Ask HN: How to stave off decline of HN?

465 points by pg ↗ HN
I was just asking RiderofGiraffes if he had any suggestions for fixing the decreasing quality of comment threads on HN (http://news.ycombinator.com/item?id=2403449) and it occurred to me that I might as well ask everyone.

Anyone have any suggestions? We're on mostly uncharted territory here.

The problem has several components: comments that are (a) mean and/or (b) dumb that (c) get massively upvoted.

728 comments

[ 3.0 ms ] story [ 384 ms ] thread
A suggestion could well be to not have threads like this one (not trying to be disrespectful!) An interesting thought is the idea that punk music was dead the first time someone said punk's not dead.
I could see that being the case with a trend/fashion, but it's not so much of a problem with other things.
With respect, the meta of late has been extraordinarily aggravating, and I do think it is a problem here.
Borrow a verse from the book of MetaFilter: people need an area for meta discussions, and it's better for the signal:noise ratio if it's in a separate but equal space.

In fact, there's lots to be learned from MetaFilter. Two long but useful videos containing Mathowie's wisdom:

- http://vimeo.com/11916466

- http://vimeo.com/21043675

I don't think you really can. This site is a community and the users matter above all the features. If the user quality takes a nose dive all you can do is hold off the inevitable with new comment rankings. Every site has a point where it gets so big it declines in quality, reddit hit that and now those who want the old reddit back are coming here.

The only way to truly guarantee it would remain high quality would require credentials to use the site, or require invite/referrals, but then that has a whole host of its own problems.

I'm relatively new so I don't know what HN "used" to be like, but in the short time I've been here I've noticed it decline. It seems to me that more and more people who aren't knowledgable or have insights to offer are joining and people like jacquesm and riderofgiraffes are leaving. It was inevitable and has happened in every community I've ever used.

(comment deleted)
(comment deleted)
I disagree, changing incentives can radically change a way a site is used and interacted with, just like many things in life. The nature of the incentives change the nature of the comments/submissions.

I think that the reason that it has happened to other communities you've seen is because they all had similar incentives. Also, I believe that requiring credentials to use the site (or using referrals/invites) like you suggest is in fact changing the incentives, as people are more accountable to a smaller community/feel more compelled to have a higher standard due to being invited there, rather than just using it as they would public infrastructure (see how people treat a public WC compared to how people treat toilets in homes they've been invited in to).

Stop showing people other people's comment scores. They stimulate argumentative comments.
I do like that idea. I think I tried it a while ago, but users complained that without comment scores it was too hard to figure out what were the most interesting comments in a thread.

Is there some other way to show what the most interesting comments in a thread are? How about if I displayed the point totals for subtrees, but not for individual leaves? Would it solve the problem if you could follow other users, and see their comments graphically distinguished in some way?

Another problem is that people use point scores as a guide to voting. It's clear from voting patterns that many if not most users vote not to express approval or disapproval, but to cause the comment to have what they believe is an appropriate number of points. If I didn't display points, people couldn't do that. Perhaps that's not a problem. But if it turned out that that's what voting was for, then this could break voting, which would in turn break the sorting of comments, which would be a problem now that there are so many.

Gut: anything that reduces the number of comment scores shown on a thread will to some extent reduce the number of dumb comments. It's a knob. I suggest you turn it. I'd turn it to 11, but 5 might do nicely.
I'd find a way to show what was a good comment with something besides points. Show that flag, but still hide the number of up votes. This flag would tell you whether it was a good comment, but not exactly how good. That's what karma points do well
>Is there some other way to show what the most interesting comments in a thread are? How about if I displayed the point totals for subtrees, but not for individual leaves?

How about dispaying the rounded log(score)? You give some indication of how well a comment is doing but it's hard to vote strategically.

Not that easy with negative and zero scores, obviously, but those usually aren't the comments people are looking for.

That is a great idea.
No offence to pg, but why did this comment get upvoted 9 times?

This makes me think it would be good to remove the name of the commenter as well as the number of points a comment gets.

Names are important because they allow you to use your own internal karma system (I will read comments from certain people no matter what their score). One problem mentioned is the overloading of what an up/down vote is - what about separating "good/flag" and "agree/disagree"?
Upvotes are sometimes used to indicate agreement.
Yes, but in this case the upvoters could have simply upvoted the parent comment pg was replying to.

Short "i agree" or "i disagree" type of comments are usually discouraged on HN, as they don't really contribute to the discussion. Usually.

pg's comments aren't differentiated in any way, but on discussions of potential new features, I'm very interested in seeing his comments. Even if they don't add on to the ideas being discussed, they're of interest because they indicate possible new features or changes to the site.

Also, username hiding is a good idea in some cases, but I think it's worse overall. It encourages snarkiness, some posters are actually worth looking out for (grellas on legal threads, pg on HN, etc), and it makes back and forth discussions difficult to follow unless you give each user a unique ID per thread. Impersonation can also be a problem. Overall, I think it adds up to more problems than it solves.

I upvote comments if I am interested in reading them. I'm likely to be interested in reading anything that pg writes because he runs this site. The bar is set rather lower for pg than for any other user and for good reason.
> Not that easy with negative and zero scores

Started logs. [f(x) = log(x+1) or log(x+3)]

You could show a simplified indicator of Great Comment, Good Comment or Bad Comment (or a similar set of values) so that users can skim a thread effectively. It's hardly a model community, but I think Engadget used to have a system like this.
This is basically an extension of limiting the display of negative votes to -4. I like it.

Perhaps there should be a "best comment in thread" indicator or something as well?

Take it step further and show 4 different smiley faces, a la 37signals.
I like the idea of showing point totals for comment threads. Entire threads (and subthreads?) with the highest total points could be near the top, which would indicate that those are the most interesting without revealing individual comment scores.
What if comment scores aren't related to karma, but rather the position of a comment within the context of the story. You up or downvote a comment to move it up or down. The effect would be to rank comments based on their perceived value rather than score.
If you remove the vote count, that doesn't mean you've removed the sorting by vote, right? Isn't that enough information for people to find interesting comments?
No, the software would still know the scores, and use them to sort comments. The scores would just not be displayed.
I think people can figure out for themselves what an interesting or useful comment is on their own without points. We need to give people some credit. The use of points to rely on what others think is valuable, useful, helpful is irrelevant. At least for me. I can give a darn what other people think is useful. Maybe I find some insight useful where everyone else doesn't.
> Is there some other way to show what the most interesting comments in a thread are?

a vague indicator instead of discrete values. People enjoy seeing their karma for a comment tick over from 3 to 4, but seeing a little thermometer rise by 1px is less noticeable and isn't so much something you can keep track of.

When you see your comment score increment, you know it's because a real human decided it was interesting - that's a good feeling, but it can encourage karma-whoring/playing karma like a game.

Maybe a refinement of this idea: show me the precise number of votes on my comments, and a vague log(score) or good/medium/bad grade for everything else.
users complained that without comment scores it was too hard to figure out what were the most interesting comments in a thread

I was one of those complainers. And that was then; this is now. I'm not sure I'd complain if the points went away again.

