> The first two you listed were downranked by the flamewar detector.

Just some feedback that I've found a number of articles fall off the FP due to the flamewar detector that I've felt were good articles/discussions. In fact, I think some of the more valuable discussions tend to have a lot of back and forth discussions relative to the votes.

But I also recognize that flamewars can also look a lot like that.

So I'm wondering if it may be worth revisiting the algorithm for this, and maybe having it factor in a few other things vs. simply the vote:comment ratio (which is what I'm understanding it currently is, but correct me if I'm wrong).

I don't think it necessarily needs to be a lot more complex, maybe simply add to it some standard deviation of upvotes/downvotes (or just a simple ratio), if that's not already part of it.

But I've seen some discussions fall off that I don't remember seeing a particularly toxic discussion happening (e.g. relatively little to no downvoted comments).

Again, happy to see flamewars fall off, but just hoping to see some more interesting/helpful discussions not get caught in the crossfire.

Absolutely. We review the list of stories that set off that software penalty and restore the ones that are clearly not flamewars. No doubt we miss a few, and also - not everyone interprets these things the same way. But if you (or anyone) notice a case of a good thread plummeting off the front page, you can always get us to take a look by emailing [email protected].

There should be some way of doing language detection to detect the relative quality of 'flaming' going on.

So the highest quality 'flame wars' can remain untouched, but downranking everything else below that bar probably makes sense.

Yes, the carrot of automation would be so much nicer than the stick of manual review. I haven't seen any system that works well enough yet though.

The nice thing is that the comments are all public so if someone wants to take a crack at building a state-of-the-art sentiment detector or what have you, they can have a go—and if anyone comes up with anything serious, we'd certainly like to see it. As would the entire community I'm sure!

You don’t really need a state-of-the-art anything here. People get too distracted with building the perfect system when it comes to use cases like this because they are paralysed thinking about the avoidance of false positives and make a bunch of sub-optimal decisions on that basis. False positives are much less of a problem with a human in the loop, and putting a human in the loop doesn’t require moderator effort.

You can probably put a big dent in the number of low-quality comments by just showing a “hey, are you really sure you want to post this?” confirmation prompt and display the site guidelines when you detect a low-quality comment. That way you can have a much more relaxed threshold and stop worrying about false positives. Sure, some people will ignore the gentle reminder, but then you can be more decisive with flags and followup behaviour because anything low quality that has been posted will by definition already have had one warning.

You're right about one thing: I didn't need to say "state of the art". A system that works at all would be great!

I don't think a confirmation prompt will help because people tune such things out after they've seen them a few times.

I hate myself for saying it, because of all of the buzz/hype, but LLMs can assist here.

You get better intent assessment than with NLP/ regex/whatever.

Plus HN is entirely in English, so you never have to worry about lexical resource gaps.

There is no off the shelf solution - afaik. In addition I have no idea how expensive running costs will be.

But something serviceable can be built.

Source: mod /t&s person dealing with these things

Even a bad implementation isn’t going to be showing this warning to people often enough to desensitise them. And if they make a habit of ignoring the warning to post flamewar stuff… that’s solved with moderation by a human. The intent is to add friction for knee-jerk low-quality comments, not solve for people who persistently, intentionally post low-quality comments after a warning.

I asked for a showdead feed to make it easier to train an LLM on for this purpose but got denied.

Not sure what you're referring to, but you don't need a showdead feed to train an LLM for this purpose. Only 2% of comments are dead, and the number of bad comments that aren't dead is certainly higher than 2%. That's the problem, in fact!

Here's one from last week:

"Ring will no longer allow police to request users' doorbell camera footage" (npr.org) https://news.ycombinator.com/item?id=39138423

I posted an on-topic supporting quote to explain why this item was newsworthy and got one unhelpful one-word response and my comment got inexplicably flagged (not the commenter) https://news.ycombinator.com/item?id=39138481

How did that slip past detection? How do I get the abusive flag on my comment reversed? This behavior seems to have managed to push an important story off the frontpage quickly. (yes there was a badly-worded dupe headline, but that's a separate thing)

If I understand correctly, you have three concerns here: (1) the story was downranked off the front page; (2) your comment was flagged; (3) a comment that replied to you was not flagged. I'll try to respond to these in turn:

(1) the story was downranked off the front page because the topic had already been discussed a bunch—for example in these threads, two days earlier:

Amazon's Ring to stop letting police request doorbell video from users - https://news.ycombinator.com/item?id=39119387 - Jan 2024 (141 comments)

Ring steps back from sharing video with police – mostly - https://news.ycombinator.com/item?id=39120892 - Jan 2024 (15 comments)

Culling repetition from the front page is one of the most important things HN's systems need to do. Actually, it's probably the single most important thing. Certainly it's best if we can link to the previous discussions so people can know where to find them—but we can only do that some of the time. Users help out a ton by posting links to earlier threads. Ultimately we need better software support for dealing with this, but that's not done yet.

