Ask YC: How to prevent social news sites from devolving?

Not necessarily. I thought about this from the beginning with News.YC, because I didn't want that to happen here. The solution I came up with is to pick a group of smart people as editors and weight everyone else's votes based on how much they overlap with the editors'. Then as long as the editors don't vote for dumb stories, they won't take over the front page.

I don't have this turned on, but I do have something in the code that shows me where stories would rank if I did. The good news is that, as of right now, the front page wouldn't look significantly different if vote weighting were turned on.

But this site is also much more targeted than a digg or reddit or mixx type of site.

I think if you stick to any given topic (hacking, politics, biotech, environmental, etc.) you have a lesser chance of the site becoming a giant cesspool of repeated lame topics.

So then: someone should make "voteupronpaulstories.com" and "allyourronpaularebelongtous.com" and "this-is-where-you-vote-on-videos-and-NSFW-content.com" to make reddit and digg better.

Interesting. I reached the same conclusion. I start by inviting trusted friends and friends of friends to sign up to the community before it's open to the public, and use their voting data as a base to measure against. And, of course, continued tweaking is also required. It's not 100% democratic because a selected few set the tone of the community, but that's not necessarily a bad thing. Seems to work great in news.yc.

As I've thought more about it, it sticks in my head: Have you tried weighting votes by the voter's karma? Use some floor for overall karma, or a logistic function, and you let the community be the editors. It could drift over time, but it would take a while to move significantly if at all. I'm willing to bet that the community principles would continue to drive submissions and votes. That inertia creates a stable orbit.

The editors seem like great training wheels and a good sanity check over time. But I'm really curious if they become irrelevant as the community takes on life. Seems like you could test this hypothesis by using karma-votes to predict edited-votes. Drift could then be measured as that prediction varies from the edited standard.

A karma weighted neural network. A variant of pagerank. The only thing to worry about is if the karma based penalty function is a good gauge of quality or just activity.

It needn't be as complicated as a neural net. Just calculate the difference between the editor-based points to each story and the karma-weighted points. I bet that's a closer match than the current no-weight version. If you played with the function, you could even try to approximate the editors from the users.

And the beauty here of karma is that it inherently isn't just activity. Fundamentally it is a gauge of quality because people make informed decisions on what to value when they vote. For instance, you could submit every story, every day, from digg's front page and this community won't reward you (much). Better to make more informed decisions about what to submit, if you value your time. The goal isn't submission - it's submission to a community standard.

No, the tabloid posters' karma would go up (presumably) and so their postings would get stronger. (I reason that if tabloid stuff got voted down, people would post less of it, so it must be finding an audience.)

My personal tactic is to be nasty to posters of crap. It has more impact than some mere 'point', and gives them guidance as to what's wrong.

Only to the extent that the community appreciates (and up-votes) what you think of as tabloid postings. Karma is a function of every submission (and comment). If the submissions aren't appreciated, neither will be the user. If a community mind has already set in, it should, in theory, be hard to shake free of that standard.

Being nasty to posters of articles that don't meet your definition of Hacker News sounds like a terrible idea to me. Spam/personally linkjacked articles are different.

I don't think that's such a great idea.

There is not, IMO, a great technical "solution". There are technical things that need to be in place to put a floor in - to keep things from going to pot, but the most important thing is to put a lot of emphasis on creating a good community and communicating its values to new members. Some aspects of that that I think are valuable:

- No politics. I think it's even worthwhile to sacrifice things that are (to me) potentially interesting, like economics, where they get too political. As long as we keep things on the subset of subjects that we know about and are interested in (tech, startups, defined broadly), we're more likely to find common ground, and be able to comment intelligently.

- Don't vote people into oblivion unless they're being seriously lame, on purpose. A 0 or -1 is enough to indicate disapproval.

- Don't say things you wouldn't say to someone's face.

- If you see something by the user 'pg', be sure to vote it up;-)

http://ycombinator.com/newsguidelines.html

Has some others.

I wasn't thinking about this regarding YC news's quality. It's remained amazingly stable since I've been using it. I was thinking about this regarding Reddit, and how the signal to noise ration seems to be degrading even more recently. The main reddit is fairly unusable for me now, although I like the science and programming sub-reddits.

I was asking for ideas for future reference regarding sites with social components. It seems to be a common problem online: how to avoid the troll issue and signal to noise degradation issue.

I was thinking that votes could be weighted based on length of stay, or karma or both. Users who've been around longer and who had higher karma are obviously more invested in the site than hoards of users that just started.

@PG: I really hadn't thought of weighting everyone's votes similar to the editors. A nice, elegant solution I think.

It seems we have nickb to thank for reposting just the good stuff from reddit here.

Rob Malda and the Slashdot crowd deserve kudos for blazing a trail and for sticking to their guns. they seem to have weathered Digg and the like just fine and will most likely outlast them all. I still stop by Slashdot regularly, but haven't visited reddit for almost a year now.

Slashdot seems to have found at least a local minima in balancing editorial selection and audience participation. Firehose, a nod to Digg, is an interesting spin on democratized story selection.

The difference as I see it is that the important things that appeal to the Slashdot crowd appear there about a day after it shows up on Reddit. On the social aggregators, you get stories a day early, but you have to filter through so much more stuff to get at interesting articles. Of the interesting stuff on social news sites like reddit, you tend to find more interesting things that never would make it onto slashdot.

> Are social news sites destined to devolve? ... How does one keep social news sites from sliding down that slope?

With moderators/editors who are responsible for the quality of content?