Today I find that comment scores are no longer a reliable guide to anything I care about. Something has changed -- most likely just the scale of the whole system.

So you have two problems: users need to be able to figure out what are good comments to read, and correspondingly you want users to vote on comments that make good reading.

Have you thought at all about discerning between vote value based on some metric? You'd need something that doesn't disproportionately disadvantage new users.

Elsewhere in this thread I proposed a scheme that meets both of your requirements: it discerns votes based on a metric (how closely the voters vote to your own preferences), and it doesn't disadvantage new users:

http://news.ycombinator.com/item?id=2405266

How about adding a star to the top log(n) comments in each post?
If you turned off gravity for comments, then ranking would indicate their relative votes. Users could vote to create what they believe is the appropriate ranking.

I also think quality was higher without gravity for comments. Now, I scan comments to find the highly voted ones. Do you think gravity for comments was a net gain?

BTW: Is it possible to objectively measure the impact of changes, as opposed to be being persuaded by articulately argued complaints?

> Is it possible to objectively measure the impact of changes, as opposed to be being persuaded by articulately argued complaints?

Yes, if and only if "good" and "bad" comments are defined (intensionally or extensionally).

Hybrid: the first comment by hotness (as now), to give up-and-coming comments visibility.
would displaying them just like the links on the front page not work? are they not currently working like that anyway?
(comment deleted)
A hard ban on politics and current events, instead of the wiggly one we have in the site guidelines now.
I second this - the politics posts do get a lot of hits, but HN needs to decide if it wants to be "awesome" or "popular"
Gotta agree.. I sometimes fall into commenting on or upvoting politics threads, but that's really not what HN is about... a harder line against that stuff would probably help.
Politics, I'm definitely with you. "Current events", though, is an extremely broad term -- what exactly did you intend it to mean?
The guidelines currently say "if they'd cover it on TV news, it's probably off topic". The predicate should be clarified, and the word "probably" removed.
The problem with this is tv news is uneducated for the most part. People end up posting things to HN that don't fit well, because it doesn't get discussed in other venues. So yeah, you're right though. It's got to be a hard line.
Can we say use common sense? They'd cover Facebook IPO on TV. Yet it would have an impact on a lot of small startups that frequent here.

"Common sense isn't all that common"

(comment deleted)
Wouldn't that ban a lot of tech companies' current actions? Ie, isn't everything happening now in Silicon Valley etc a <i>current event</i>?
Going after a specific topic misses the point a bit. A general ban on topics that tend to be driven by emotion would work better.
Unfortunately a lot of discussions about technology are driven largely by emotion.
(comment deleted)
Talk to any startup founder or early employee about the emotional roller coaster they went on. It's an important part of it.

Start up and hacking are emotive subjects.

And things where discussions are occasionally productive. You can be emotional about BSD, but it's possible to have a productive discussion about BSD vs. another OS.

Have you ever seen a typical political debate? Those are driven by emotion. They're full of catchphrases and insults, and nothing of merit.

The problem with that is that it's hard to say exactly what counts as politics. E.g. is an article about economic inequality like this one

http://news.ycombinator.com/item?id=2402027

politics? I don't think so. Economic inequality is a very interesting social phenomenon, though it is often written about in a heated/political way.

It's interesting, but is it useful for a busy founder?

Or to make that more concrete, would you forward that link to multiple portfolio CEOs and suggest they take time out of their day to read it and discuss it?

Not sure what 'founders' and 'CEOs' have to do with hacker news.
(comment deleted)
I assumed that the target market for HN is people in startups and people about to start startups.

That said, if the target market is just 'hackers', just modify my question to refer to 'hackers' rather than 'founders'.

HN is for hackers, not startup founders. There just ends up being a disproportionate amount of stuff about startups, because YC is in the startup business.
Clearly not everybody reads the site guidelines. Perhaps enforcing that periodically would help.
And therein lies the problem. As HN and Y Combinator itself have grown, as well as the general startup climate, HN has become increasingly startup focused to the point where many would describe it as "for startup founders" rather than "for hackers". And there are a lot of new users who like it that way (or assume it's supposed to be that way). But then you have the old guard like tptacek who would like nothing more than to see HN filled with only hard tech/science posts. That's a gap that isn't going to be filled with any algorithmic tweak or new comment layout.
I would strongly prefer an HN that only had startup posts. I think this part of the thread is going off the rails, though. The problem is, "how do you improve comment quality on HN". We can probably avoid debating the premise of the question, and just focus on the (plentiful) ideas themselves.
(comment deleted)
(comment deleted)
(comment deleted)
Hey, when I joined it was called "Startup News".
This tiff seems to highlight what's perhaps the problem: is there really agreement & understanding on what HN is supposed to be about?
I think political stories and economic trends that relate to startups would also be appropriate (for example startup visas, growth of middle-class in China) as these are area that impact startups in the long term.

Obviously if the discussion on these threads is purely political then they probably shouldn't have a place here.

I don't see a guideline saying something has to be useful or forwardable among CEOs.
I'll go ahead and suggest, since it's on topic in this thread, that passive aggressive snark like this is one source of HN's decline.

One can disagree without being snarky and disagreeable.

How was it passive aggressive or snarky?
(comment deleted)
(comment deleted)
Perhaps it was intended as a non-snarky, non-aggressive statement, but I still have trouble reading it as anything other than passive-aggressive snark, especially given the context of the conversation.

I find myself wishing I had a 'meh' flag for both your comment and my reply. Not quite a downvote or a report, but... "meh... this is not a useful addition to the conversation."

The downvote button is the "meh" flag! Why would you hesitate before downvoting such a comment? Slay evil instantly!
It's interesting, but is it useful for a busy founder?

Aren't founders as busy as they are in part because they hope to reach the top 1 percent of income in their country? I would want to know what my country's policies are toward high-income people, and what social trends are influencing those policies, if I were pursuing a high-wealth business result, as at least some founders are.

politics? I don't think so. Economic inequality is a very interesting social phenomenon, though it is often written about in a heated/political way.

The comments should count as much as the articles do in determining if a particular topic is allowed on HN or not. Because the comments are just as important. To me, at least.

A big part of why I first started reading HN is the the comments in the articles. I could find the articles in half a dozen other places. Still can. But, the comments here aren't usually anywhere else.

And the best comments are typically better than the article itself. If the comments turn political and into people arguing politics with one another, thats not interesting to me anymore. I don't usually find the article being commented on to be very good, and the comments are even worse.

Seeing people argue politics just makes me close HN and find something else to use to kill the next few minutes. I've actually avoided reading articles that I likely would have otherwise found interesting just to avoid seeing the vitriolic comments that inevitably follow.

> Economic inequality is a very interesting social phenomenon, though it is often written about in a heated/political way.

I think that if you look at the resulting conversations on those sorts of articles, it usually retreads extremely familiar territory.

Also, 90% of that kind of article I've ever seen tends to be someone "broadcasting" a view they agree with ( "hey, this guy's right on!" ) rather than presenting actual new research in economics.