(2) Your comment https://news.ycombinator.com/item?id=39138481 was flagged by users. We can only guess why users flag things, but in this case I'm pretty sure I know why: comments that do nothing but quote from the article, or try to summarize the article, are considered too formulaic by readers here. If you want to say what you think is important about an article, that's fine, but please do it in your own words and share your own thinking. To simply paste a quote from the article is too superficial. On HN the convention is to assume that readers are smart enough to evaluate an article for themselves.

(3) The reply https://news.ycombinator.com/item?id=39138536, which only said "and?", was definitely an unsubstantive comment that deserved to be flagged (and killed) even more than yours did. The reason it escaped detection was simple, albeit unsatisfying: pure randomness. We don't come close to seeing everything that gets posted here—there's far too much. I've flagged it now.

Your comment was flagged by users. We can only guess why users flag things, but in this case I think I know why: comments that do nothing but quote from the article, or try to summarize the article, are considered too formulaic by readers here. If you want to say what you think is important about an article, that's fine, but please do it in your own words and share your own thinking. To simply paste a quote from the article is too superficial. On HN the convention is to assume that readers are smart enough to evaluate an article for themselves.

(I copied this from the parent comment so I can link to it when this comes up in the future).

I have a compliant : sometimes there a proliferation of anti-scientific posts, in example I can mention those related to the "50 years nuclear battery", I remember particularly one from techradar.com that was especially misleading and anti-scientific and more similar to a PR campaign then scientific information, they was stating che you can power a smartphone or a drone with a betavoltaic battery (millionth of Ampere ). This is only an example, I noticed similar article , often related to green energy with the same anti-scientific cut and sometimes anti-scientific is a euphemism. Could nice to have a way to report them , even for occasional readers like me. Often the same articles have approval posts that IMHO are bot made. we live in times where scientific fraud amplified by the media is becoming a serious problem and I think everyone should do more to stop the phenomenon.

Trying to assess what's scientific vs. anti-scientific is outside the scope of what mods can do. I have my opinions just like you do, but hashing these things out is a community process, not a moderation issue. We could put our fingers on the scale if we chose to, but nothing good would come of it, so we don't.

The flags on the last item don't seem to be made in good faith. This looks like abuse of the flag system to me. Is there a system for monitoring flag abuse?

By "the last item" you're referring to "Avoid Async Rust at All Cost", right? Personally I don't think that's abuse; I would have flagged that post if I'd seen it. That's despite the fact that I agree with a lot of what's in the post. The title is just too inflammatory. And there are more inflammatory bits in the post, such as saying the feature is "objectively bad", and saying that a community member's post "gracefully omits" some information (where the word "gracefully" sounds like an accusation that they were being disingenuous). Totally unnecessary. Chop off the inflammatory bits and you'd have a perfectly good blog post making an interesting point, but as-is that post was not going to lead to a productive discussion.

Of course, it's only inflammatory because async is a darling to more than half of HN :)

But if we get into that we'll trigger the flame war detection.

I assume you mean this one:

Avoid Async Rust at All Cost - https://news.ycombinator.com/item?id=39102078 - Jan 2024 (62 comments)

I can make an argument either way there. The argument in favor of flagging it could be: Rust is one of the most-discussed topics on HN; Async Rust in particular has had a ton of discussion [1], including a major thread just a few days earlier [2] - therefore this post was very much in the follow-up category [3]; the article was arguably rather low-quality, especially by the standards of this much-discussed topic; its title was flamebait and arguably misleading as well since the article seems more about async in general; and generally it was more of a drama submission on a classic flamewar topic than an interesting technical piece. I'm not saying all that is fair but it's easy to imagine good-faith users flagging for such reasons.

I checked the flagging histories of those users and only saw two cases where a user had previously flagged a different article about Rust, and one was years ago. For typical examples of other stories that the same users had flagged, see [4] below. A few of those might be borderline calls but I don't see abuse of flagging there. It's important to remember that even when a story is on topic for HN, flags are legit if the story has had a large amount of discussion recently. Otherwise HN would consist of the same few discussions over and over, and we have enough of that as it is!

