The story the OP is complaining about was flagged by users. Moderators never saw it (edit: wrong, we put 2010 on the title by mistake, see downthread [1]). Had we seen it, we would have turned off the flags. There's a long tradition of people looking at HN data and posting about it. Edit #2: since the 2010 thing was our mistake (an accident of sleep deprivation by the looks of it!) I've invited foob to repost the original article using the second-chance mechanism described at https://news.ycombinator.com/item?id=11662380 and the links there. I think he's planning to do that tomorrow.
The [flagged] annotation only shows up on stories that are heavily flagged, i.e. enough to kill the post. User flags have downweighting effects long before that.
Story rank on HN is determined by upvotes, flags, software, and moderators. Moderators downweight stories to the degree that they don't fit the site guidelines. This doesn't happen by upvotes alone, unfortunately; certain stories routinely get tons of upvotes regardless of how good they are for HN—e.g. anything sensational, indignant, or meta. If we didn't have a compensating factor, those stories would dominate the front page every day and HN would no longer follow its primary rule: "anything that gratifies one's intellectual curiosity". Of course that means HN is subject to our interpretation of what counts as "intellectual curiosity". HN has always worked that way; before we did it, pg did, and he trained us to do it so it would work as before. There's no way around the need for moderator intervention on a site like HN—the clue is in the word 'moderator' itself: left to its own devices the system runs to extremes and it needs a negative feedback loop to dampen it.
When YC is involved, we do this less than usual as a matter of principle. When HN itself is involved it's a little bit different, because the hypnotic power of all things meta causes HN upvoters to go into an upvoting trance. Meta is basically crack, so we routinely downweight such posts—but only so much, to compensate for the crack effect. That's what I've done here, which is why the post is now at #7 rather than #1. It should probably be lower, but I want to illustrate the point that we intervene less, not more, when judgments about ourselves are involved. As a further example, a moderator actually turned off software penalties and user flags on this post this morning, which is probably why it went to #1 in the first place. That's more than I would have done but it shows how seriously we take that principle.
None of this is new information, btw. I've posted about it plenty over the years and am always happy to answer questions.
> "If we didn't have a compensating factor, those stories would dominate the front page every day and HN would no longer follow its primary rule: "anything that gratifies one's intellectual curiosity". Of course that means HN is subject to our interpretation of what counts as "intellectual curiosity"
I appreciate the straight talk. I had some questions about the above. Suppose there was a post on the front page that was popular, and "gratifies one's intellectual curiosity", but it happens to conflict with the interests of
A) HN,
B) YC, or
C) Specific individuals within YC, such as sama or pg
Can you guarantee that such stories will not be negatively moderated for that reason alone?
In the interests of transparency, can posts that are negatively moderated be publicly flagged, so that people can form their own opinions on whether the mods are using their powers in a fair and unbiased manner?
Most of us HN readers appreciate the work you're doing in moderating this forum. We're all in favor of mods using their powers in order to further the site's stated goals, but we're also very wary of mods using their powers in order to further their own personal agendas. We're worried that without sufficient transparency, it's only a matter of time before mods start abusing their powers. Hence why posts like this, and accusations of censorship, are such major topics of interest.
When stories are negative about YC or YC-funded companies, we moderate them less than we would if they were about something else. That doesn't mean we don't moderate them at all—it would be foolish to leave open any loophole that large on the internet—but we're conscious and careful about doing it less than we usually would. That is literally the first thing pg taught me about how to moderate HN, and the first thing I've taught every other moderator. I've written about this many times: e.g. https://hn.algolia.com/?query=by:dang%20moderate%20yc%20less... and https://hn.algolia.com/?query=by:dang%20pg%20chair&sort=byPo....
An example: the other day someone reposted an old (2011) controversy article about Airbnb. It was a me-too post of the kind users frequently submit in response to something currently on the front page, which tends to be bad for HN because it's repetitive. Also, the article was misleading in several ways (as I understand it) and long out of date in others. Our usual moderation practice is to penalize such stories—i.e. that are derivative of something else or clearly misleading—but we didn't do any of that because it was about a YC-funded startup. We're very clear about this with all YC founders and everyone who works for YC.
Stories about HN itself are a little different because of the evil catnip power of meta, but it's still the same principle: we don't suppress stuff because it's critical of us.
The idea that more data would clear everything up is a fallacy. "Just make everything public" is the sort of argument that sounds great on paper, but I think it would more likely cause just the opposite. People already take things out of context and construct fun, rage-inducing narratives about how awful we are—not because we're special, it's just a fact of life on the internet. Why would it ever stop? Of course it will never stop. What it would do is consume far more of our time and energy than it already does. That would steal resources from the things we should be doing to make HN better, basically giving a small number of users (some well-intentioned, others not so much) the power to do a giant DoS attack on the good things of HN. Hacker News doesn't magically self-regulate, unfortunately; it takes a lot of work to keep it interesting and stave off entropy, and that's what we need to focus on. Every time one of these things comes up, it takes hours, sometimes a whole day, to answer people's questions about it. There's a burnout risk too.
So I think the balance we gradually arrived at over the years is the right one: be open about the principles we use to moderate HN and answer questions in specific cases. Unless something big happens to change my mind, that's how HN will continue to work; I believe it's what serves the community best as a whole.
We're all in favor of mods using their powers in order to further the site's stated goals, but we're also very wary of mods using their powers in order to further their own personal agendas.
HN is quite well moderated, IMO (and I have been a moderator, and I spend a LOT of my time rolling my eyes at how most mods are doing it wrong, damnit). Your comment is a kind of comment I see a lot on HN that sticks in my craw. I blogged about that very recently, so, it isn't you and I don't want to get into a pissing contest, but this is my cranky feeling about remarks of this ilk:
"Get a clue, people. Hacker News is an awesome discussion forum because it serves a business purpose. Please stop acting like there is something bizarre or ethically questionable about the mods making sure that this free service actually serves the business that pays all the bills for it. Geez"
I don't agree with the the assertion that HN mainly "serves a business purpose", perhaps in its early days its main purpose was to funnel prospective people into YC (second step from pg's essays, I stumbled on HN this way back in the day).
While you consider the arguments as "inane", I think there is merit in people trying to understand the mechanisms on which HN is being moderated. The "Don't worry your pretty head, it's being done right" argument only drives up more suspicion, because this is exactly the way less than savory practices in government and elsewhere are hidden. I don't think this is the case with HN but trying to argue that the suspicions arise not (just) because people are just jerks or stupid.
I don't agree with the the assertion that HN mainly "serves a business purpose", perhaps in its early days its main purpose was to funnel prospective people into YC (second step from pg's essays, I stumbled on HN this way back in the day).
If HN serves no business purpose, why does it employ full-time paid moderators?
Obviously: They illegally magic up money in their basement from printing presses to support their idealistic desire to give all their time away for free, in spite of the inconvenient need to somehow also eat and keep a roof over their head.
I don't understand what you're trying to say. Which part of my comment do you disagree with. I specifically mentioned that I and others appreciate the work the mods are currently doing.
You are wary of them furthering their personal agenda. I quoted that.
There seems to be this assumption that good only occurs when some Christ-like martyr falls on their sword for the benefit of other people. The entire world seems to think that the only good men are dead men who died as martyrs to a cause and if someone is getting something out of the deal, they are evil predators.
I loathe this idea with every fiber of my being and would like to see it wholly stamped out of existence. I believe HN is as high quality a forum as it is because it has a purpose that serves the needs of the people who created it and run it.
I have done moderating for free. At the scale of HN, unpaid mods would absolutely not work. I wholly approve of them having a personal agenda here. I think the fact that they need to meet a standard of enlightened self interest is part of why they do so much better than most moderating teams.
The best way to optimize HN's value to YC is simply by having it be as interesting as possible. That's the reason the community comes here, and the community coming here is why HN has value, so to do it any other way would trade a global optimum for a local one—basically trading the goose that lays golden eggs for some eggs. We're in an unusually fortunate position this way. If HN were a startup constantly needing to get its numbers up, we'd be under business pressure to do things that made HN less interesting.
