I think the best way to handle this is to apply a sort of social equivalent to the PageRank algorithm to users. Thus, when a user votes up/down an article, its points change as a function of that user's karma level. Consequently, users with greater karma have greater direct input into the community.
Eigenvalue centrality is a fascinating algorithm that provides such a use case: http://en.wikipedia.org/wiki/Centrality#Eigenvector_centrali...
Another benefit of such a system is you would no longer need to enforce karma thresholds for downvoting, as this sort of meritocracy arises naturally from the algorithm.
It would be interesting too because over time as a user's karma rises or falls it would change the current score of all the articles and comments he's voted. That would be interesting for two reasons: 1. It could cause old, interesting articles to bubble back up onto the front page months or years after they were originally posted and 2. It could help alleviate people's anxiety associated with their own karma score-- it would be something that you know bounces around in an out-of-your-control manner and not something that monotonically trends upwards. The loop between user's karma, article and comment points would be closed. We could use our fancy math to analyze the stability of the system and figure out the proper controller to insert in the loop to prevent positive feedback making it all go to the rails. It could be beautiful.
With popularity comes the masses and that will always lead to lowest common denominator winning out. Hopefully there will never be any lolz cats on here!
Great job btw pg.
You could establish influence as a factor of a user's "distance" from the community's founders: you'd have the founders, the founders' friends, the founders' friends' friends, and so on, each with diminishing influence in community moderation.
However, this and other suggestions are not new, and pg doesn't tend to comment on them.
So in the end it's all a waste of effort.
edit for additional explanation: Democracy is another name for the tyranny of the mean. You can try to educate the mean all you want, but at some point it's a losing strategy due to the sheer weight of the numbers.
Further thought: stackoverflow is an example of a community that has implemented an incremental rights model successfully. Universal rights are cleaner and more elegant, but ultimately give a community over to the mean.
- The community of readers will keep growing.
- New readers will continue to find it easy to comment and participate.
- New readers will mostly enjoy and prefer the kind of content PG and many older users want to discourage (not a big reach, since there are tech-oriented sites with millions of readers and participants having pretty shallow discussions.)
Then we have a different problem; we're trying to suppress the majority of users and rule with the minority. Incremental rights with a simple "karma" model won't help us in any way I can see, because the majority will vote up people who contribute the kind of content we don't prefer, and those people will have the karma to perpetuate that content.
(As I understood, that was your original point here, but I don't see how incremental rights is a good way of tackling that issue, at least not the way SO does it.)
Imagine a 30-day waiting period between joining the site and being able to influence content. You won't stick around unless you like what is already there.
pg, you've done a hell of a job thus far. i've never seen an internet hangout get this big and this old, yet remain this polite and civil. but as external forces conspire to drag the site down toward the mean, fighting them off will require increasingly severe countermeasures.
If you don't want the character of the site to change then don't allow new users to vote. It annoyed me on StackOverflow at first, but now I see the wisdom in it.
In this way, ultimately PG determines the character of the site by spreading voting power directly and indirectly via his voting.
I think it makes more sense for voting power to be associated with karma level.
This could be gamed, or lead to runaway karma effects within a small, incestuous cabal.
If I were pg, I would handle it similarly to amichail's suggestion, but instead of voting power going to submitters of articles that I like, I would have it go to voters of articles that I like.
In other words, I would restate the problem as, "I can't be awake 24 hours to moderate every single submission, so instead, I'll assign invisible voting power based on how reliably a given voter's running and recent preferences match what I would want to see on the front page."
I would add some random jitter to the algorithm, to mildly penalize people whose votes are too similar to my own.
The bonus to consistently good submitters would not be better voting privileges, but better submitting privileges (better initial placement and/or slower decay).
And I would silently give voters and submitters whose patterns negatively correlated with my preferences corresponding penalties, and use their poor taste as valuable data on what not to put on the front page.
If you wanted to go that way why not make a user's 'voting power' be based on their own karma / time on the site. This rewards users who have been around longer and have been 'peer reviewed' through their karma.
"Feel free to enhance the discussion" or "Feel free to add value to this thread"
Any other ways to word it worth considering?
http://www.metafilter.com | Note: Help maintain a healthy, respectful discussion by focusing comments on the issues, topics, and facts at hand—not at other members of the site.
http://ask.metafilter.com | Note: Ask MetaFilter is as useful as you make it. Please limit comments to answers or help in finding an answer. Wisecracks don't help people find answers. Thanks.
http://metatalk.metafilter.com | Note: Everyone needs a hug.