Also, any article like that tends to be something anyone can have an opinion on, tending to attract people that are more interested in those sorts of discussions than about "hacker news".

IMO, at least.

I don't actually think articles on "Economic inequality" belong on hacker news. In fact, it's articles like that which are a big difference between this site in the early days and now. I'd prefer a site that stuck to technical and startup business topics but perhaps that isn't hacker news anymore. The fact that these articles get voted up implies that I must be in the minority.
We could have choice, would have to be in conjunction with either a tagging system or a completely separate list like subreddits. With tagging in your profile you could have things to opt out of like politics and current events and tech gossip.

The when these submissions popup all you need is someone to tag them as one of these things. You could either assign people to do this or weight it by karma or something.

Technically it just means you're in the minority of people who upvote things.

Hang out on the front page more and vote there, like RiderOfGiraffes mentioned in his farewell post (or the ensuing discussion, can't remember). It is really important that if you are dissatisfied with the quality of HN (as I currently am) you should do something about it.

imo the "anything that gratifies one's intellectual curiosity" guideline and related submissions add a lot of value to HN.

Besides, I don't agree with the notion that mean-spirited or dumb comments or ideology-based downvoting/upvoting are limited to politics and current events. Any thread about iOS, Android, Microsoft etc. is also likely generate a lot of those comments and ideology-based up/down voting :)

Require articles about (say) economic inequality to be scholarly.
tptacek, it would be great to require everything on HN to be "scholarly" or sufficiently serious, no? But the question is how to do that.
Simple; they would need to be published in journals, or have bibliographies, or have a cite record. "Scholarly" isn't really a subjective term.
I believe scholarliness is not simple. Examples that muck up a naive attempt at defining "scholarly":

. Econophysics, fuzzy logic, systems theory -- are those really serious fields? Could they become serious? The answers are subjective.

. Interdisciplinary Journal of Problem-based Learning -- too specific? is this just a journal that someone started to enhance her reputation in a self-defined field? how long has it been around and how long will it be around?

. Douglas Hofstadter -- scholar or dilletante?

. the blog of Terence Tao -- it is serious but is not published in a journal

. http://arxiv.org/find/math/1/au:+Jormakka_J/0/1/0/all/0/1 -- a guy who claimed to solve more than one Millennium Problem and is widely considered a crank. Articles were published in journals.

. http://arxiv.org/find/math/1/au:+Perelman_G/0/1/0/all/0/1 -- a guy who did solve a Millennium Problem. Articles were not published in journals.

Listen to Frank Wilczek talk about his feelings about not having his genius recognised: http://www.learnoutloud.com/Catalog/Science/Physics/The-Univ... (Q&A at the end).

Academics play the karma game with much higher stakes.

I would also add that topics about problems with possible software solutions are also relevant.

Problems in politics have led to startups like Votizen and OpenCongress.

Where there is a problem there is a startup. And where there is a problem that people are willing to pay money for a solution there is a business model.

I would say: yes, that's definitely politics, and it should definitely be banned. I'd be prepared to use a fairly broad brush in determining what "politics" is.
What about adding a separate "flag as politics" option to stories? That way, if enough people feel that it is off-topic for the HN front page, it could get banished to a separate news.ycombinator.com/politics area. Not deleted, just removed from the front page and easily accessible to anyone who wants it.
I really like your suggestion. It could be extended even further.

If readers are willing to tag stories not just as "good" or "bad" but with categorical labels, then suddenly the news aggregator "knows" a lot more about the data its serving up. You could have categories at the top, preferential sorting for users, and more.

I came here to say this and can't believe I had to scroll down so far. All you need are accurate tags and a way to filter by them. There is no point punishing people for submitting or voting up something the community is not interested in. Just make sure they tag it correctly so that it can be found or ignored according to reader's preference.

Note this discussion is exactly akin to deletionism on wikipedia – if we can categorise information effectively, why reject any of it?

Facebook / Twitter analysis, if it has 10,000 shares on Facebook/Twitter chances are we can find it somewhere else. Would any of us have missed that Color raised $41 million dollars? Probably not.

maybe ensuring the articles are at least a week old, that instantly removes 99% of the current events which are generally irrelevant and if it is relevant it will still be so in a week.

You might want to consider just shutting it down for 2 months, the great users will likely come back. It will definitely lower the dunbar number quite a bit.

Just a ban on "current events" (aka news)?

Past political essays, e.g. by Cicero, will generally be intellectually interesting.

Perhaps HN could ask the submitter to include a short justification as to why a busy hacker should take time to read this article.
I can't see this working, but heck, I'm willing to give it a try.

Two main problems: 1) everything is politics and current events. That is, everything touches in some way on what's going on in the world and how you feel about it. Maybe I'm applying too broad a brush, but I can see politics and current events in every post on HN -- and always have. About the only exceptions would be the driest of technical articles. If you want a board of Erlang innards, go for it. Other than that, it's always going to be Steve Jobs, EFF, which VC trashes which other, etc -- all gossip. (And gossip is just a broad term for current events and politics)

People post, comment, and vote based on emotional response. You can pretend to cut that out by banning, say, any mention of political parties or politicians, perhaps any pending legislation, and past legislation, how economics prevents or helps startup creation, etc -- but I think you'll just trade one monster for another.

2) I've been thinking about this for a while, and the key problem here is that the commenting system promotes learning how to fit into a community of hackers, not actually doing anything useful. We think of the system as being some sort of logical function to take all kinds of input and provide the "best" stuff for hackers, but it's exactly the opposite: it teaches hackers how to get votes from other hackers. In other words, spend 3 or 4 years on here, like JacquesM did, and if you're lucky you become an expert on what to say and do to make large groups of hackers agree with you. The karma system and large crowds isn't training the material to be more targeted to hackers -- it's training hackers on how to target other hacker's response with their material: how to fit in, what to post to get a rise, how to succeed without really trying, etc.

You can blame the topics for being too emotional, but I think it misses the point. We've always had all kinds of issues on here. What we're getting is a larger pool of people actively gaming the system in hundreds of ways: what kinds of snide comments might sneak by, what's the best time to run flame-war submissions, who to praise, what topics to champion or deride. That's a function of forum size, not system parameters. And that's not even getting into the fact that with a large enough audience, somebody is going to get pumped about just about anything that appears on the front page.

There's nothing nefarious about this. It's all just people being humans. After all, if you are in a conversation, do you speak in order to provide factual information? Or do you speak based on your current emotional state and to effect the emotional state of others? Why would we expect HN to be any different?

Nyah. That proposed rule -- if you could somehow miraculously define it with the precision required -- wouldn't give you the results you desire.

> What we're getting is a larger pool of people actively gaming the system in hundreds of ways ...

We're getting politics, in short, even if by a different name.

(comment deleted)
How does that address comment quality?
There's more than enough hacker-appropriate linkbait to fill the front page, if that's what the readership wants.
Especially, a ban on "startup news". e.g. AngelGate was not intellectually interesting, but it was important industry news.