Secondarily, HN has job ads and (more recently) 'Launch HN' startup launches that get special treatment. The job ads are written about in the FAQ and the launches eventually will be, but we're still figuring out the best way to do them. These are concessions HN grants to YC in return for being funded by YC, and our intent is to have them be both secondary and obvious. I agree that it's fair for HN to make these concessions and especially with your reason for saying so: it's the way to have a sustainably high-quality site, which does not come free. What HN doesn't do is make non-transparent concessions, like moderating anti-YC stories differently. But I've covered that elsewhere in the thread.
> "I wholly approve of them having a personal agenda here."
Any forum with a policy like that will never gain the credibility and respect that HN currently enjoys. If someone wants to create a private forum and moderate it in a way that furthers their personal agenda, as opposed to the community's interests, more power to them, but they should disclose it publicly. As community members, it is entirely reasonable if we ask them whether this is indeed the case.
Thankfully, Dang disagrees with you and isn't moderating HN in a personal-agenda-driven manner (see his post above).
Fwiw I hear you and Mz arguing much the same thing here. It's 'personal' in the sense of self-interested but not in the sense of a personal agenda beyond that... or at least we try hard not to be that way.
> I want to illustrate the point that we intervene less, not more, when judgments about ourselves are involved.
Given "ourselves" is all of us here, a way to illustrate that in a trustworthy way would be to expose the meta data around the story. Which articles were flagged and removed, who flagged, how much it mattered, etc. If not in real time, due to possible exploitation for ranking tweaks, then perhaps in an acceptable offset of time.
I also recommend implementing a cost of suffering flag type for articles which use negative emotional responses to spread the information in the article, especially when it is dissonant and viral in nature. See https://www.youtube.com/watch?v=rE3j_RHkqJc for context.
This article seems reasonable, but here's one indication it's flirting with recursive irrational behavior:
> I have to admit that I found it a bit comforting that I wasn’t the only one who thought this all seemed a bit fishy.
Confirmation bias is still a bias. Making an argument with bias prevents exploring other explanations, such as errors in HN code, activity periods (lunch for example) and sharing of stories within other aggregates which elicit a negative response (a large company monitoring the comments may not agree with post, and consensus there effectively moderates it down).
As a regular user, I have no interest in being witch hunted because I flagged a story. Votes and flags should be anonymous as far as other users are concerned.
It has to cost something to hide good content if you expect people to take the time to produce or submit quality content and engage with the community. Otherwise you'll end with the things no one cared much about.
This is highly irrational and speculative in nature. If an entity flags all articles by another entity or group, they should be held accountable for their actions, which could be shown to be biased. Further, a "witch hunt" would be driven by irrational decision making processes in the "hunter" aggregate, which is the point of exposing the meta data in the first place. Stopping recursive irrational thinking is the goal here.
Incorrectly indicates agreement on concensus of the aggregate. I agree with your assertions. Moderators, as infrastructure currently stands, serve an important role in shielding users from the truth of things.
>Given "ourselves" is all of us here, a way to illustrate that in a trustworthy way would be to expose the meta data around the story. Which articles were flagged and removed, who flagged, how much it mattered, etc.
Publicly pointing out who flagged a story seems like a bad idea. I think the data should always be anonymized. I just think it will lead down a bad path to ostracize people for specific votes or flags.
I do however think it would be interesting to see some data after the fact for some stories say ones that reach 500 points or greater.
Publicly pointing out who flagged a story seems like a bad idea.
Possibly better idea:
Moderators could take a look at unfairly flagged stories and silently adjust flag weights for flag-abusing users.
This is of course based on my unscientific thought that the majority of interesting (IMO) stories flagged of the frobt page are removed by competitors (political or business-wise) rather than moderators.
The other post incorrectly had "2010" in the title and I made a comment about how this should be corrected. The "2010" was then removed.
If, like you say, the moderators never saw the story then who made these changes? I find it hard to believe that foob would put the wrong year in the title of his own post.
Edit: ok, I think I know what happened now. A moderator put 2010 on the title by mistake. I haven't checked the timestamps on the user flags, but it's possible that people flagged the article because they thought it was old, in which case the title change would have acted as a moderation penalty even though obviously we didn't mean it that way. I was wrong to say "moderators never saw it" because I checked the log for moderation penalties, found nothing, and didn't look for title changes. So I made a mistake too.
We're sorry for the mistake. I don't recall this particular thing ever happening before, but plenty of other moderator errors have. We try to admit them and correct them as soon as we know about them.
I know how difficult it can be to run a volunteer review system, but we have done it for years and have learned a few lessons. One thing you should do here is remove privileges for a bit from the moderator in question. Reviewers need to have some consequence for mistakes, otherwise they will tend towards moving faster and faster in their actions, with lower quality as the result.
As much as I love this place, every time I find out more about how the sausage is made, I lose faith in HN a little bit more.
Look at just the stuff from this case - heavy handed uneven constant moderation (you claim moderators dont see a bunch of stories, but those that you do see like this one, often get adjustments), misreporting, questionable transparency, too much hardly visible power of flags, etc.
I know the rules, but It really seems like at best we are a few moderator changes from having this place turn to something that wouldn't be valuable to people like me.
Are there other online discussion forums that also go into such detail about how the sausage is made you can point to as an example?
From the ones I frequent, HN via Dan and team is the only one I've come across which frequently and directly engages the community in a detailed and genuine way about their policies.
I'm sorry that any user who loves this place would feel that way! I've tried to answer some of the concerns you mention in comments elsewhere in the thread—not that you'll necessarily agree with the answers, but if there's something unaddressed, let me know, and if you don't get a reply, hn@ycombinator.com is best.
Would there be a way to be a bit more specific about flags? E.g. "inaccurate title", "duplicate post", "advertising"?
There's been times when an older article is posted and adding the year in the title might give better perspective. Or other times where it's clearly a PR post for a company and we just want it off the front page.
We've been toying with doing something like this. The original intent of vouching was to rescue good posts that happen to be [dead] (e.g. because a banned user posted them) which is why we didn't implement it that way at the time.
I think that this is a great idea. I imagine that it's very difficult to differentiate between users flagging things that really don't belong on the site and flagging things that they dislike for more personal reasons. Showing a "vouch" link to high-karma users on stories that are being flagged would make it easier to separate the two cases.
And besides, if a submission is flagged off the front page by that point I would think the damage is done short of resetting the gravity of the post to allow it to be in the same position on the front that it "should" have been. I'm not even sure if this is something they can do.
E-mail worked for me when I needed to ask them a question. The whole thing was resolved in less than a day, and it was painless. From what I've seen they're pretty active here in comments too, and seem inclined to reply to reasonable questions posed to them, including, "Mind if I send you an e-mail detailing some concerns?"
I honestly don't know, but I'd guess that there are only so many mods, and a lot more users/posts that demand their attention. To be honest though, I base this on experiences moderating communities in the past, not particular experiences here.
I'm sure that kind of thing could work, especially with the group of users here, but it would probably represent a full-time job for a mod just to police that page and make use of it. They may simply not see the benefit?
Moderators would not need to police it, users would use the page to rescue incorrectly flagged posts.
In any case, someone other than HN (e.g. Algolia) could construct a page/queue of flagged posts for review. It would be used by a subset of the audience for the "new" page.
No permission needed from HN moderators. Users already have the power to vouch for stories, they only need a timely way to review.
I think the rate of objectively incorrectly flagged posts is very low. The author's previous post is an example. Flagging meta posts is not objectively incorrect. Flagging posts that suggest nefarious motives by the moderators is not objectively incorrect. Flagging posts that are technically mediocre (i.e. not reflecting knowledge of HN's moderators or the long published general ranking algorithm) is not objectively incorrect. The only obvious objectively incorrect situation of flags killing a story is when there is organized effort to flag a story to death for ulterior motive...and I suspect that is probably already covered reasonably well algorithmically.
Technically and opinion aside, given the weighting function, what is supposed to happen when someone vouches for an eight hour old thirty-five point story? Does it move to page the middle of page two? How is justice done?
To quote Dennis Moore blimey, this redistribution of wealth is trickier than I thought.
In the weighting function, wouldn't vouching invert the changes that were previously effected by flagging, which in fact took place on an incremental basis? Or are you saying that vouching is not needed?
Generally, I think that a vouching queue looks like another way for people who like the sort of submissions that HN tends to flag to further promote that content.