In the old days comments always had to be previewed in context (with a full refresh) before they could be posted. The "New Post" pages have huge search fields for dupe avoidance, dupe search on preview, stern warnings about what is allowed, and occasionally a row of siren.gif from the Drudge Report.
I would suggest one additional thing, which is a little bit more invasive. After a user submits a comment, show a confirmation question like "Are you sure you are making a positive contribution to the conversation? [Yes] [No]"
I believe the posts are only part of the problem, though. The post obviously needs to exist, but it's the votes that make a comment insipid or insidious. Votes are multipliers: nobody cares about a stupid comment with few upvotes and a great comment with many upvotes is an even greater comment. It's when the votes are distributed the other way around that problems start.
Furthermore, a poster that gets downvoted will probably learn from the experience whereas if he gets upvoted, he's encouraged.
The problem and the solution therefore probably lie in the voting system. There's a few options I can think of right now:
* Giving high-karma individuals more voting power. Maybe tiered or maybe on an analogue scale. This has the problem that not all karma as it is now is deserved. (Take mine: I seem to get most upvotes for snide remarks and snarky questions, not for the posts where I want to make a point.) That said, karma probably still is an indicator of quality of posts, so it may work.
* Make upvoting cost a bit of karma. I don't think this will work well, since you'll end up with a closed economy where only the top X% of the users can really participate. And X will be decreasing when more players come in.
* Give a certain amount of points each day to spend on votes. (I'm not a user of that site, but I believe Slashdot uses it. Is that correct?) This will make upvotes a scarce resource, making people be more careful what they upvote. Of course, if unused points get deleted at the end of the day, people may try to spend unused points on things they wouldn't normally upvote.
* Up the current karma-limits. Right now you need X karma for feature Y, but with the influx of new users, karma has inflated somewhat. It's much easier to get to X if 10 people read your posts than it would be if only 1 person read it.
Just a few things of the top of my head. I'm not the smartest guy here, so it may not be a whole lot, though...
"I'm not the smartest guy here, so it may not be a whole lot, though..."
Please don't ever think that. 'Smartness' is just a function of few things which are in your control, at least in the long run. Thinking "I am not smart" is self-fulfilling.
I'd also place a bet that not too many people with lots of karma are flippant spammers; even the ones I think can be jerks sometimes (myself totally included) at least are putting in an effort.
I would also add one item to the list. Make all comments cost Karma. If you are saying something that you do not think has the potential to be upvoted, just don't say it.
At least in this case if people want to be strident or acerbic, they have to work for it.
Another possibility would be to look at the words people used. I'm pretty sure you could train a spam filter to recognize lame comments.
I think you are onto something here.
Also, not related to the topic so much, I find it funny that I often see people say "I do not care about points." but are also the most likely people to make suggestions about new ways to implement the points system.
It is supposed to reference the idea that there isn't any intrinsic merit in verbosity, and that we often prize writing that is pared down and says only what it needs to - prose or code, and as a group prize languages that allow conciseness with joy going to Ruby, reverence to R and J, and fear and loathing to XML.
Particularly PG who has written more than once about Lisp and how it allows dense code to express a lot, so much so indeed that he reduced function creation to 'fn' in the language that backs this very site on which you are suggesting longer comments are the answer.
(See? Not so insightful, sharp and cool now, is it?)
Instead, you should debate the merits of the idea. I was actually not so much advocating the idea as I was bringing it up for discussion - you can tell by the tone of my comment. And, while I questioned myself for even bringing up the idea of a length requirement, your response makes me lean a little bit more towards the idea.
When voting, I vote up for "interesting point", for "you put a lot of effort into this reply" and for "I agree" and "ha!". I feel conflicted downvoting a comment I like even if it's not contributory, and I don't want to upvote a comment I disagree with even if it's a good contribution. Split the two up so the conflict goes away.
You are playing into the cult of personality
It's what the site's design encourages - usernames are prominent, vote scores are prominent and tallied as the main measure of a person, popular comments get votes. Big thoughtful comments and deep discussion get squashed into nesting and pushed aside by the ever increasing flood of new submissions. "Furthering the discussion" as you put it ... what does this mean to you? Who is supposed to benefit and how, and does the site design encourage that?