Eliminating the exciting news of the moment would make the site less entertaining for those not seeking intellectual stimulation. It would also make the site less popular. Both are good.

A ban would logically include YC company launches - revealing the conflict between YC's business interest and intellectual interest. It would be a defining moment, for what HN is.

"Startup news" could be a sister site, if important.

(comment deleted)
The comment flag button could be changed to really mean something; for instance: sufficiently flagged comments can stop collecting upvotes.
Would also be nice if the flag button were visible without clicking the comment.
Sounds like it might lead to abuse. If a comment is -that- bad, it should be worth taking the time to flag it.
That would probably end up like the "report" button on reddit. Instead of being used to report spam or inappropriate content, it's often just used as an additional downvote, thus devaluing it. Keep flags difficult to use makes sure that the user really wants to flag a comment.
Maybe flagged comments could be killed in a similar manner as flagged stories are, rather than just stopping them from getting any further upvotes.
That wouldn't do anything currently. Only really atrocious comments get flagged, and they always have huge negative scores. Though of course if flagging had more effect, maybe more people would do it.

Another thing I've considered is having specific types of flags on comments, and having them have different effects. E.g. there could be a flag for incivility, and if you got enough of those (maybe in proportion to your total number of comments) you'd actually get kicked off the site temporarily.

Something to keep in mind though is that, as the size of the user base here grows, there's going to be an increasing number of people who feel comfortable using "this button will contribute to banning this user"-type controls less than judiciously.
Make flags public, and their outcomes. "16 people flagged this comment." "16 people flagged this comment, and it was tagged uncivil. User was banned for 2 days." "3 users incorrectly flagged this comment, and have lost flagging privileges for 2 days."

alex is right that automatically escalating based on #flags is a blunt instrument. You'd need to perform the escalation manually, you'd need multiple people doing it, and you'd need the decision-maker to attach their name to it ("kn0thing marked this uncivil") so the watchers can be watched in a lightweight manner.

One transition from "current" state to "flag does something" state might be to make the flag button more obvious, to make it have a more immediate effect, and to encourage its use.

My perception of the flag button now is that it's the bat-signal; it means that something is so bad that it actually warrants individual admin attention.

One of the Dutch news sites I frequent also has qualitatively different forms of upvoting. E.g., you can upvote for humor, upvote for insight, upvote for new information. By applying different metrics to different kinds of comments, it might be easier to create the kind of balance you want.
What site is that? I'm curious.
When I see something dumb and/or mean that has been highly upvoted, I don't consider flagging it, because I don't think it will have any effect.

However, I don't see an algorithmic way to use flagging in such a situation. The new control mechanism would be subject to the same problems that led to the comment being upvoted in the first place.

I'd like to see the mods experiment with a more authoritarian approach. When a comment sucks, I want the mods to impose their will upon the populace, informing everyone that the comment sucks, regardless of its upvotes. Flags could play a role.

There's a lot of discussion here about comments and votes, but only one or two remarks in this thread note the clearest differentiator I see between "old guard" and newbie users: the meaning of the up/down arrows.

Old guard would prefer to vote on a "Contributes/Detracts" axis, while new users vote on an "Agree/Disagree" or "Like/Dislike" axis.

Arguments about this erupt in threads, with newer users generally saying, "It's a democracy, this site is what the majority want it to be, and if most of us want it like this, your loss." But such an approach devolves into pop pablum.

I believe most other ideas here would be unnecessary if the meaning of the up/down arrows could be resolved either socially or algorithmically.

The "correct" solution would be to offer a quadrant, with contributes/detracts on one axis, and agree/disagree on the other. But that would require a rewrite.

Instead of a rewrite, I'd experiment with temporarily changing the arrows to say something explicitly supporting well reasoned comments:

  17 points by uptown 1 hour ago [ contributes | detracts ] link | parent | flag
Some would undoubtedly still interpret these as a rightness axis like "agree/disagree", so perhaps an even more familiar pair of terms:

  17 points by uptown 1 hour ago [ content | spam ] link | parent | flag
However, the term "content" might lead to voting up every valid content remark, so the positive word should be something with more of a value judgment, while still being a word that can apply to points of view with which one disagrees:

  17 points by uptown 1 hour ago | interesting* | spam | link | parent | flag
The idea with the vocabulary choice is that a neutral comment would not be clicked on, and "interesting" is directly in the HN charter.

I think this type of vocabulary is more in line with the desire to see well reasoned or contributory discussion flourish.

* Mouse over the word "interesting" could tooltip: "This comment made me think."

// This account is ~400 days old, but a prior anonymous account is ~800 days old, giving some perspective on the trend over time.

EDIT: Single click, long delay, resulted in double post. Edited this one out to save page space.
I think the idea of categorizing comments is good. I would cite as a somewhat-working example Wikipedia's article flags. Wikipedia has a serious problem with deletionism, but they have ways to mark articles in meaningful ways, like too long, or too short, or irrelevant, or excellent, etc. They can also mark individual contributions as "citation needed" (a flag I would really like to have on comments, see http://news.ycombinator.com/item?id=2423582 as an example of a citation-needed comment getting, I feel, way too many upvotes).
This is a separate issue, but one thing that is not perfectly clear to me is what an upvote/downvote is supposed to mean. Does it mean that I agree with the comment, or that it adds to the discussion?
I think it would be an interesting experiment to add clearly labeled agree/disagree buttons in addition to upvote/downvote.
How about notifying people when their comments have been flagged and pointing them to the site guidelines?
Some policy/feature/system to aggregate related stories ("killing" stories that duplicate stories that already have active threads, and posting a link to the "duplicate" story in that thread, or something similar to that --- I'm being minimalist here).

A lot of dumb comments appear to germinate on threads that are the 3rd or 4th take on some tech news story about Facebook or Apple.

To me, this is the biggest problem with HN rather than comment quality. Tons of related, and even duplicate, stories happen all the time which fragments discussion. Not to mention the duplicate posts which happen a few months later. Thankfully someone usually remembers and posts a link to the previous discussion but that shouldn't be required.
I think it's an interesting idea to be able to "group" related posts that occur within a certain time range of each other. If five posts on the new iPad 7 come in within the same three days, someone can drag them all together into the "iPad 7" thread.

But this?

the duplicate posts which happen a few months later

Once we're up to a timescale of months, or even weeks, we're no longer being sensible. Instead we're exhibiting FAQ Syndrome: The irrational fear that someone, somewhere, is saying something that isn't entirely original.

I think the cult of originality is actually a big problem at HN, and other "news" sites as well. The important things in life are not particularly original, and they do not change particularly quickly. A site that is determined not to re-discuss previous topics is doomed to discuss nothing but ephemeral trivia. The great thing about celebrity gossip is that it is always new! We can manufacture celebrities at whatever rate is needed to keep the front page fresh. But we can't manufacture Knuths as needed; we've only got the one set of Maxwell's equations; new books on the scale of K&R or SICP don't come along every day. But if we discourage the constant reexamination of these classics they will get placed on the dusty shelves and we'll see nothing but discussions of the latest gossip and bling. You know, like we have today.