I don't think flagging is creating problems in terms of HN's quality. As I indicated, there were lots of reasonable criteria under which to flag the previous post. To me, this post has characteristics that meet those criteria as well.
Where vouching might have a role is a quality story with very few points that is flagged dead based on misunderstanding or malintent. A story that hits the front page has had many eyeballs upon it and if it is killed from there by organic means, that's the way the system is supposed to normally work. Even when I think the story does not deserve it.
Do you want it show right away? I assume that showing it right away (as soon as possible) will only cause it to fall off even faster and fewer people bother to look at it and possibly upvote it.
I assume that similar to HN meta stories resulting in upvoting trances, seeing [flagged] leads to ignoring the submission.
I would guess that most front page articles have some number of flags from concern trolls, and the down-weighting effect is probably not linear, so it just wouldn't be very informative.
1. Downvotes after Xkarma you can downvote frontpage items
2. Flags _soft_ downvote people unlock easiers
3. Mods have a ""super"" flag which is the ranking equivalent of aging a post pre-maturely. So the weighting algorithm calculates the post as being much older then it actually is allowing it to slide off the front page faster.
The experience wouldn't exist without flagging in the first place. They're an integral part of HN's system, without which it would be overrun with spam, sensationalism, and other kinds of posts that don't belong here.
It's true that the effect of flags isn't perfect in every way, but we have to evaluate this on a whole-system level. I'm not aware of any way to make the whole system work significantly better (otherwise we'd have done it already) and certainly getting rid of flags wouldn't be one.
Humble suggestion: Have a separate page on the site that shows stories which have had strong negative penalties applied to them, whether from being flagged or moderators.
Suggestion: Make anonymized graph data available on submissions and flaggers so we can spot flag/upvote cartels. This would make for a great final project for some college course.
I suspect that the software does that automatically on the server. Given that 'cartels' are likely to be automated and hence at web rather than human scale, at least some of the counter measures probably should be as well.
Well those are the flagrant cartels. I hear there are pure-human cartels too. I see emails and Facebook posts and en-masse emails asking for Product Hunt up-votes from time to time from some VC associates. I imagine there is a similar thing going on on HN (but what do I know...i'm just a tiny startup staring at PyCharm most of the day.)
Because it is mostly a site for press release type content, I suspect Product Hunt has less incentive to discourage those sorts of get out the vote activities versus Hacker News. And Hacker News users have more incentive to flag posts that rise for reasons other than interesting content or comments. I mean, if I have an informed expectation of what I will get on the Product Hunt front page and it is different than my expectation about the HN front page in part due to the way get out the vote activities are treated.
Or just monitor the stories that get flagged off the front page (~25 over several weeks according to the post) and take appropriate measures (e.g. restoring with relative time and penalize flaggers). Of course they first have to decide whether they want users who cares about the stories they submit and therefor think that non bad things getting flagged off the front page is a problem.
Good point -- but certainly you can agree that after a certain point threshold flagging is increasingly less valuable and should be more to simply let the mods know to take a look.
Anything is preferable, at this point, to the methods used. Perfectly good articles are getting dropped and it comes across as censorship.
Have asked this before but any reason you don't have in your profile that you are the moderator? Seems very clubby and exclusionary to only have experienced or frequent users know who you are on this site and what role you play. Ditto I used to say this about PG and others (paul and other partners).
Is the secret handshake that important to the success of HN? Why throw newbies for a loop or make them have to think about this?
> There's no way around the need for moderator intervention on a site like HN—the clue is in the word 'moderator' itself: left to its own devices the system runs to extremes and it needs a negative feedback loop to dampen it.
Uh... not really?
Left to its own devices, the system will include both extremes and non-extremes. In order to cut extremes out you have to sample what the extremes are, detect them and then minimize them.
The sampling comes by letting extremes pop up. The detection comes by letting users tag stories by content and emotional weight. You can minimize them with moderation tags and meta-moderation flags and automating balancing algorithms.
Currently you support story flagging, comment flagging, up/down voting. You have no way to identify what the significance of these are, though. Without better tools for your community maintain itself, of course you need moderators. But there are simple ways around this that would cut most of the moderation out and allow the system to simply re-weight stories as they become more obviously extreme in one or multiple tag/flag configurations.
Here are the categories of the 30 newest stories at time of writing: social media, privacy, programming, sysadmin, espionage, consumer products, politics, health, games, science. The stories vary in emotional rating (1-5, meh -> i am upset). As tags come in for any given story, trends emerge about how that story is perceived: it is relatively emotional, it is political and science based, and it is upvoted very quickly. Clearly, this is more likely to be controversal than the low-emotion programming topic that takes an hour to reach front page, so the latter should get higher priority, because it is less controversial and more "important to HN".
Notice that no moderator is needed to figure out what is controversial and change the balance, and we didn't even need to do something complicated like analyze speech patterns in comments or story content.
Assuming they have basic protections to stop malicious users massively-downvoting stories they don't like? The number of [real] users rating the story improves the balance, and you can use the same system they do now to prevent voting by brand-new users, things like that.
> Meta is basically crack, so we routinely downweight such posts
Interesting that you make a drug analogy. What's wrong with letting HN users have their crack? Unlike actual crack, there's no danger of side effects, dependency, withdrawal, etc. Maybe the "upvoting trance" is the expression of a sincere communal desire? It's not like it's "trashy" content... meta is simply self-reflection. Why penalize it?
Maybe the fear is that the front page will be overrun with meta content... that makes sense, but I feel like one or two meta posts couldn't hurt. I think self-consciousness is a positive attribute for a community.
I call it 'crack' semi-facetiously and semi-affectionately, but it's an interesting analogy to play with.
> Unlike actual crack, there's no danger of side effects
There's a big danger of side effects to HN itself. When taken excessively—the natural tendency—it makes the site less intellectually interesting, which alienates the best kinds of user (and attracts the less best kinds). Maybe if we could be like David Cross and just do crack once each...
> one or two meta posts couldn't hurt. I think self-consciousness is a positive attribute for a community
This I agree with but HN is extremely well-resourced in this area. The intention behind penalizing such stories, as I tried to explain above, is not to suppress them but to countervail.
Speaking of meta, perhaps there could be a Meta: area strictly for meta discussions, similarly to Ask HN:? Then the meta addicts could congregate and upvote all the want there? A slight modification might need to be made so that Meta: prefaced titles never appear in the regular submission queue, or if they do they start with a huge negative factor.
This is more or less what works for Stack Overflow.
How about if there was a bit associated with every post that indicates whether it's meta or not. A submitter can set it when they post if they like; a mod can set or clear it later. Perhaps users could flag posts as being meta as well, i don't know.
Then we just take all the meta posts off the main page, and show them on another top-level page like Ask HN or Show HN.
I i think i'd be happy for meta posts to appear on the 'new' page. There aren't that many of them, so they won't flood it. If they get upvoted, they get promoted to a different place to substantive posts, so they won't clog it.
It's a bit like a GC which allocates large objects in the nursery, so you can still, eg, create a scratch buffer quickly, but promotes them to a large-object heap instead of the normal tenuring generations.
I'm in no way affiliated but I think it would be to cut down on what "shitposts" that are just meta-related and are low quality posts. While this one seems to have interesting theories, over on Reddit there is a scourge of highly upvoted posts that do little to nothing to motivate discussion or contribute to the community.
Isn't there another factor on rank based on ones Karma or is that a per account thing?
Aka, I've noticed that most of my stories I submit never reach the front page, but a nearly identical, or sometimes the exact link will make it to the front-page. I can only speculate that there is some high karma aspect to the ranking algorithm that works against high karma HN users.
Kind of seems opposite of how it should be but maybe there is a good reason for this approach that the mods can share.
Not based on karma per se but there are various ways for an account to get penalized. People who are worried about whether that's the case are welcome to email us at hn@ycombinator.com. We'll either tell you there's no penalty or offer to take it off in exchange for following the site guidelines a bit better.
This might be slightly off-topic, but is flagging supposed to be more like downvoting or reporting a post? The guidelines say it is for when a, "story is spam or off-topic," but that doesn't say what flagging means, just when to use it.
It's both. It's like a downvote in that it affects rank, and like reporting a post in that we monitor the flagged articles and take action based on flags. Sometimes the action is to turn the flags off.