Since the site owner pg is in on this thread, the community may benefit by Paul getting some feedback and ideas on how to improve the site.
The site's design does encourage conversation of course. Comments are threaded to encourage back-and-forth, and voting moderation promotes good comments and helps to keep the discussion civil.
I too like to get some karma, but I'll sometimes write something that I suspect may be down-voted (or even that I may not believe) just because I like to discuss things. Karma doesn't really matter on this site or affect your experience (other than being able to down-vote comments), and I think that fact encourages people not to treat it very seriously.
I'd like to see a more sophisticated quality detection system - a learning text classifier of some sort. PG might have enough experience in that field of programming to come up with something useful along those lines.
Okay that took nine.
One or two comments to the effect of, "I agree with this, but the way it was phrased is toxic to the site, so I downvoted it" can't hurt either.
I don't think yours is a problem that requires a technical solution. You've already got a critical mass of people who give you extraordinary deference. Just use them to to fix it.
I was thinking it would contain the offending comment, a little message about it being flagged for not meeting community standards, then briefly list the community standards. A little word of encouragement and a button to show they've actually got the message- and off they go. I don't think it would need to be confrontational or make them actually agree to anything (since they might have good reason to disagree). But a gentle reminder about acceptable behavior could help new people get acclimated.
With that said, I'm sure it has been mentioned before, but maybe restricting new users from commenting at all until they have held the account for a significant period of time would help.
But forget about the "black swan" event of Rob Pike posting here on a Go story. Think about the Zappos redesign story. It's not black-swan-crazy to think the UX guy from Zappos might comment here. And lo, he did! There are lots of "normal" people who are close to the stories we post that have a moment to comment on them.
Don't be! Another way of looking at it...
Everybody is really good at some stuff and not so good at others. Use this community and its experts as your sand box for the stuff you need to work on. Better to practice here than out there.
and afraid I'll get torn up for saying something dumb.
So what? The real experts here would never "tear you up". They'd just teach. Anyone who'd tear you up probably has issues of their own. You mustn't let that bother you. And even if it did, just log off and come back fresh later. We start over every day.
Glad you stopped lurking, heed. Welcome.
That's a good heuristic for what a lot of us are looking for here: not necessarily agreement, but a chance to learn. Some comments promote learning more than others.
I'm not so sure, if you say something uninformed about computer security I completely expect tptacek to tear you up, for instance.
On the other hand you'll learn some real stuff in the process.
Kind of like a karma multiplier so you get an instant separation of quality and then you can look at penalizing those that take the down delta rather quickly.
Choosing to assume someone is guilty before they commit an offense would surly drive bright people away.
Anyway good comment and don't get to intimidated, many of us are just wind bags anyway, bright sure but wind bags none the less.
eg if a user with 1 karma upmods a comment, it adds 1 to its score, but if a user with 100 karma upmods a comment, it adds maybe 10 to its score.
Effectively tipping the balance in favor of those who have karma, and away from newbies.
(That's if we think those with lots of karma can be trusted to have a good feel for what is useful discussion and what isn't of course).
I think it might work well though - You could make a trollish comment, and a few newbies would upmod, but the big karma nicks would downmod, vastly reducing your karma, and thus the influence you have over upmodding/downmodding others comments.
Effectively your "ability to judge good comments" would then be tied to your "ability to make good comments".
Just an idea...
Also the whole downvoting thing doesn't even work half the time. I usually see the down vote button in only like 10% of the threads, the rest I can only upvote.
w = ln(k * m)
w = weight of vote
k = karma of voter
m = median points per comment by voter
Instead of feeding into the post's score, it could also feed into the 'buoyancy' of a post (how long a fresh post gets to stay at the top of a page before beginning to sink to the bottom).Good luck with that.
I've seen content that was much more on-topic killed off ruthlessly, it creates a sense of class-justice.
Whoever posted it and how long they have been a member should not be the deciding factor in killing stuff or letting it be.
Anyway, of course it is your call, but that's how I feel about it.
I've been wondering for a while now if this flagging is useful or not, it doesn't seem to have too much effect, if any.
I wasn't referring to the 'auto kills', but to the editor kills.
I properly also triggered the spam filter, because I just went to the thread and downvoted all the comments.
Now why on earth would you do that ?
If and when it isn't doing that, said group of people should prune away the topics they don't want - otherwise it will grow into a general discussion site. Like pruning a shrub or guiding a vine. Yes, it's a bit dictatorial and a bit undemocratic, and that's fine because it's not a popularity contest trying to target the majority of people.