I always wished HN would feel more like academia, which cycles like the seasons. Every year, you discuss all the classics again for a new audience of newbs. After a little while, you've heard all the classics and are ready to graduate, or become a professor. This is what makes me miss the days when this was "Startup News" and was more explicitly tied to the YC cycle, the time when you could tell that a new YC class was starting by watching for the influx of new people.

ISTM you're thinking of merging duplicate content as a punitive measure. I think of it as a mostly positive thing: I wouldn't mind seeing a discussion on SICM once a quarter--but it'd be nice to have all the older insightful comments on it readily available.

Perhaps there should be a "hard merge" that prevents duplicates from sharing the front page, and a "soft merge" that multiplexes comments from duplicates at any time range under any of the submissions?

"Constant reexamination of classics" sounds like bliss to me. I would love to see more of it. nostrademons' comment (http://news.ycombinator.com/item?id=2403317) about feeling like he's absorbed as much as he's ever going to get out of 15 minute blog posts, and hungering for deeper learning, is one that deeply resonates with me. I remember books. The last 5 years, say, of internet use has retarded my learning, and I'm ashamed for letting it happen.

I wish there were places to go online to teach or be taught about topics in a discussion-forum kind of way. I'd love, for example, a Lambda the Ultimate for newbs. LtU is an awesome, high-quality site, I just wish I could understand more than 10% of it.

I tend to use HN when I'm too tired to work or just waking up. Those are also the times when I'm not interested in buckling down and studying a hard subject. But surely I could do better than the kind of learning I get from HN (and other casual internet use), which is random access to shallow, scattered tidbits, as if training for a massive Trivial Pursuit tournament you never signed up for. If only there were a community where people devoted themselves to teaching and learning classic topics in a sustained way, but that still had the social and casual aspect. News as such doesn't interest me that much. I'm mostly just consuming it by default. If anyone's read this far - my apologies for adding a mostly off-topic comment to an already crowded thread.

Maybe give high karma users the power to merge/suggest/link this stories, as Stack Overflow does, and make HN moderate itself better.
A related problem is blogspam. If the original source material was submitted a while ago but didn't get many upvotes, there is a huge incentive for people to mirror the content or link to a site that links to the original.

I think a more useful metric than "number of votes since submission" would be "number of votes in the past hour/day/whatever." That way an item could be submitted days in the past, but still get on the front page if it had enough upvotes recently. For testing purposes, this could be an alternate front page like http://news.ycombinator.com/classic

Another option would be to have moderators manually de-dupe stories or fix them to link to the original source. While they're at it, the mods could ban or reduce the karma of users who submit blogspam or duplicate stories.

This plays into RiderOfGiraffes' experiment with leaving comments that tied together related (or identical) discussions, which I was rather sad to see him discontinue (especially in cases where two threads were discussing different aspects of a story). I don't recall, though; was he automating that, or was it a purely manual effort?

I suspect a "closed as duplicate" system, ala StackExchange, could work well here, assuming a high enough bar were to be set for who could do (or vote for) such a thing.

He mostly did it manually. He had started to pull a system together to automate it, but (iirc) some people rather resented it, so he discontinued that.
Have you considered adding a down vote button like Reddit has? I know you can flag comments above a certain karma level but I think that either giving everyone the option to down vote or having a lower karma threshold to down vote would allow the community to regulate itself.

Just a thought, but it seems to have worked for Reddit. This puts a lot of responsibility on the community to keep the quality of the discussions up, but I think enough people on here care about the quality of the community to help out.

IMHO, that hasn't worked so well for Reddit.
(comment deleted)
Isn't that what downvotes are supposed to be for?
(comment deleted)
With an auto-flagging system, a comment might get stomped out by a small number of trolls.

I kind of like how YouTube does their voting system (showing both the number of upvotes and the number of downvotes).

(comment deleted)
I don't think so. Up vs down is a 1-D dispute. Categories add dimensionality.
A privmsg feature, available to people who cross a karma or karma average feature, that would allow gruseom to tell people offline that their comments are dumb. Sometimes it's good to make an example of a dumb comment, but other times it just begs for an unproductive fight.
Interesting idea. That would be pretty easy to implement: it could be an ordinary comment, but that would only be visible to the sender and the receiver.
Note that this won't "fix" the comment problem, but every time it is used, more likely than not, it's going to improve the comment threads, because the alternative is publicly calling someone out.

Please label it "the gruseom button".

You've called me out in public before. Shame is an effective teacher.
Except with sufficiently snarky/provocative comments trying to shame the commenter might look from the outside like something very similar to petty disagreement.

I think some people here are learning that hugely self-confident, strongly opinionated, obvious writing tends to spark a strong agreement reaction on the readers, who quickly upvote a comment that adds nothing to a discussion. These are a problem, as they encourage snark and posing over effectively arguing things out, but they are very hard to treat as disagreeing with them is likely to cause knee-jerk reactions in many upvoters.

> Except with sufficiently snarky/provocative comments trying to shame the commenter might look from the outside like something very similar to petty disagreement.

While I agree with the general case, in my specific case I called somebody an "idiot" out of anger and tptacek called me on it. It was quite mortifying and I've tried to watch my words since.

Strongly agree with where you're going here. Every time I see a "senior" or "better" HN handle call someone dumb, I think: well, who says you're so smart? Who makes you the dictator of "good" comments?

I never see "nice" tellings-off. A "nice" telling-off might be: "reader123, this comment is mean. please be nicer"

That is exactly the sort of telling-off I got. It was polite and to the point. After I edited and apologised tptacek deleted his remonstration.
(comment deleted)
I was thinking about this in the car and came up with the same thought, that it could just be a comment that no one but the OP (and mods, perhaps) saw. I am definitely guilty of posting some bad comments as far as the HN definition goes. I blame living on USENET in the late '80s and early '90s and the WELL, but the reason is immaterial. The problem is that many times I don't really see the comment as particularly bad, or I haven't thought about how someone else might interpret it, didn't realize how I was being dumb, or I simply forget the tone of the venue I'm in.

When someone points one of these issues out, I often feel bad about it. However, I usually don't edit my comment to remove or reword the offending portion because I think it's rude to the person replying and, oddly enough, those who later read it. Because they've made a good point, if I fix up my post it breaks continuity, possibly makes them sound like an idiot, and feels like I'm trying to cover up my lameness.

If I got a similar note that was private I'd feel free to make the changes without as many of those worries. Especially without concern for how they'd take it as presumably they're looking for me to shape up not score points on my misbehavior. I also think that some people that might say something are loathe to reply publicly in fear of just making it worse.

Of course it could seriously backfire if people used it just to be mean back without fear of the community observing. But it seems like it might be worthy of trying.

I really like this idea. It'd disincentivize trolls because they're looking for a public flamewar. Turning those flamewars/disagreements private would hide a lot of the noise.
My thought was to have a selection of canned responses that encourage this sort of feedback without allowing it to devolve into arguments.