Is there a threshold to the amount of flags you look at? If I am the only person to flag your comment here, would a moderator wind up looking at that flag?
I take a little bit of offense to this statement. The article identifies stories that had sudden extreme drops in position and then speculates that this could be caused by a hidden flagging threshold and/or moderator intervention. I strongly suspected that moderation was responsible in some of the cases but I also emphasized that this was because I didn't expect those stories to get heavily flagged and that we couldn't know for sure. All of these stories were effectively removed from the front page whether it was due to a hidden threshold or a moderator. I thought it was interesting to look at which stories were removed even if we could only speculate on the mechanism in each case.
You've confirmed that there are hidden thresholds in place that severely penalize stories and that the moderators make manual adjustments. I don't think that that makes the post completely wrong even if the moderator adjustments tend to be smaller and all of the stories I looked at were affected by automated means.
Do you have any thoughts on whether any of the listed stories are appropriate for the front page? As I said before, a few of these stories seem like "bad" posts but most of them appear fairly typical to me. Flagging is obviously necessary but maybe there is an issue with the user-specific weighting or how one of these thresholds kicks in if there seems to be a high percentage of collateral damage? I'm sure that you've spent a lot more time thinking about this than I have so I'm genuinely very curious to hear your opinion.
You're right, and I wrote that in haste, so I've taken out "It should really be lower than that because the article is completely wrong" and replaced it with "It should probably be lower".
Edit: I took a look at your list and don't see any pattern in those stories. Some belong on HN and indeed got a lot of time on the front page. Others clearly don't belong on HN. In a few cases moderators intervened to penalize the story or turn penalties off, but in a solid majority moderators didn't intervene at all. In a few cases a bunch of points were dropped by the anti-voting ring software so the 'real' score wasn't as high as it seemed.
To me all this shows, and the reason I described it as 'completely wrong', is that you can't reconstruct what's going on from timestamps, point totals, and story rank. There's simply too much data that isn't public. I know that's frustrating to the large section of HN's userbase who love to know what's going on and figure things out. But I don't believe making it all public would be helpful, so here's the balance we've arrived at over the years: we make public the the principles of how the system works, including moderator intervention, and we answer specific questions about specific cases.
This may be true for locks and cryptography but a message board is not really a 'security system' nor does it fail completely and catastrophically if someone manages to figure out the details of the ranking algorithm or spam countermeasures.
I can't really think of any good forms of 'security through obscurity'. Is the elimination of buffer overflow vulns and sql injections a form of obscurity? Is SSL a form of obscurity?
SSL is based on obscure prime numbers. Another example is user passwords (obscure text). Sessions and API tokens, too. Credit card numbers, garage door openers, and SIM cards all rely on hidden information. Even door locks are a physical form of hidden keys.
But you're right, there are some forms of security that don't require obscurity. For voting systems though, I would categorize them as "cat and mouse" systems, which unfortunately fall into the obscurity category.
Honestly this is typical of your usual responses to be dismissive, faux outraged... And also just incorrect, factually incorrect.
Someone finally has enough data to confront your obvious bias and enough people saw it that you had to back up and apologize but this is a rare exception.
Mods removed submissions and individual posts they find distasteful and it has nothing to do with HN guidelines, they are applied unequally just as some members get special treatment. hN is just not that honest of a place to have an open discussion on a variety of topics.
> There's no way around the need for moderator intervention on a site like HN—the clue is in the word 'moderator' itself: left to its own devices the system runs to extremes and it needs a negative feedback loop to dampen it.
Then maybe the system needs to be changed?
> None of this is new information, btw. I've posted about it plenty over the years and am always happy to answer questions.
Is there anything in particular that should make us trust the information you provide? You could be lying for all we know.
An article comes out about how Hacker News has been censored, and Hacker News censors the post... because of "meta crack effect."
Without sounding too snide, it sounds like what's happening is that we do all of the work by posting articles and comments, and hacker news controls the narrative.
Shows stories which have hit the front page ever, in the order of their posting. If it's currently on the front page, the link is orange. If it's not, it's black. It's very interesting to watch how frequently highly upvoted and commented posts turn black, while their temporal peers remain.
Anecdotally, there appears to be trend of positive/neutral news about YC companies remaining on the front page the longest, the latest shiny technology sticks around for awhile longer than average, and pretty much any negative news disappears almost instantly.
For example, as of this instant in time, there is an article about Angular2 which remains on the front page while more highly upvoted and commented articles about laptop security, AT&T discrimination, and a Nintendo Switch CVE discussion are all gone from the front page.
> For example, as of this instant in time, there is an article about Angular2 which remains on the front page while more highly upvoted and commented articles about laptop security
For front page rankings time matters more than total votes. A submission with 10 votes can be ranked higher than one with 50 if those 10 votes came in minutes rather than over the course of a day. It's not surprising that controversial submissions that get less up-votes (and more flags) don't get ranked as high.
As much as I like HN, I'm not a big fan of the secrecy around moderator interventions - what gets censored, what posts get re-titled, etc.
I can understand they might want to keep the ranking algorithm and anti-spam techniques secret, but stuff that are manually censored by a moderator should be indicated as such, maybe by some automatic message like "This post was removed due to [reason]".
Some websites manage to fight spam while remaining reasonably transparent (eg. StackExchange, where pretty much everything is documented - flags, closing reasons, edits, etc.).
Any time I've reached out to dang (https://news.ycombinator.com/user?id=dang) via email, he's always been responsive and super helpful and pretty transparent. Moderating HN isn't an enviable task IMO.
The problem with that is that they have various tools like adjusting things off the frontpage or banning comment threads to the bottom of the list; but the UI itself does not provide information on those. If the mods do any of these things but don't also leave a human comment on them, then it's unlikely people will notice.
>the UI itself does not provide information on those //
Which is down to the varying desires of the community and the administration. The admin/mods have at least in part a goal of promoting their commercial ends whilst the community in general are neutral wrt those same ends. If the mods upgrade YC companies visibly and downgrade competitors, for example, then the chances are that will cause problems between the admin and the community.
Based on that I can't see YC ever agreeing to do their moderation in public. I think we have to remember that this is not a neutral forum but has an inbuilt bias, through moderation, towards the administrations benefits.
I think the perspective some people are taking here is that if it's invisible in the UI, then how would they ever know to ask? Short of running a continuous scraper and then compiling statistics as article did.
I'd be all for tagging a generic "Downvoted by moderation" onto any article this happened to. The poster will probably still be annoyed, but I imagine most would agree with moderation most of the time. And if they like, they can contact.
What they're less likely to agree with is this happening invisibly. And if it's invisible, then how do we know whether we (as a community) agree with it or not?
PS: Side note, strongly against "toggle a non-default option" as a solution. Filter bubbles are incompatible with democracy and true, diverse community.
I remember a few weeks back I was looking for a comment reply to my comment. The original post got flagged and the comment itself didn't appear in my comments block!
I was really disappointed as the comment I was looking for wouldn't even show up in search.
I've got my account (with 9+k point) banned. Which is, of course, annoying, but the most annoying thing was the comment left by moderator "we banned". How about "I banned"? Faceless corporate "we" does not look well.
There was also a lie about warning which never happened, but hey, "not lying" is not in the guidelines, so…
And while some of my comments might be harsh I sometimes wonder if overzealous moderation leads us to creation of some kind of Stepford here.
There was also a lie about warning which never happened, but hey, "not lying" is not in the guidelines, so…
I think you've unfairly jumped from the evidence you see (I don't recall receiving any warning) to an accusation that Dan was lying. Since lying about this doesn't seem to be to Dan's benefit (why risk adding false information?) it seems more likely that Dan may have confused you with another user, or that you may not have seen (or remembered) the warning.
Scanning back through your comments, I fairly quickly found the warning from a couple months ago that I think Dan was referring to: [redacted]. Likely you didn't notice this (or didn't consider it to be a warning?) but even if he's wronged you elsewhere, you probably should apologize to Dan for the false accusation.
while some of my comments might be harsh
Having just scanned through a couple months of your comments, I think that's unfortunately an understatement. A surprisingly large percentage of your recent comments are, as Dan warned, "personally abrasive". If your intent was to cause offense with each of these comments, I think Dan was right to ban your account. If the offense you were causing wasn't intentional, then there would appear to be a cultural misunderstanding that would be to your benefit to address.
the most annoying thing was the comment left by moderator "we banned". How about "I banned"? Faceless corporate "we" does not look well.