Often when googling for a problem and finding a result, there's a languishing forum post that goes
"problem" "irrelevant suggestion" "irrelevant suggestion" "unhelpful but well meaning comment" "me too" "answer"
and I wish someone had enough care to weed out the drivel.
I think code and culture can shape this curve in useful and beneficial ways, but fundamentally, once an aggregation site gets popular, its quality declines. And the site gets popular because it has high quality.
So we are thinking about trying to shape the curve with code and culture; that's good. I bet there's a more 'meta' solution though.
I'm not really proposing this, just throwing the idea out there. For a given site, once popularity starts pushing down voting and submission quality, make a new site. Obviously something as trivial as http://news2.ycombinator.com/ won't work. I'm not sure how this could be accomplished, but it's interesting to think about.
Regarding the exclusivity of the core group of users, I understand that would tend to be the outcome. But I think the 'forking' method would be a little more dynamic than just locking down the user list. It would allow new, high quality users to come on board, if they really wanted to.
I realize how 'classist' this sounds, and it makes me a bit uncomfortable. But as we all know, good things came come out of dwelling on uncomfortable topics.
Anything involving kharma or other metrics is also likely to fail, and will probably lead to lots of benign but likely popular posts so that people can get more "power".
Basically I think your options are to freeze the community (or make it invite only) or to be ok with the fact that trolls will start showing up and all we can do is ignore them. I don't particularly advocate either of these, but generally believe less extreme measures are superior.
There's a fine line between trolling and radical new ideas and sometimes radical new ideas are a good thing.
No. If signal:noise gets too low, good people will leave.
usenet was designed in a more naive time, and it has zero defenses against disruptions. and that's why it died.
The user's upvote count scales up per some formula, and beyond a certain point, it becomes uncapped.
I wouldn't grandfather anyone, either. KISS, and so that it also addresses the problem with the current population.
Potentional problems with this:
+ In line with the general policy of transparency that is practiced here, you may want to show users how many upvotes they have. But you may not want the load of additional profile page hits that might occur as people check their counts. Perhaps it would be a lower load to show that count in a fashion similar to how karma is shown in the banner next to the user name. However, I don't have a suggestion for a design that wouldn't make the banner look too busy.
+ A scarcity of upvotes might promote more low-value commenting. Perhaps a user's minimum time between comments needs to be defined and similarly throttled. (Simply cutting off comments when the count hits zero will merely result in people turning to comments before their count hits zero.
To take the glass half full view, perhaps such throttling would not eliminate participation, but would help people observe and become accustomed to the community's values before handing them the keys to the kingdom. And if they don't respect those values, it's going to hinder them from progressing to greater influence.
Including karma in the formula allows for high value contributors to progress more quickly. Tenure keeps a one hit wonder or clique of friends from gaining undue influence.
I only realized a couple of days ago that one can flag comments by following the comment's permalink -- why not make this more obvious and promote its wider use as an experiment?
This is not the same as just not counting the votes. Point here is by preventing voting you're explicitly communicating your expectations to the new user.
How about using your ideas from your antispam series* to limit maximum karma points awarded to a comment? Would it trigger too many false positives/negatives?
I remember someone applying a learning network or something like that to filter HN articles a while back and it didn't turn out that bad.
That way you can acknowledge the wit in zingy one liners and agree with political and ideological statements while still shoving them down to the bottom as being poor contributions. You still get the feeling of being able to chime in while not facing the "I agree so I don't want to downvote" dilemma.
Would it be difficult to integrate comment filtering for /classic and /noobs while they act as somewhat of a controller and items behave like views?
Please keep us updated as to your thinking & actions. It has been consistently interesting to watch you manage HN's evolution and growth.
(For the noobs, I don't think this has happened before)
The biggest danger is not the obviously bad comments, though. It's the meretricious ones-- the zippy one-line putdowns and strident political statements-- because the newer users actually vote these up.
The big surprise for me is how much the mere voting power of the new users is changing the character of the site. A zippy one-liner that a year ago would have languished midway down the thread now becomes the top comment.
So while I agree with you that the right thing to do about overtly troll posts is to silently flag them, I'm still not sure what to do about the subtler and more dangerous decline I've been seeing.
I suppose I should be encouraged we made it to 40k in decent shape. Maybe I'll be able to come up with some kind of fix.