But anything that encourages feedback and discourages defensiveness will be effective.

Yikes, if my name is the pronoun for that behavior, I must be doing it way more than I thought I was!
You're just one of the people who I go out of my way to read, and among those people you're the only one who goes out of their way to help police thread quality. I'm sure there are people who do it more.
Perhaps a way to include a private note with a downvote, so that downvoted commenters can better understand why they were downvoted (without that meta discussion being public).
Maybe a 2D vote. The up-down dimension we have now, and another to say why the up/down vote. This would combine the up/down with a Slashdot-like Troll~Insightful dimension. That would be some useful feedback.

Sometimes I don't know if I am being downvoted because I am being too vicious (I get carried away from time to time) or because I am just completely wrong.

This is by far my favorite suggestion, although I think that the messengers should be hand-picked, not based on karma.

I make a lot of lousy comments, because I don't really have a lot of respect for the quality of discussion here. But, you know, if I got private feedback a few times -- even from people who were powerless to punish me -- that my comments were bad, then I would actually stop.

> A privmsg feature, available to people who cross a karma or karma average feature ...

I don't agree. You'll just create another incentive to game for karma and that's what I've seen at sites with karma-linked features.

I'd suggest that either

* nobody has the feature

* everybody has the feature

* only paying customers have the feature

I disagree that the quality is declining. I think you're just suffering a misapprehension of the quality of old.
Comments are certainly declining in quality. For example this thread from a submission I made a year ago:

http://news.ycombinator.com/item?id=1215549

The quality of discussion was much higher than on most posts today. And that was only a year ago.

Depends entirely on the post. Esoteric technical issues (as in the business model of advertising) have a higher level of debate than more relatable articles.
Compare it to some of the recent articles on Paywalls, there's almost no discussion of business models left.

Instead you have people ranting "content should be free", "paying is an outdated model", "the nyt are stupid" without any kind of coherent argument.

I was writing an open letter to HN on my blog for this topic, but this now seems a more appropriate place to reply (apology about the style which seems out of place in a comment):

Once upon a time Hacker News was called Startup News, it was a place to share links and discuss between people passionate about startups. Good links and discussions stayed around for days, every aspect of startup life was discussed.

Sadly that time has long gone. As I write this, on the front page of HN there are maybe 4-5 stories out of the top 30 that relate to startup topics.

Articles relevant to startups are being pushed out by generalist tech and programming articles that are better served by the many many subreddits on these topic. While it's open to debate whether these are on-topic on Hacker News or not, HN is far less about startups than it used to be.

Many contributors to HN don't even see it as being about startups anymore, even contributors who've been involved in HN for over a year are talking about it as a tech or programming site. The startup stories that reach the frontpage tend to be on technical topics, the non-startup tech audience of HN now means stories focused on the non-technical aspects of startups such as marketing and raising money make it to the front page far more rarely than they once did.

I remember complaining at one point about the number of stories about A/B testing on the front page. I wish I could complain about that now.

Take a look at Gabriel's Ask YC archive - it was created to address the startup questions that frequently turned up on HN, for many of these topics I can't recall when I last saw them discussed on HN.

There are a hundred social networking sites that serve the tech community from proggit to dzone, what differentiated HN was the focus on the startup community. That focus is dying out, and we're becoming just another tech social news site.

I don't think we can make HN be more about startups again, the audience has changed too much for that, and it wouldn't be fair to the non-startup tech community that's come to rely upon HN.

So instead I'd like propose that HN stays as it is, but pg creates a new HN called Startup News, which has startups at it's heart as HN once did.

It was certainly more interesting then; I wonder if it's because the community was smaller or because the topics were more focused.

I don't value HN as a resource for entrepreneurship anymore. There are some interesting technical conversations here still; but you can find those all over the place.

(Of course there is still a lot of information of value for entrepreneurs here... it's just buried under the flood of everything else.)

So, rename it from "Hackers News" (which to me has meant "Slashdot that doesn't suck" because that's the type of posting I've observed here) and back to "Startup News." I'll bet most of the non-tech-startup-biz nonsense will go away.
Presumably the name changed to reflect the fact that there are many things unrelated to startups that the founding "hackers" wished to discuss. HN isn't entirely about startups now, in the same way Amazon is not purely about books any more.

You seem to dislike the non-startup material, yet those older members who were around at that time must have liked the discussions sprouting from the "more generalist tech" posts, else the name wouldn't have been changed.

Thus, I reason that it is not the range of topics submitted that is the problem, but the quality of the posts and subsequent comments. I believe the older members valued intelligent discussion on any topic (centered around tech-startups).

I think a solution should concentrate on improving the discussion of topics, promoting those that spark the "best" (for some definition of "best") conversations. Therefore I suggest:

More liberal use of the downvote button by those that have the ability (over 500). Number and score of comments should play a (larger?) factor in ranking stories. More aggressive moderation of "off topic" or vacuous submissions and comments.

When the name was changed to "Hacker", the term Hacker wasn't used in the tech sense but rather in the same sense that it's used in the YC application form. As in a clever unorthodox solution to a problem.

But over time people took it to mean hacker in the technology sense of the term, and thus we now get reviews of Ubuntu on the front page, which you would never have seen a few years ago.

Make it invite-only to post threads/comments and quietly associate the inviter with the invited person.
Thus ensuring that we don't get spontaneous comments from e.g. the UX guy at Zappos on that story about Zappos. A non-starter, I think.
There will be pros and cons to every system. How else do you think we can decrease the growth rate of the HN userbase, which is probably the root cause of the decline in quality comments and threads
That could be worked around, contributing could be invite only, but also allow a non-invited contributor to post but their post would be collapsed (or at a negative score) by default until an invited contributor voted it up.
Won't it make it just an echo-chamber that would increase group-think? I like the idea that new people can join and contribute or learn from others without having to first curry favor to get invites.
That will certainly ensure only the acceleration of the decline. The ecosystem will be isolated and perpetuation of 'preaching to the choir' will ensue.

How would have all of the now-members of both HN and the greater 'community' have stumbled upon this resource if it was invite-only?

HN would still be open for everyone to see, commenting and submitting threads would be invite-only.
Thank God you've noticed. I seem to recall that you brushed off this observation for quite some time.

One thing I've noticed repeatedly in the online communities that have scaled succesfully (in a cultural sense) is that the founders/owners/admins tend to take a very active role, both proactively by being role models and also by stepping in and settings things straight whenever they feel the community is straying too far from their vision. Reddit is a good example of this. Joel's forums at joelonsoftware, which fostered a very tightly knit entrepreneur community, were also heavily influenced by the omnipresence of the site owner.

Unfortunately this is not an elegant technical hack, just simple hard work on the part of administrators.

Get rid of the whole point system. I go to HN for the community, not to collect points. It seems to provide incentives for the wrong behavior, even tho I understand that it was originally intended to do the exact opposite.

A community stands or falls on the quality of the interactions. Therefore to a certain extent, you have to let it thrive or die on its own.