I agree with you about the use of "we". It often comes across oddly when it seems clear that the decision was made by an individual. I don't think a blanket switch to "I" solves it, but it seems like there should be a better phrasing.
First, I don't remember mentioning that the account you are pointing to was mine. And while it is true in this case this kind of exposing does not sit well with me at all.
Second, it is a bit of a stretch to call Dan's (I still have no idea who 'dang' is, there is no name or personal address on his profile, but based on your doxing skills I will believe you) comment a warning. You can see that in other cases he's a bit more explicit ("stop doing that or we will ban you"). Also, I may be wrong, but if you read his comments a bit of different treatment is applied, I guess this is up to personal likes and dislikes.
As for offensive comments… well I am firmly in a camp of thinking that offense is taken, not given. Sure, one can be excessively rude, but I do not think this applies here. Calling opinion you consider bullshit is worth banning? Well, sad day indeed, then. More so because you can find plenty of stronger expression here which are posted without repercussions—likely because mods agree with them?
Cultural differences? Close enough I guess. I am still not sure if shaping this place into a Stepford where everyone is all fake smiles and nods in agreement is a best way, but it is not up to me to decide.
I am glad that at leas on topic of "we" we are in some agreement. As I've mentioned previously there may very well be solid reasons for this kind of anonymity, but it still feels very inhuman.
I've had dang detach a post of mine, and IIRC give me a few other similar warnings. In the case of the detach, I was way out of line based on a misunderstanding of the original article (in my defense this misunderstanding wasn't cleared up until after my post).
But what it might come down to is, I'm from New York and a lot of people running HN are, if not actual San Francisco flower children of the 1960s, then perhaps their spiritual descendants.
And to flower children, the tone matters. They don't like you 'harshing' them. The opposite of how in-your-face New Yorkers often communicate. The HN moderators want you to make your points in a more subdued and less personal manner than New Yorkers might want to.
And it's not just dang. I've often been downvoted because of the tone of my posts, rather than because of (and even in spite of) the content.
So, if you want to play on HN, you need to play by their rules. It's their site.
In credit to HN, I'll mention a book written by Ken Hamblin, supporting the USA. The title is "Pick a Better Country" (if you don't like the USA).
So, pick a better alternative to HN? But I don't think there is one. HN probably is as good as it is because of, rather than in spite of, pg and dang and all the other (oppressive?) overlords?
I don't remember mentioning that the account you are pointing to was mine.
You clearly didn't. I happened to recall your comment about "we" from the thread where that account was banned, and thus knew where to look. I thought at the time about commenting to agree with you, but decided I was too late to the party for anyone to notice.
this kind of exposing does not sit well with me at all
I think there's a difference between "doxing" and a link showing that a warning was given to the other account, but I don't know where the line should be. I'll ask to have the link removed.
I still have no idea who 'dang' is, there is no name or personal address on his profile
I'm not sure why he's chosen the ultra-low-profile approach, but at this point it's clearly something that he's consciously chosen. Here's the official post introducing him several years ago: https://blog.ycombinator.com/meet-the-people-taking-over-hac....
Second, it is a bit of a stretch to call Dan's ... comment a warning.
Arguably, but I was only aiming for the lower bar of showing why Dan might have considered himself to have given you a warning, and thus might more charitably be considered to be mistaken rather than "lying". And while he didn't use the words "this is a warning", I do think most people would have correctly taken his admonition as a warning.
Calling opinion you consider bullshit is worth banning?
No, disagreement is fine, even if it involves telling someone that something they said is completely wrong. The moderators are (properly I think) very sensitive to the difference between saying that a particular idea is stupid and foolish, and claiming that person who said it is a stupid fool.
It's possible this distinction is considered more essential in North America than elsewhere. Personally, I'd suggest writing to 'hn@ycombinator.com' and asking if they would unban your previous account. If nothing else, you'll probably get a better explanation of where the line is between acceptable and bannable.
More so because you can find plenty of stronger expression here which are posted without repercussions—likely because mods agree with them?
I'm sure there is occasional bias in this direction, but I'm also sure that there is a conscious attempt of the moderators to compensate for this by being more forgiving to those whom they personally disagree with. I don't know what the end effect is, but I'd guess that (by design) the two effects mostly cancel out in the long run.
I am still not sure if shaping this place into a Stepford where everyone is all fake smiles and nods in agreement is a best way, but it is not up to me to decide.
While there are probably companies and even governments that have fallen prey to this problem, I'm doubtful that excessive civility has ever resulted in the downfall of an online community. It's fun to envision the endgame, though, with the "moderators" desperately trying to goad people into being more assertive and combative, and being met with unflappable peace and harmony.
It seems likely that the mods talk to each other and critique one another's decisions. It's quite reasonable to say "we" to mean "the moderators of HN as a collective group" if banning your account was a group decision. The fact you're getting a comment from an individual doesn't mean it was necessarily just that person's decision to ban you. It's pretty uncharitable to assume moderators don't take their 'job' quite seriously.
Also even if the moderator made that particular decision alone, he will not have to stand up for it as an individual if it gets challenged. Any sensible organization will back him up as long as he acted in good faith and did not break some basic rules. He is, at all times (except when not acting as a moderator), representing a larger entity.
(They might of course still have an internal disagreement over an issue, and maybe take internal action for bad decisions, and might overthrow that decision if warranted, but not by blaming it on one individual.)
To be fair, sometimes this is a security feature so that folks don't go off hating one individual moderator. I've seen this in games more often, but I imagine it works out here as well. It sucks for normal folks, but sometimes that's the way it works.
If they want to 'protect' the moderators they should use more neutral language focusing on the actions. You can easily say "This account has been x" instead of "We've x this account" or "x has been changed to reflect y" instead of "We've changed x to reflect y". It's less personal, but that's the point.
I think that works best; a moderator is / should be only a member that happens to be responsible for maintaining the site's user behaviour policy and tidiness of articles posted, and thus decisions shouldn't be an "I" or "we" thing, but "the site". Of course, that's an utopia and no amount of rules will cover every situation.
It may sound stupid, but it'd make me feel better if moderators still used "I" even if otherwise completely anonymous. Still add human touch and personal responsibility to the act even if there is no way to call some particular person out.
Well, yeah, I kind of get it, but still smells fishy somehow.
You still can be anonymous and say "I did something" now it just sounds like you are washing your hands and avoid being responsible.
OTOH, I was never a moderator of any kind, so I have no idea what problems they have to deal with.
When you get banned I'm pretty sure you get banned. So that account/accounts is/are closed but the person is banned. Clearly it's not enforceable but indicating a new account of a banned person would surely have that account removed?
Not necessarily. If we see some sign that the user is turning over a new leaf, we're happy to give another chance. The purpose of banning isn't to cast anyone out, it's to preserve a place for civil, substantive discussion. I sometimes hear people defend uncivil comments by saying that other people shouldn't be so fragile. But it isn't individuals who are fragile—it's the community.
Plenty of users have gone from being banned or penalized to being positive contributors on HN. Once someone understands why the rules and moderation are the way they are, things almost always go fine. It's not, for example, about "needing to sugarcoat everything", "avoiding uncomfortable truths", or any of that kind of explanation. It's about the extreme weakness of the social contract on the internet.
The site is being crushed by traffic right now -- but without reading the article, I've also found that some stories that I thought were important that were scrubbed from HN's front page just about as soon as I saw it (when I doubled back to read the comments)...
While I realize I'm not entitled to explanations, some transparency would be appreciated. Maybe it could even be automatic, whenever a mod removed something forcibly from front, they could leave a comment and it'd show up on some page?
[EDIT] - After reading the article, if a mod did indeed take down the post because it discussed reverse engineering the rank algorithm, I think that's pretty naive. Security through obscurity isn't a thing, and the better response is just to make a better algorithm, not try and suppress knowledge about it.
I say this naively myself, as I've never had to maintain a ranking algorithm with these many users who depend on it (or any at all for that matter), but surely the problem isn't intractable?
