sorts of reasons. No idea how successful they are, probably tough to tell.
The spookiest thing of all is that most of the effect might be genuine grassroots action. Picture a Chinese Nationalist poster here, genuinely independent tech enthusiast and happens to know enough English to participate in an English forum. Perhaps they are genuinely annoyed by what they see as westerners meddling in their internal politics, which there is a long history of. Perhaps they flag what they see as clickbaity stories likely to lead to a bunch of China-bashing out of genuine annoyance. They don't need to be paid or leaned on by the CCP at all, they just actually feel that way.
Dammit, now I sound too apologetic about it. Sigh...
The pattern seems clear that these users are flagging the more sensational kinds of submissions that tend to lead to predictable discussions and flamewars. There's room for competing opinions about which of those are/aren't on-topic for HN, given the site guidelines; if you or anyone want to understand how the mods look at it, I recommend the explanations at the links below. But clearly the flagging behavior in this case was in good faith.
https://hn.algolia.com/?dateRange=all&page=0&prefix=false&so...
https://hn.algolia.com/?dateRange=all&page=0&prefix=false&so...
https://hn.algolia.com/?dateRange=all&page=0&prefix=false&so...
https://hn.algolia.com/?dateRange=all&page=0&prefix=true&sor...
https://hn.algolia.com/?dateRange=all&page=0&prefix=true&sor...
We have banned people in a few cases for serious $BigCo astroturfing but there's always a grey area in the Venn diagram around "PR operation" and "overzealous fan". You can't tell those apart without a smoking gun and those are hard to come by. Fortunately, from a moderation point of view it's a distinction without a difference because the effects on the site are the same.
Also FWIW, my sense (and we do have circumstantial evidence for this) is that even when these things are PR, they're somehow haywire (e.g. a contractor gone rogue), not official strategy, and if high-enough execs found out about it they'd probably shut it down. That's just speculation though; informed speculation, but not highly informed.
There was an interesting report in German TV, where they analyzed a paper looking for bot patterns in Twitter. That paper named some offending accounts, so what they did was PM one - and it turned out that it simply belonged to a pensioner with strong political opinions and a lot of free time. Interesting to look behind the cover some times (through I do think that TLAs realize this power and don't let that slide, to some extent at least).
> I'll be honest, at first I wasn't too keen on you moderation style, as I found it too heavy handed.
It's interesting how viewpoints diverge - for quite some time when I started reading, I actually did not realize that HN was moderated. If I may ask, where did you encounter so much heavy moderation?
I do not want to single out a single company, but would like to use this particular example to ask you the following: Please keep in mind the level of manpower and persistence, some of these corporations can call upon for their strategic objectives..
In 2020 Microsoft had, apparently, 106 lobbyist companies working on its behalf: https://www.opensecrets.org/federal-lobbying/clients/lobbyis...
and 94 in 2021 https://www.opensecrets.org/federal-lobbying/clients/lobbyis...
Looking at the website of some of these companies, offered services include and quoting: "Third party influencer outreach" :-)
I think social media (sorry for calling this site that) vote manipulation detection will be one of the defining problems of the decade.
A couple places. The one that bothered me most was that titles would get changed without asking or notification to the poster. Sometimes they would get changed to something I didn't think made sense, and then I looked like I had done that, since there was no indication that it was changed. I guess I'm still not a huge fan when it happens to me, but I see why it happens.
I also didn't like having my comments detached or cooled. If you reply to a top level comment with a good comment that happens to generate a flame war under you, it will get detached from the top into it's own thread, and that just felt weird because it made it look like I made a non-sequiter top comment and also stifled discussion (which was the goal of course).
Also if you make a comment that gets a ton of votes but is perceived as off-topic, they will put a flag on your comment that makes it fall in the rankings. So based on the points and time it should be up at the top, but instead will be near the bottom, sometimes under the comments with negative scores.
Lastly, I have dead comments turned on, and I would see dead comments that I didn't think deserved to be dead. Eventually I got enough karma that I could vouch, which helped.
Those were my main moderation complaints. I still don't particularly like when it happens to me, but usually when I see it happen to other people I think, "yeah that makes sense".
This one is interesting to me, because I have emailed the moderators to do exactly this for highly upvoted comments I feel take the discussion into what I feel are the wrong places. I can understand that for a new commenter such tangents might be novel, but for someone who’s been around here for a while I am curious if you oppose such actions for the nth time that someone drags “here’s my article about new C++ feature” into “honestly C++ just keeps adding too many things, discuss”.
Want to censor a thread on HN? Flag it with a few different users, or turn the thread into a shitshow so that the "flamewar" tools will be triggered, or moderators will be forced push the thread off the frontpage.
Tao Te Ching, Ch. 17,
With the best kind of rulers
When the work is complete
The people all say
"We did it ourselves."
In re: Option 2:
https://xkcd.com/810/ "Constructive"
> [[A man is talking to a woman]] Man: Spammers are breaking traditional captchas with AI, so I've built a new system. It asks users to rate a slate of comments as "Constructive" or "Not constructive". [[Close up of man]] Man: Then it has them reply with comments of their own, which are later rated by other users. [[Woman standing next to man again]] Woman: But what will you do when spammers train their bots to make automated constructive and helpful comments? [[Close up of man again]] Man: Mission. Fucking. Accomplished. {{Title text: And what about all the people who won't be able to join the community because they're terrible at making helpful and constructive co-- ... oh.}}
Cheers dang.
I can attest to this: at one of my old companies a post related to us ended up getting removed, just because so many of our engineers (entirely independently of the company) voted or commented on it. After that there was a very strict instruction from the company _not_ to engage with any posts about us...