Solely my opinion, but I see points as getting in the way, motivating bad behavior, and not relevant to why I come to HN.

Karma can also provide a subtle (or not so subtle) incentive to good behavior.

For instance, I've seen many instances of newbies making an "lol" or "wtf?" comment only to be quickly downvoted in to oblivion.

Hopefully it doesn't take much of that for them to get the point.

Also, the downvoting provides a quick and easy way for the community to express displeasure at someone violating the community's norms, without needing to write a long post explaining just why a comment consisting solely of "lol" or "wtf?" isn't appropriate.

On the othe hand, it's clear that karma is not a panacea, that it can be gamed, that it can encourage an echo chamber effect, and that it scales poorly when a site becomes as large as HN.

Make voting on comments cost karma. Alternately, make new story submission require at least 100 minimum karma level. I suspect the effect here would be to reduce the number of frivolous and spammy submissions. When more high quality submissions are the topics of conversation, the quality of comments will go up.
Slippery slope... I think Digg is the ultimate example of where the "elder statesmen" become the only people that matter.

Anyone who's smart should be able to contribute w/o passing some sort of exclusionary bar.

"Anyone who's smart should be able to contribute w/o passing some sort of exclusionary bar."

The difference is that to get even one or two stories a week on the front page of Digg, you basically had to devote your entire life to the site. Whereas getting 100 karma doesn't take more than a couple weeks of casual use.

> new story submission require at least x minimum karma level

This seems to have worked well for stack overflow. Just spitballing here pg, but what if you extended this to commenting as well? HN could be open to browsing, but to comment, you needed to be invited by another member with x karma? You could quantify this by requiring 10 karma to comment and 100 to post, and allowing a user to gain x karma by getting an 'invite' from an existing user with 100 karma.

Of course, without data, it's hard to draw conclusions about the fundamental issues.

A real part of the value of HN is that you often get comments from the people involved in the story itself - whether it be the author of a post or the developer of a piece of software being discussed or Matt Cutts giving the lowdown on Google's policies.

Some of these people just create an account to make a few contributions on specific threads and then they're gone. Having an invite system would increase the friction in being able to do this.

I like this idea, but it would need some tweaking before being used.

If voting on a comment gives karma to the commentator and it takes karma to vote, then people would need to comment to vote. I think this type of zero sum game would increase the number of poor comments from people who want to vote and an increase in poor comments is what we are trying to prevent.

I live in Argentina and I usually speak in Spanish. So when I want to make a comment I have to switch to English mode, write the comment, run a spell checker and then submit it.

I think that an initial 100 karma level for submissions will make the site more difficult to use for the not native English speakers.

A hard limit on the maximum upvotes a comment can get. Say, 25.
Good idea. Also maybe a hard limit for how much a user can earn from submitting articles.
Also a limit on how much Karma you can get from one article submission. A lot of times it feels like luck when someone submits something and gets hundreds of Karma for it because they posted it first.
I'd try to severely decrease total # of comments.

Really bad comments are not the root of the problem. Simply having large number of mediocre comments crowds out and discourages thoughtful discussion from starting at all.

I'd say:

* create some real cost to making comments

* make bad comments disappear/not display at all with time

* make things less democratic -- to encourage good behavior identify users who have this behavior and make this behavior more prominent programmaticly

Making it cost karma to comment would be one way to do that. I could also do something like slashdot and reddit do, and not show comments below some threshold.
You think it's a good idea to penalize people for commenting? How does this help people new to the scene? "Thanks for joining in the conversation, here's a penalty" This only rewards people who have been around in the scene for a while, while penalizing all the new people.
Penalizing comments is also a more graceful solution than invite-only, because I doubt Jason Fried or Joel Spolsky care too much about their karma score here; the only people this dissuades are people who are commenting to game.
It doesn't dissuade anyone who doesn't care about their karma, or who feels strongly enough about what they want to say.

A karma penalty is a very soft slap on the wrist, at best.

It won't stop trolls, or assholes, or anyone with an agenda.

Subtracting a fixed amount of karma also gets less effective the more karma one has accumulated. So people with a lot of karma will be able to get away with more than people with less karma.

This has an upside, in that it allows more valued members of the community to express themselves more freely. The downside is that they can act like assholes without much repercussion. If the community winds up rewarding them for acting like assholes (by, say, upvoting their assholish comment) then that's even worse.

Of course, any community that not only tolerates but encourages assholes is not a community I want to be part of, and I've left a number of communities over that sort of behavior.

But there are other solutions short of "love it or leave it". I describe one such alternative in another comment in this thread:

http://news.ycombinator.com/item?id=2405266

People who have been around longer also know better what the community expects.
It's not penalizing, it's just acknowledging seniority. Sort of like how on Slashdot you can't vote on other peoples' comments until you accumulate enough karma.
> acknowledging seniority

Why is seniority good in and of itself? Privileging seniority seems anti-democratic and anti-newbie.

Or a karmic penalty for upvoting a comment. That way you have to be respected by the community before you can upvote a thread. Maybe it would help curtail negative or snarky comments, if only because people upvoting have been around longer.

Stackexchange kind of does this. You need a certain (low) amount of karma before you can post certain types of submissions.

I think it's important to let new users have a visible voice, but giving older users greater powers for moderation might help preserve the older attitude of the site.

If comments cost karma it would lead to less discussion, which I think would lead to better discussion.

Popular comments will make more karma than they cost, so users will still be encouraged to leave comments that will become popular.

It seems that a system like this will be even more sensitive to what community considers popular. For this to work well you'll need to make sure that comment being popular correlates with it being good. To improve on that you'd may need to further reduce inefficiencies (e.g. time-of-day vs popularity) and maybe implement un-democratic measures if "voice of the community" still doesn't correlate with good.

I'd split test this system (and any other change like this). Have some posts that have these new rules in place (this should be publicly visible) and some that don't. See how this affects the results.

> If comments cost karma it would lead to less discussion, which I think would lead to better discussion.

This would probably work well if combined with the private messaging function mentioned elsewhere on the thread.

> If comments cost karma it would lead to less discussion, which I think would lead to better discussion.

I disagree. I think it would lead to a mix of bland groupthink and fashionable rebellion, with no room inbetween for the merely thoughtful.