> Is there a system for monitoring flag abuse?

There are some software protections in place against that, but like all such protections, they don't catch all the bad cases and they have false positives as well.

We review the flags and turn flags off sometimes. I would not say it's perfect because although we try to look over all the flagged stories, it has to be done hastily (or one wouldn't get much else done). That makes it easy to miss things. However, users often email us at [email protected] when they think a story has been unfairly flagged, and in those cases we always take a closer look. I don't know what percentage of the time we turn flags off in such cases, but it's not a low number. So if we include "users sending emails" as part of "the system", then yes, there's a system for monitoring flag abuse.

Last point: this is a pretty typical case. I'd say it's borderline but in the end I probably agree with the flaggers. If the topic of Rust (and async Rust in particular) weren't already so thoroughly covered, and/or if the flamewar aspect hadn't been there, then I'd probably disagree with the flaggers.

---

[1] https://hn.algolia.com/?dateRange=all&page=0&prefix=true&que...

[2] https://news.ycombinator.com/item?id=39061839 - the word 'async' appears over 200 times in that thread!

[3] https://hn.algolia.com/?dateRange=all&page=0&prefix=true&que...

[4] You Don't Have to Be a Jerk to Succeed - https://news.ycombinator.com/item?id=39228231 - Feb 2024 (21 comments)

Birth rates are falling in the Nordics. Are natalist policies no longer enough? - https://news.ycombinator.com/item?id=39191651 - Jan 2024 (151 comments)

New tires every 7k miles? Electric cars save gas; tire wear shocks some drivers - https://news.ycombinator.com/item?id=39175675 - Jan 2024 (64 comments)

Google layoffs: Tech giant to cut down 30k jobs, says report - https://news.ycombinator.com/item?id=38791297 - Dec 2023 (6 comments)

Code will make me rich and famous - https://news.ycombinator.com/item?id=38336699 - Nov 2023 (2 comments)

The NSA Invented Bitcoin? - https://news.ycombinator.com/item?id=37599194 - Sept 2023 (61 comments)

Leaving the Web3 cult - https://news.ycombinator.com/item?id=36803267 - July 2023 (47 comments)

How the Military Is Using E-Girls to Recruit Gen Z into Service - https://news.ycombinator.com/item?id=36471105 - June 2023 (97 comments)

Alphabet plans to announce its new general-use LLM called PaLM 2 at Google I/O - https://news.ycombinator.com/item?id=35866435 - May 2023 (5 comments)

Is your husband/ boyfriend gay? LGBTQ - https://news.ycombinator.com/item?id=35734086 - April 2023 (0 comments)

dang, thanks for taking the time for such a thoughtful response. I didn't know about the policy regarding topics that have been on a lot lately, that makes sense. I've not been around as much lately and hadn't noticed that this topic was well-trodden.

How can users downrank headlines? I only have an option to upvote them. While it's not too frequent, there are things that make it to the front page that I'd like to express my disapproval of.

User flags, once they've accumulated above a certain threshold, have a downranking effect. Pretty sure this is in the FAQ: https://news.ycombinator.com/newsfaq.html.

I'm curious why there's no actual downvote for submissions. Was that ever discussed on here? I did a quick search for prior discussions on the topic but didn't find anything.

To me, "flag" means "this is a serious violation that requires moderator attention". Something I'd want you to see and deal with because it's bigoted, illegal, spam, etc. I wouldn't flag something simply because I didn't think HN was the right audience, or because I personally dislike the topic. You seem to be encouraging me to use it simply as a downvote.

I'm not going to start flagging things, nor do I feel that strongly about the lack of a downvote, but if flags are effectively downvotes behind the scenes, and if that's how users are treating flags (which they obviously are, from other comments on this thread), I think the UI should have a downvote button.

I assume there's been discussion about this before and I'm curious about the thought process behind the decision. I don't find the FAQ to be informative about this.

The only person who could answer that is pg because that design choice was part of the embryo of HN.

He must have been thinking something though, because Reddit was originally his conception and he was an influence on the earliest development of Reddit as well.

It's good that this is in this thread, as I bet a ton of power users (I check HN multiple times a day for years but likely only a time or two have glossed over the FAQ), did not know FLAG could be used as a downvote tool. Interesting choice by PG, I agree with the previous comment, we have all come to know FLAG as a violation tool on most platforms. Now we know.