> Security through obscurity isn't a thing, and the better response is just to make a better algorithm, not try and suppress knowledge about it.
Obscuring an algorithm or making it more tedious to reverse may not make it perfectly secure, but that's not the goal. It's not like actual information security, where loss of the encryption keys means your product is broken or your database is on the Internet. You're just trying to minimize the workload on humans who act as a back-up for the few posts that slip through.
If an email spam detection algorithm was public, spammers could precisely craft their content to slip through. If the heuristics for showing a CAPTCHA were public, bots could automate their requests to avoid it. If a ranking algorithm was public, people who might financially benefit from the front-page traffic could force content there through vote rings and sock puppets.
If the algorithm is secret, far fewer will be able to do so, and this small fraction of abusers can be handled by humans.
You can always find reasons why something has to be a certain way. But at the end of the day, you can't expect to attract and retain curious hackers/entrepreneurs in this way. The author presumably spent some time compiling the original story, the math symbols and graphs are quite nice for instance, just to have it removed without explanation. It's not a very good way to treat people.
Does it matter? The author seems representative of exactly the type of users you want in a community like this. Long standing user account, high karma, produces decent comments and great content (the visualizations in the Swype article is another example [0]) and even puts HN comments on the blog. Whatever mechanism that ends up discouraging people like this in favor of trigger happy flaggers is broken. Fortunately for the community the author didn't go "fuck it", but instead wrote another sensible blog post which makes them an even better community member.
Once it's public (even unofficially), people will begin to rely on the current implementation and then complain if it is ever tweaked or changed. They build towards the value it grants and then complain that the value was arbitrarily stolen from them.
That's for sure. I rarely visit SEO websites or forums, but the few times I did I saw lots of complaints from people who seemed to think that Google owed them something after tweaking their page ranking algorithms.
The debate is not really whether the algorithm should be made public or not, it's about the method used to try and keep it secure.
If the algorithm can be reverse engineered then trying to suppress the knowledge that is already "out there" will only create an illusion of it being a secret, and the fewer people know about it the more damage they can potentially do (i.e they more they can financially benefit from their knowledge).
It's the same as with information security - if you discover an exploitable bug then chances are someone else has already discovered it too (or can discover it any time) so making it public is one of the most sensible things you can do.
It's not the same as with exploitable bugs, because exploitable bugs are fundamentally preventable. Not necessarily in aggregate; but individually, all bugs can essentially be patched given enough time or effort. There's no benefit to keeping them secret if their threats can be neutralized.
As I outlined in another comment in this thread, algorithms that do not offer or adopt significant authorization constraints (as quantified by time/monetary costs) cannot be "fixed." This is fundamentally why reverse engineering e.g. HMAC signing algorithms, search results ranking, spam filtering or front page listing algorithms is possible. The generous usability requirements do not allow for authorization that would mitigate reversing the algorithm, even when it's not embedded in an untrustworthy client.
Suppression is essentially all you can do to prevent reverse engineering, and suppressing the knowledge of how to reverse engineer an algorithm is in effect the same as suppressing the algorithm itself.
What does "verifiably" mean for you? Are you talking about provable security?
First establish an upper bound, worst case scenario cost (as a function of time + resources) to fully reverse engineer the algorithm. Use that as the comparison benchmark, and if you can come up with a design that eliminates any reverse engineering efforts with fewer costs than worst-case, you've done it.
Here's where that breaks down: "ungameable" is not precise enough to establish worst-case bounds for, in the same way that we can establish worst-case bounds for breaking an MD5 hash (brute-force it - what does "brute-force it" mean for gaming a ranking algorithm, or reverse engineering more generally?). Other than that, were you to come up with such a measurement, it would almost assuredly increase the costs of reverse engineering to infeasibility by increasing the authorization controls in place and decreasing the usability requirements.
Well, probably not for such a broad description of "ungameable", but there is an entire research field dedicated to try and come up with such algorithms/systems: Mechanism design!
I'm pretty new here, so take this with a grain of salt. My impression of this place is that it's a pretty highly regulated forum. In particular, it's designed to serve the needs of YC, and is not particularly free. That's the good and the bad of HN. It's a walled garden essentially, but instead of a wall it has very busy gardeners.
Honestly, it's why I like it, but I also recognize that behind all of this is a business, not a chatroom.
Edit: To clarify: there are so many places online to speak whatever is on your mind. What's lacking are places to have a decent conversation, any conversation, that can remain a conversation. How many places like this, with this many users, exist? I can't think of many.
We're on a discussion board web site. Using it to discuss the subject would seem the obvious place, right? Perhaps the mods could use their own site and be responsive to everyone...
As others have pointed out, the [flagged] annotation means heavily flagged. A story can be downweighted off the front page by user flags long before [flagged] shows up. Indeed that's what happened to the submission the OP is complaining about. Moderators never saw it.
Here's an example of a submission that does not include the [flagged] label, but which has a moderator saying it got flagged, and that the flags are the reason it's not on the front page. https://news.ycombinator.com/item?id=13741276
Just a few flags from high karma accounts can knock a submission off the front page if it isn't getting enough counter up-votes. It wont be marked as flagged.
There can be the reverse though. A controversial post may get flagged repeatedly but the impact of flagging could be disabled by mods to not disable the post.
I am of the mind that Mods can also prevent submissions from an account from ever being displayed in the page rankings.
Many months ago, I had several submissions that were visible when I was logged in, and yet when I accessed HN as a guest, they were not in the submission ranking.
That's something that happens. It's always worth talking to the mods to sort that out. There's a bunch of stuff to detect vote rings, spam, etc etc and sometimes it unfairly blocks the wrong stuff.
From what I've seen mods are happy to fix any problems.
As others have pointed out, the [flagged] annotation means heavily flagged. A story can be downweighted off the front page by user flags long before [flagged] shows up. Indeed that's what happened to the submission the OP is complaining about. Moderators never saw it.
I agree with you overall regarding transparency. It would be nice to see removed pages listed somewhere, or perhaps just the fact that it was done with a counter and maybe a timestamp. But as someone who reverse engineers a new mobile app nearly once a week these days: candidly, you can summarize nearly the entire discipline as security through obscurity :).
You use the word "suppress", but we generally can't use words stronger than that in security. We deal in abstractions that are hard to reasonably quantify, so we do so by approximations of monetary or temporal costs, and frequently both. We also try to limit absolutes. Reverse engineering exists as a discipline because 1) contrary to popular opinion, security through obscurity is valid, if incomplete and 2) there's frankly not much more that you can do in many cases other than obfuscation.
There are situations where algorithm secrecy gains you nothing defensively and is actually a strategic disadvantage, such as in encryption or hashing. But in situations where you fundamentally cannot discriminate between authorized users, such as in email (spam), search results (SEO) or, here, Hacker News (front page), you cannot rely on the strength of the algorithm to properly discriminate between users, because that's not its intended purpose. In these situations, obscurity is essentially your only remaining option.
To be fair the Hacker News moderators have more control of the ranking algorithm being reversed as it's on a remote server they control, as opposed to embedded in a client deployed to inherently untrustworthy hands. And once the information is out, it's out. But I don't agree that they have a trust or transparency imperative to keep that sort of submission on the front page. Even if the information exists, there's no reason to make it even more accessible. They can remove it and also improve the ranking algorithm.
If you want to design a general purpose web application without significantly reducing usability, functionality that is not restricted through authentication or higher levels of authorization is susceptible to reverse engineering. Being that the ranking algorithm is not client-side, there are fundamental protections we cannot bypass, but much of it is still inherently obfuscation. There is rate-limiting of course, and you have to log in, but the inherent inputs and outputs can still be somewhat flexibly assessed over reasonable timespans because there are hard usability requirements in place.
The tl;dr: algorithms which cannot be gamed because their inputs have significant quantifiable and controllable time/monetary costs do not require secrecy - these are excellent for implementing authorization. Algorithms which do not have such costs are not appropriate for authorization and, unless also paired with significant authorization constraints, require some degree of obfuscurity.