It also actively disincentivises posting constructive comments on threads few people are likely to read, as commenting has a negative expected value. A constructive suggestion in a page dropping towards the bottom of the Ask/Show HN might get an upvote from the author if they vote, has a negligible chance of garnering upvotes from anyone else, and yet is more potentially useful to at least one member of the community than any number of eloquently-stated opinions on the 'openness' of a particular platform, whether we're in a bubble yet or the idiocy of the USPTO.
If comments cost karma, some users will just submit more stories to try to earn more karma so they can keep commenting. That could decrease the overall quality of articles. One option would be to stop awarding karma for stories so that they only way to earn karma is through good comments. That would conflict with comments costing karma though because you'd never be able to earn the initial amount.
One feature of Slashdot that I did like when I still used it was the option to sort comments by point value, which usually meant excluding everything below a +4. Great time-saver, for those who are not interested in scrolling through long threads. For those who have more time or interest in reading through the discussions, they still see everything.
This doesn't really address the main problem, which is that "poor quality" comments are being upvoted.
If it costs karma to comment or vote, then you also have to limit the number of submissions people make. Otherwise people will spam HN with controversial topics to accumulate enough karma to vote/comment. That said, if you could only submit one link per day, and it cost karma to up-vote or comment, I think the result would be much-enhanced. Also you'd have to give new members 100 points or something, but maybe they wouldn't be able to use them for the first week.
Maybe something along the lines of N free comments a day, or 1 free comment per thread, just to get the ball rolling.
I'd be careful of comment thresholds at least to some extent. It's pretty well known across many of these services that if you reply quickly to a new story you're much more likely to get upvotes. Similarly with replying to a top post. If they need to be written quickly they often aren't as high quality or carefully considered, or may even fall back on some of the easy bad posting styles you're trying to eliminate. If you take the existing karma whoring incentives to do this and add to it that you may never even get read if you don't get in early that might cause a lot more people to play that game. And cause a drop off in participation from people who might only be able to respond to a thread when it's hours old.

I think in general you already have most of a filter in that bad comments get pushed down and the lower sections of comments seem to be read much less often.

But that doesn't solve the problem of bad comments that get tons of upvotes, in fact it will probably aggravate that problem.
Risk aversion and loss aversion might lead to excessive silence if every comment cost karma. Maybe the cost could kick in above a certain speed/number of comments, or above some other threshold.
One possible implementation is to scale the number of comments a person can make in a time period dependent upon his median comment score over some previous last time period.

I've got no citation at all, but I'd bet that people who have N comments left in the next week will be more miserly about using them. You could also make exceptions for replying to replies and that kind of thing, so as not to artificially limit back and forth (which is sometimes a good thing; for example, tcptacek and zedshaw always have good conversations, if a bit argumentative).

It'd cut back on people who shotgun "funny" comments and have them land occasionally without disrupting people who try to only comment when they have something valuable to say.

A set of super voters would also be good, but, again, not super democratic.

Great ideas. The act of commenting on a thread should cost you a few Karma points. Right now there is now down side to commenting. If you comment and it cost you say 5 points, and you don't get 5 back, you'd probably think about it a little more next time.

Also, on the 'less democratic' idea, up-votes from users with more karma could be weighted heavier. Quality comments would rise up faster.

I think you're right about mediocre comments being the real problem, and about identifying good users and rewarding them programmatically. I would take this a step further though and apply the same logic to story submissions.

For example, bloggers who write insightful stuff on a regular basis should get some sort of bonus when content from their domains is upvoted early on. Maybe each upvote would internally count as 1.5 upvotes for the first 100. Similarly, bloggers who have a reputation for writing linkbait should get some sort of penalty.

Right now you can easily spend ten hours writing an amazing blog post and have it not even make the front page. This provides an amazingly strong disincentive for intelligent people to contribute, especially when the front page is dominated by vapid current events gossip. The heart of the problem is that the current system is set up to reward people for submitting garbage from TechCrunch and to punish people who try to make thoughtful contributions of their own.

Similarly, content that takes longer to read should stick around on the new page for longer. Otherwise the front page gets dominated by fluff. Again, there are lots of people who would be willing to spend 10+ hours writing a 3,500 word essay designed to benefit the HN community, but they don't because they know that there is essentially zero chance of it hitting the front page. Right now by the time the first few people finish reading, it's no longer on the new page so any upvotes basically count for nothing. This problem gets vastly worse as the amount of content submitted increases, so if nothing changes then we're probably only a couple more iterations away from having the front page be dominated by pictures of cats.

On the other hand, I've been ridiculously productive the last few months now that HN is basically unreadable, so maybe the decline in quality isn't such a bad thing after all.

I would strongly prefer we not add features that make HN more echo-chamber-y. For instance, I like John Gruber's writing a lot, but I'm not so much a fan of every one of his posts being on the front page.
I agree, but I don't think that invalidates what I'm suggesting. I think the solution is just to not give Gruber a bonus, or else to even give him a negative bonus.

The idea is to reward people who are writing good stuff but who aren't making the front page, not to reward people who already have all their stuff upvoted.

How about giving reliably good sources 10 "soft upvotes" to start with so that they'll be visible, but not let their score increase past 10 until 10 users have actually upvoted it. That way, a story won't get much staying power on the front page until users have validated the initial assumption that the article is good.
Another option is to have some social cost to making comments. The easiest way to do this is to force real names - look at e.g. Techcrunch (but there are other examples) to see the impact on comment quality.

I can't see it working here, for a number of reasons, but it's an interesting thought experiment.

How would you know that someone's used their real name?

That might work when there are a small number of participants and everyone has met everyone else face to face or when there's at least someone to vouch for the identity of every participant, but it doesn't scale very well to a community the size of HN.

I agree with gleb and Alex3917 that a large number of mediocre comments are the true problem.

I used to participate in the comments because the conversations were stimulating and the community was small. The community's too large for everyone to talk now, but HN has been the best tool for the intellectually curious to date and that doesn't have to change. The bar for commenting just needs to be higher.

When I read the top comments nowadays I'm expecting them to be written by:

- the author of the submitted article - the subject(s) of the article - employees or close relations to the subject(s) of the article - experts in the subject matter

At the bottom I expect to find comments such as product feedback or links to the print version of the article and minor but useful stuff like that. Smart people with interesting things to say shouldn't leave comments here--the community's too big for that now.

I'm not sure how you programmatically enforce that. It might be as simple as changing the commenting policies and have the users adjust their self policing.

Something slashdotty -- i.e. qualitative moderation, not just quantitative moderation -- would help. If you had seperate upvote buttons for "amusing" and "informative", this could factor into sorting.
Rather than making the site invite only, how about some means of differentiating read and write access, i.e. the amount of times you can upvote or submit is tied to your karma.

That way the information is still accessible to everyone, and if someone new has something to interesting to contribute, that info will still surface if it's picked up by vetted users.

Create positive and negative moderators but make the roles mutually exclusive.

The positive mods can promote stories and comments beyond normal up-voting and the negative mods do something similar with down-voting/flagging.

People can become 'supermods' based on karma, election, or something more arbitrary.

(comment deleted)
A community can grow only so large before it has to provide some personalization so it is not trying to be everything to everyone.

Reddit has subreddits and you can choose the ones from which stories appear on the front page. HN can start with allowing users to 'frontpage' other users aka whitelisting by showing stories from only these users on the front page. The next logical step is allowing blacklisting. Version 2.0 of this would allow whitelisting and blacklisting of content-sources (sites), in addition to users, so that I could blacklist certain blogs if I wanted to.

This will result in some fragmentation of the community, but in my opinion, it will keep HN interesting for everyone. This may also reduce the need to answer subjective editorial questions such as - we don't allow politics, but is open-source-politics politics? Is coverage of world-changing-elections allowed?