I'm surprised by this thread too.

I never flag anything because it's recorded on my profile, and I don't want stuff recorded on my profile that isn't useful to me. I only upvote submissions and comments that I intend to refer back to in the future. Upvotes are simply bookmarks to me, so my only tool for voting on the quality of conversation is downvotes. Which, apparently, I can't do for articles without spamming up my profile.

Actually I just checked my profile and saw several flags that must have appeared on a mis-click, just like how sometimes upvotes appear on a mis-click. Fortunately, unlike mis-clicked upvotes, you can still remove these.

Do you happen to remember if the misclicks were on mobile or not? I'm planning to add a confirmation screen to cut down on flags-by-misclick, but the current intention is to make it mobile-only.

Almost certainly they were on mobile, where the small font and use of touch screen to scroll makes mis-clicks much more likely. I'd be happy with a confirmation screen because I rarely or never intend to flag, but by the sounds of it there are a lot of liberal flaggers who could find the extra step annoying.

Not the original poster, but in my case it's always mobile.

Thank you for your tireless work. HN is a breath of fresh air compared with the rest of the internet thanks to it.

I've ran into this a few times, not just with flagging but with hiding too. It would be really helpful to have a confirmation dialog and/or a banner on the screen after the action was taken that would let you undo it. To answer your question: I think most of mine were on mobile but it would be nice to be able to have it in all environments if possible.

Tbh if you just upvote what you like and do not vote what you don't like it's almost the same thing.

The one exception is if some group organizes to upvote something that fits their agenda / business plan. But in this case it's generally something worth flagging and it gets flagged?

Why don’t you make the system transparent? This will save you a lot of effort answering questions.

"Transparent" means different things to people, but if you mean a full moderation log: I think most likely it would produce more such questions and effort, for no clear gain. I've written about this over the years: https://hn.algolia.com/?dateRange=all&page=0&prefix=false&qu...

Here's how I look at it: if trust is present, then we don't need to publish a full log, as long as we answer questions when people ask them. That degree of transparency has been available here for many years. If, on the other hand, trust isn't present, a moderation log won't create it. It will just generate more data for distrust to work with—and distrust always finds something.

Thus our focus is on building trust with the community and maintaining it. That happens through lots of individual and group interactions, answering questions whenever we get them, in the threads or by email. That's what I spend most of my time doing.

We're never going to take the community's trust for granted because it's what gives HN the only real value it has, and it would be all too easy to lose. But I do think that this approach has proven to work reasonably well for the bulk of the community. If people learn they can always get a question answered, that's a powerful trust-building factor.

Equally clear is that this way does not work for everybody; but that's always going to be the case no matter what we do. I don't mean that we dismiss such users' concerns—quite the contrary, I make extra efforts to answer those users. I'm just not under any illusion that we can satisfy everybody. It's satisfying enough if a few people can occasionally be won over in this way—which does happen sometimes!

While you argue against transparency, keep in mind that you are doing it in full public view.

Envision an airline withholding safety records, a car manufacturer keeping crash test results private, a restaurant refusing to provide health inspection logs, or a government refusing to disclose details of its budget allocation — all claiming that transparency would only complicate matters and provide "more data for distrust". In each case, the flawed nature of your core argument becomes obviously evident.

I fully expect your mindset and behavior to never change (unless forced), but just wanted to point out that your argument against transparency is a cop-out and that you're on the wrong side of history here.

I don't think you can reasonably compare the importance of transparency in your examples to that of editorial decisions in a private moderated community.

In the first set, the stakes are far higher, which is why the collection of objective data is legally mandated in the first place.

In the second set, you have only the subjective opinions of people who have an explicit goal to cultivate a specific variety of community. As members of that community, we select into the cultivation regime under which we participate. Not everyone will share the same preferences, and that's OK.

It's perfectly reasonable for private communities to have their own sets of rules and regulations. I'm simply pointing out an example of the all-too-common situation pervading certain communities where clear mechanisms of ideological projection and indoctrination (i.e. censorship, non-transparency, etc) are present but unseeable to many members of the community since it supports their existing ideological beliefs. a.k.a. Cognitive dissonance

That's interesting! But I don't think the "flawed nature of my core argument" is obviously evident—and I don't think the community would consider that obvious either.