I can comment on the "Books that Aaron Swartz read, loved and hated" story. OP posted that link to HN and it made the front page. The story was soon flagged by the community (people viewed it as Amazon affiliate spam) and it fell off the rankings. When the moderator dang noticed that the story had been flagged, he decided to restore it (https://news.ycombinator.com/item?id=13840869). So yeah, that explains why this particular story fell off the front page quickly but does not have any [flagged] or [dupe] tags attached now.
I've learned that it's best not to jump to conclusions based on what you think is true (however sound your analysis might be). Always ask the other side(s) for an explanation. In this case, you could have sent an email to the mods asking for an explanation. If you find their response unsatisfactory, go ahead and write a post explaining why.
> There are also a few articles that are thinly veiled affiliate spam. For example, the ShelfJoy links about books that Aaron Swartz and David Bowie loved fall into this category. They get called out in the comments for being very low-effort lists of Amazon affiliate links.
OP uses this story as an example of having "moderator fingerprints on them" i.e moderators censoring posts, which is not true. What happened was the opposite : the moderator uncensored a post flagged by the community.
Well, the author tried to account for that by saying that normal flagged posts don't go down near as fast.
If what is being said in other threads is true, high karma accounts clicking flag could push a post off the front page without showing the "flagged" tag, then that could be an explanation instead. Which would then mean that maybe these high karma users have too much power.
Especially when it's a bit too easy to flag an article by mistake due to progressive rendering causing a misclick, or just being careless. I know I've done it, noticed, and then gone back to unflag. I wonder how often I've clicked off by a few pixels and didn't notice.
That's an interesting point. I hadn't even considered the possibility that a single user would have the power to drop a story by hundreds of positions. If that's the case then those users do have a lot of power.
- This is a bug / happens randomly; you just noticed it because you were looking (i.e. as you analyse this data); all the posts it's happened to before and since went unnoticed. That's supported by the evidence of your analysis; most of the results don't look any different to other posts.
- It's not the link, but the related activity. Presumably if you're running analysis on HN data, there are a lot of HN requests coming from your machine. Maybe any posts made by your IP are therefore treated as suspect (i.e. the sort of protection you'd expect to avoid automated posting or upvoting... just without that extra sophistication). Perhaps the other posters had something similar... Would be good to see if any of those posts were by the same author; as that may add weight to this theory.
- Other variables... Maybe the algorithm has rules which cause this behaviour under some conditions; e.g. posts made the previous day (not 24 hours ago; but rather before midnight UTC / something like that) lose weight when midnight hits; so posts made moments before suddenly lose enough score to knock them off the top spot; whilst those which had more score before midnight, or were posted just after survive... Many other possibilities such as this may exist; and we'd only know by looking at those variables in the data... What else is common about the posts which are in your post's club vs those which aren't?
Good points! I use the official Hacker News API [1] and my requests come from an ip address that is completely disconnected from my personal account. Even if the API usage were a red flag, there would be no way to automatically connect it to me.
You're definitely right about there being miscellanious rules in there. Something that I mentioned in passing in the article is that many stories exhibit a significant drop in position once they're 15 hours old. If you look closely at the typical story trajectories you can also see various other jumps of about 10-30 positions which I would guess are triggered by these various rules.
The stories listed in the article exhibit very different behavior where they jump hundreds of positions instantaneously. It's absolutely possible that this is triggered by some automatic mechanism but if that's the case then there's an enormous amount of signifance being assigned to the corresponding rules. If there's some random component to the ranking then I highly doubt that it would be responsible for jumps of this magnitude.
I try to emphasize in the article that I do think it's possible that there's a hidden flagging threshold that's responsible and that the data can't tell us with certainty whether or not that's the case. I just personally find it unlikely that that's what happened for all of these stories. If you ran a site like Hacker News then would you put an admin link next to each post that pushes it off of the front page? I know that I would.
I didn't understand why the guy's PayPal story was removed from the front page -- the one about PayPal seizing a $40k USD balance without warning, allegedly due to a 2% chargeback rate over several years of doing business and hundreds of thousands of dollars in successful transactions.
I thought that the point of HN was auto-moderation? Perhaps now that HN has seen great increases in popularity, the quality of content has to be more carefully controlled, lest the quality of posts on the HN front page slowly enter a death spiral towards that of reddit.
“Think of how stupid the average person is, and realize half of them are stupider than that.” ― George Carlin
I don't think the Paypal story was "removed". That story is now on page 2. Yesterday, it stayed on the front page all day. It's decay from the 1st page to the 2nd page a day later seems comparable to other stories with ~200 comments.
The author's meaning of "remove from the frontpage" means abruptly disappearing from the front page and you still don't see it in the subsequent 6 or more pages.
That's not quite true. "Buried" just means "gravity has been increased." Sometimes the gravity is adjusted massively, which moves stories to the 50th page. But it's very common for stories to go from page 1 to page 2. Even more common is to go from rank ~5 to rank ~29.
I flagged the Paypal story because it lacked supporting details that would allow me to evaluate the claims. For example, what was the person selling and how was it priced. Since one of the few leaked details was that the account involved USD to Canadian Dollar currency exchange, such information might have been relevant due to US currency control laws. In addition, people in the thread indicated that the dispute rate was high for e-commerce and the nature of the transactions might have clarified that.
In the end, the absence of detail meant there was no actionable information in the post, just 'Paypal sucks'. And I did not find that intellectually interesting.
A few months ago I wrote a tool that scrapes HN and detects moderator activity (and provides interesting stats in general). I posted it twice and it didn't receive much attention, and the database storage requirements starting to get out of hand, so I killed it. It seems like an appropriate time to bring it up again. The source code is here:
If there's interest, I can put this back up and start pruning old data so it's more maintainable. The data I collected shows a lot of questionable moderator activity and a lot of abuse of flagging. I'm also unhappy with HN sending all comments on paywalled posts (which are against the rules) to /dev/null, when they're usually at least willing to talk about things.
I wrote a tool that scrapes HN and detects moderator activity (and provides interesting stats in general).
Sounds great! It might be good to distinguish between "moderator activity" (which to me implies that a human reads and makes a decision) and "moderation activity" (which covers automatic adjustments such as the frequently faulty "flame detection algorithm").
If there's interest, I can put this back up and start pruning old data so it's more maintainable.
Yes, I think it would be very useful to have this available.
>Sounds great! It might be good to distinguish between "moderator activity" (which to me implies that a human reads and makes a decision) and "moderation activity" (which covers automatic adjustments such as the frequently faulty "flame detection algorithm").
Unfortunately this information isn't made available. More analysis of the data is necessary to identify trends that can be used to distingish these.
Historically, the mods have not killed a post for discussing Hacker News meta, although on occasion they apply a penalty to meta submissions. (the original post only had 32 upvotes, which is enough to get swallowed)
Presenting the facts out of context isn't helpful. The original post might have "only had 32 upvotes" but, as mentioned in this article, it received about 20 in less than an hour. More than enough to get to the front page, where it was steadily receiving more upvotes until it just vanished. The post was not "swallowed".
When a story gets to the front page more people will see it and some may upvote it. Likewise more people will see it and some may flag it. I'd hypothesize that longer term or more active users are among the people who are more likely to use flags. Likewise, I suspect that long term and active users are more likely to not want to see meta posts on the front page.
The point is that the post wasn't swallowed in the usual sense. We now know that the original post fell drastically because it was being flagged by users.
Does HN actually suffer this? I know its done on Reddit and by who in one case; just having a family member active in politics gets you good insight how many sites they try to manipulate.
I was hoping that HN's flag system would sufficient for the community to self censor. Perhaps we need a flag on comments too, something you can't see. I would also prefer that posts don't transition through the lighter grays as they get down voted but disappear completely once the dead threshold is met. That would prevent some piling on that does happen
Another possibility is that voting rings for _other_ posts flag maliciously to remove the competition and give their posts a relative boost. Maybe author can find correlations with other posts that were showing on the front page at the same time as the flagged ones and see how _their_ rankings got boosted due to the flagging.
I've chatted to mods on various sites over the years and I've heard that "voting rings", particularly the automated ones, often try to obscure their upvotes by also upvoting obscure stories and downvoting competing stories.
Likely if IP A posts a story and IP X,Y,Z always upvote it first, it likely flags that as suspicious. Or if IP X,Y,Z go directly to the link and upvote, instead of organically finding the post.