It could be fun to look into it together, but the fun stops at "I fully expect your mindset and behavior to never change". Why dance if someone wants to kick you in the shins?

I hear you, but relying solely on the community's perception, particularly when it's predominantly of a certain mindset, doesn't guarantee objectivity. After all, recent years have seen an accelerating trend where leftist-controlled media and social-media companies have normalized the suppression of opposing viewpoints under the guise of 'fairness'.

It's one thing for private communities to set their own rules, but it's quite another to publicly tout these restrictions as universally fair when they're essentially projections of biased viewpoints coming from human-beings in positions of power. Actual transparency permits genuine openness and dialogue, rather than masking personal ideologies as righteous principles.

I have to admit, I laughed when I realized you are using a throwaway account, making this a strange work of performance art.

My original account was shadow-banned by HN years ago (probably by @dang) due to some overly heated political discourse bs.

Hence my interest in debating @dang's straw-man response....and nowadays, using throwaway accounts is often the only way others even see opposing viewpoints on topics like this.

I think the fundamental difference is that an airline or government is working according to objective rules and regulations, while HN is not. HN is trying to build a community, and I think that communities need subjective rules rather than objective

People will game it. We don't need a transparent algorithm when we have transparent results, e.g. enable `showdead`, or the OP's project.

If you have nothing to hide, why not make all story and comment removal history publicly visible, like Wikipedia edits.

That would create a bureaucratic nightmare for no significant gain.

Edit: see https://news.ycombinator.com/item?id=39234189 for a longer answer; and also krapp's comment at https://news.ycombinator.com/item?id=39232795, which makes a similar point.

Just enable showdead if you want to see all of that. It's 99% botspam.

That would create one more thing for people here to complain about. People here would just accuse the mods of faking the mod log to hide their "real agenda" whatever that is.

Wikipedia can and does vaporize edits.

I don't think revdel can actually fully delete a revision, there's always at least a revision entry left, perhaps with no user name or summary.

Lol, OK, throwaway665544

Or include the URL rather than just the HN ID so readers can follow the links.

yes! good point. Edit: I changed my GP comment to say "link" instead of "ID".

HN ID? I don't see that in the FAQ, maybe it's defined elsewhere?

edit: oh duh. thanks all, answer was 'right under my nose'!

The url for this page is https://news.ycombinator.com/item?id=39230513 so the id is 39230513

I changed my comment to say 'link' instead of 'ID' so everyone can follow the same links.

Thanks _kst_: https://news.ycombinator.com/item?id=39232594

Presumably its the id parameter in the URL?

Most likely in the URL, id=3923

I don't necessarily want to dissect every little story, but this post was a funny edge case:

https://news.ycombinator.com/item?id=39203106

a tame story that got some discussion, but was marked as a dupe. But I didn't see any other posts linked in the comments as expected. I search for other submissions and see two other posts... with 0 comments:

https://news.ycombinator.com/item?id=39190710

https://news.ycombinator.com/item?id=39186297

I don't really have a critique or solution here, I imagine false negatives are an inevitability. Just sharing.

We try to, and often users help by, posting links to the previous discussions in the thread. But there isn't enough time to do that in all of them.

In this case, you can see from https://hn.algolia.com/?dateRange=all&page=0&prefix=true&que... that there had been a lot of submissions, and from https://hn.algolia.com/?dateRange=all&page=0&prefix=true&que... the ones that got comments.

Of those, https://news.ycombinator.com/item?id=39165981 had been just a few days earlier. And it turns out that there actually was a link to it in the later thread: https://news.ycombinator.com/item?id=39204186, but this comment was flagkilled, probably because of the personal swipe in it. (You can still see it but only if you turn on 'showdead' in your profile.)

I could go on! because there's endless detail one can go into about these calls. But you "didn't necessarily want to dissect every little story" :)

Ahh, that would explain it. Showdead is off on my account and I guess I didn't find every result on my end. That's a shame. But thank you!

> could go on! because there's endless detail one can go into about these calls. But you "didn't necessarily want to dissect every little story" :)

Yeah, i imagine if I went down every tiny rabbit hole it'd be a full time job. I'll leave that to the professionals haha.

Thanks for replying with added context, didn't really mean to add more to your plate with this!

No problem! I see these threads as opportunities to explain things to the entire community so I try to make the explanations as thorough as possible, and to answer every question that I see. (though I'm sure I don't see them all - if anyone has (or sees) a question that didn't get answered, you can always let me know at [email protected])