If you take a look at the comments, it's theorized there that the story got pulled not because of moderator action, but because people abused the flagging mechanism. Given the content, and given the principal person under discussion, this seems pretty likely to me.
Why would it be abuse if people flag it as being off-topic for HN? The article is about some community code of conduct text which the article author comments on by adding their own interpretations to it. It look political and it uses inflammatory language.
The tug of war between upvotes and flags, as commented by dang, seems a strong sign that the article is very much political in nature. The question is if it also gratifies one's intellectual curiosity, but personally, in this case it did not do that for me.
I'd say tech culture is 100% fair game for Hacker News. In fact, it's been the subject of a fair number of front page articles recently (look at the Fowler stories, for instance). The flagging suggests to me that they didn't like what was being said, rather than they thought it was off-topic.
A good test for this that the moderators could run is to see if those that flagged this article flagged any other articles about LambdaConf's policies and if so, which.
Another one is to compare its performance to other, similarly political articles by the same author (e.g. his free software posts) and see how they performed.
The main problem I have with this is that similar to Reddit, HN is advertised as open-minded, free-talking place by the team themselves.
When you then secretly censor stuff, because it doesn't fit your agenda, be it politically or financially, it makes you look even more like a hypocrite.
Branding yourself liberal while employing fascist methods (censoring and banning) seems to be a trend, not only on the internet.
The list of stories the author claims "have moderator fingerprints on them" does not seem to have any kind of common pattern at all which would support such a claim.
Edit: to be clear, with common pattern I mean the topics of the submission (obviously they have one common pattern, which is dropping out of the front page quickly). They do not reveal some secret agenda moderators might follow or something like that.
Apparently I'm going to go against the flow to say that I'm perfectly fine with the way HN is moderated. In fact, I think the moderation is probably the reason why I come here. If magazines, TV shows, and journals have editors, it seems completely natural to me that a quality website should have one too.
More often than not, keeping a community from turning into 4chan requires some heavy moderation (reddit's AskHistorians comes to mind, with entire threads nuked at once), and it's often a thankless job. I'm happy that HN managed to keep it's overall spirit, and I thank the mod team for that.
I predict that one of the biggest issues in tech over the next few years will be 'silent moderation.' Tech companies like to present the illusion that it is all 'just an algorithm' but that is deceptive.
Silent curation and other practices like shadow-bannning are unethical and symptomatic of a mentality that seeks to avoid confrontation. If things go well we'll see more transparency over time. A good start for a site like HN would be to create another page that shows just the titles of the submissions rejected (no links). People can google for those titles if they are interested.
I wrote in detail in 2013 about how the Hacker News algorithm works and the penalties that can drop stories from the front page. My analysis was based on reverse engineering the algorithm from observed behavior and comparing with the published Arc code. This latest analysis seems kind of reinventing the wheel.
Interestingly, my 2013 article also suddenly dropped off the front page. Apparently it somehow triggered "voting ring detection" and was penalized. (I'm not part of a voting ring of course.)
I don't think I understand your figures correctly, because none of them (except the last) seem to show any significant drop. There's a drop of about 10 positions in one of them but I wouldn't call that particularly significant, and it even climbs back up after twenty minutes or so.
Sorry if I didn't make this clear enough in the article but all of the stories in "Suppressed Story Trajectories" either drop off by hundreds of positions or disappear completely after their plotted line ends. They all match the criteria laid out in the previous paragraph. I used a larger y-scale on the last figure to show the full trajectory because that was a special case where the story didn't disappear completely and the trajectory after the jump was relevant.
Here is an example of three submissions, two flagged enough to get kicked off the front page (for poor use of sources on a contentious topic) but not get marked as 'flagged':
It seems to me that it's healthy if we as a community are able to discuss issues with how the community is policed. Some may find it boring, but I'd argue that it's useful.
I agree. I thought the write-up was fair and reasonably level headed.
If something popular (and surely by definition of being on the front page it is) is suddenly removed, people are bound to be interested in the reason why? Was the source discredited? Was it just a copyright issue? A simple filter for spiked stories would be good, just with a note on the reason why.
Of course HN don't have to implement this, but it would be of benefit to the community.
But the answer is always going to be "users flagged it", or "the flamewar detector was triggered", or "the vote ring detector was triggered". And advice is always going to be "if you notice something unusual email the mods to discuss it, because they can fix it if you send them an email; they might not see it if you leave a comment in a thread".
My point is you don't notice every post that gets removed and there is currently no transparancy on the reason why.
HN doesn't have to be transparent, it's just a site with it's own agenda (by that I don't mean evil agenda, but it is there for a reason) but if you want to grow the community, I think clearly identifying why things were removed is a reasonable thing to ask. If every one it marked "flagged by users" I'd worry that there is no manual intervention.
I did an Ask HN one time something along the lines of "how would you game the HN ranking algorithm?". It seems like good security practice IMO to have a discussion on what is publicly known / discernible about the ranking algorithm.
This was during the time of the election so I was thinking along the lines of political astroturfing, but also to guard against companies unfairly promoting their products or suppressing posts related to a rival company. For instance, if someone really wanted to keep a discussion off HN, all it would take is to tangentially start a flame war over some sensitive issue and watch the ranking algorithm punish the ensuing vitriol.
I don't really read this terribly as complaining. It is an interesting analysis of some posts (including his own) that seemed to behave atypically in a non-transparent manner to the user base, that may suggest some manipulation of the ranking. At no point does it seem to take any accusatory tone, or complain. The hypothesis may be incorrect, but it is not unreasonable, and it does seem relevant to readers of hacker news.
I don't think being "bored" by something is generally considered a good reason to flag content. I don't see how it could possibly be considered off-topic either.
I'm pretty happy with the way HackerNews handles posts and can only recommend to them to aggressively moderate in the future, too.
There are way to many toxic users, trolls, shills, astro-turfers, voting rings, paid advertising, political organisations, disinformation campaigns, and other 'special interest' parties on the Net to be able to do without strong moderation.
"There are way to many toxic users, trolls, shills, astro-turfers, voting rings, paid advertising, political organisations, disinformation campaigns, and other 'special interest' parties on the Net to be able to do without strong moderation."
That you worry about these things on the internet instead of real life (like the ones in our government) is rather telling.
Not every story. But some stories are cherrypicked by the mods to be shown for a short time in the front page. It's like a second chance for some good submissions.
The [flagged] annotation only shows up on stories that are heavily flagged, i.e. enough to kill the post. User flags have downweighting effects long before that.
Story rank on HN is determined by upvotes, flags, software, and moderators. Moderators downweight stories to the degree that they don't fit the site guidelines. This doesn't happen by upvotes alone, unfortunately; certain stories routinely get tons of upvotes regardless of how good they are for HN—e.g. anything sensational, indignant, or meta. If we didn't have a compensating factor, those stories would dominate the front page every day and HN would no longer follow its primary rule: "anything that gratifies one's intellectual curiosity". Of course that means HN is subject to our interpretation of what counts as "intellectual curiosity". HN has always worked that way; before we did it, pg did, and he trained us to do it so it would work as before. There's no way around the need for moderator intervention on a site like HN—the clue is in the word 'moderator' itself: left to its own devices the system runs to extremes and it needs a negative feedback loop to dampen it.
When YC is involved, we do this less than usual as a matter of principle. When HN itself is involved it's a little bit different, because the hypnotic power of all things meta causes HN upvoters to go into an upvoting trance. Meta is basically crack, so we routinely downweight such posts—but only so much, to compensate for the crack effect. That's what I've done here, which is why the post is now at #7 rather than #1. It should probably be lower, but I want to illustrate the point that we intervene less, not more, when judgments about ourselves are involved. As a further example, a moderator actually turned off software penalties and user flags on this post this morning, which is probably why it went to #1 in the first place. That's more than I would have done but it shows how seriously we take that principle.
None of this is new information, btw. I've posted about it plenty over the years and am always happy to answer questions.
https://hn.algolia.com/?query=by:dang%20downweight&sort=byDa...
https://hn.algolia.com/?query=by:dang%20countervail&sort=byD...
https://hn.algolia.com/?query=by:dang%20constitutional&sort=...
https://hn.algolia.com/?query=by:dang%20curate&sort=byPopula...
1. https://news.ycombinator.com/item?id=13858850