It seems that a social news site only stays relevant as long as the people in charge are interested in maintaining it. As it is, reddit's problem is that they put "everyone" in charge, and, as it gained popularity, "everyone" was therefore free to change their mind about what content is acceptable.

To use a successful example besides Slashdot/mefi, look at the n-sider forums (discussion community focused around Nintendo products). Although membership is free and content is provided by the community, the quality of the site has remained generally high for the 6-7 years I've been visiting it. Why? Simply because the moderators are quick to warn (and ban) members that don't contribute interesting discussion.

Basically, idealized sites such as reddit misunderstand where democracy exists regarding the internet--since the ability to found competing communities is cheap and unrestricted, it shouldn't be reddit's responsibility to give everyone a voice.

The problem isn't people who are actively trying to break the site, the problem is people who think they are making valuable contributions but aren't. It's really easy to ban someone who is being malicious. It's much harder to ban someone who is well-intentioned but useless.

1> Early adopters come in

2> Because they are early adopters, they tend to know about lots of cool things

3> The site becomes popular

4> Late adopters join

5> Site gets diluted, irritating

6> Early adopters move on

7> Site is a large, popular wasteland

I'd point to Metafilter as a site with unusual longevity due to a closed subscription process. They recently covered the Hexayurt Project, however, and I was not impressed - a bunch of braying jackals who hadn't RTFA screaming their ignorance in public. So even Mefi is on the decline, apparently, at least in the comments. The front page is still pretty decent.

But that, really, is the key: you need to stop adding posters when the site is cool and hip, to draw out the period in which the site is providing early-adopter content.

Moderation systems won't do it: they get swamped by bad moderators over time, witness slashdot's meta-moderation corrective.

Good luck.

I think the problem at a digg or reddit is that everyone's vote is considered equal. Yet, why should someone who's just arrived have as much say as someone who's been there for over a year? In real-life communities, voices take a while to accumulate. Why should online communities be any different?

I'd argue we expect that implicit hierarchy to provide a trustworthiness. If someone's submissions aren't being up-voted, they have the choice to either conform or find a community better suited to their interests. To the extent they match the community principles is the extent to which their own voice becomes heard and amplified.

But you might frustrate newbies ("I can't get links onto the front page! Too snobby for me, I'm going elsewhere!"). And users are ultimately ad-bait at places like Reddit. If Reddit could get Digg's traffic by losing whatever quality they have, they would in a heartbeat (or Conde-Nast would make them), whatever regret the original developers might feel about it.

All the newbie has to do is submit content that resonates with the community. Think of it as a game that they can play by choice. Sooner or later they'll figure out what fits. And I'd venture that people do this anyways. It's just that the power of the original community voice got watered down by treating every vote equally.

You hit the nail on the head. The next "big thing" for social news will be an implementation of an "editors" concept. Social sites don't have the lock-in they like to think. Users moved from friendster to my space and now from my space to facebook.

Social news is where search was before Google came along. A social news site that solves the editing problem (designed to naturally implement "star editors", perhaps) can still win big here by virtue of having better content, being less game-able, and not degrading over time. Someone do this, and send me an invite, please!

Social sites which use a democratic process are just that - democratic. You may not agree with the masses but these are the stories they like. I don't see this as a problem. The problem I really see is that 'we' only consider what we think appropriate/intellectual content as acceptable content. These sites are designed to promote what the masses vote and they do a darn effective job of it.

"Whatever copy sells magazines at the newsstand is what the editors are going to put in a tabloid"...isn't this what every startup hopes for - serving the users need?

PS: I do see spamming as a problem but that can be tackled separately.

Both digg and reddit have karma systems so that each vote is not equal. I think that as a social news site grows (or any community), you have to subdivide the community somehow - essentially create a collection of niche sites.

I personally believe that this is where the data you collect about a persons voting should be used to recommend articles for them specifically. Then you can go to the main page to see what the community votes up, or you can go to a personalized page that is designed for you based on your voting and what other similar users vote for.

Its a mix between democracy and personalization and it allows for the users who really do want to be the lowest common denominator.

In ancient greece, voting was mandatory. If everyone on reddit were FORCED to vote, I wager that the quality of stories would be vastly improved. I believe that reddit is currently being run by a small minority of the users.

But how would you force people to vote?

Instead of a "comment" link, have an "upmod and comment" link and a "downmod and comment" link, so that in order to comment on a story, you have to express an opinion on it.

Sturgeon's Law applies to people. This must be accepted and dealt with. Choose between dictatorship (omnipotent, zealous editors) or ease of compartmentalization (a la Usenet.)

This has been asked before, a lot of the comments seem to revolve around the idea that it is unpreventable.

As sites grow in popularity, they attract more "typical" users. These new users represent the common interests moreso than what a bunch of early adopter tech-geeks want to see :)

It can sometimes be shocking to remember that the large majority of people on the Internet are not logical and rational.

> This has been asked before, a lot of the comments seem to revolve around the idea that it is unpreventable.

It's a hard problem, but not impossible. The structure and function of the site itself plays a large part in governing group behavior.

Reddit worked for a long time because it had smart people, and a voting system that responded to whatever its userbase wanted. Now it's in rapid decline because it has a growing majority of not-so-smart people, and the voting system still responds to whatever the userbase wants. Reddit also suffered from being known as the site where smart people hang out, which brought lots of Digg refugees over to it (thereby destroying the quality they were looking for).

I'd be very interested in seeing a list of reddit headlines with votes weighted by seniority on the site. It might resemble classic reddit.

One dimension of the problem is that misleading, attention-whoring headlines work. (Maybe the sites even like these, because they generate traffic in discussion threads.)

Headlines could be rated separately from the link; perhaps a smaller group of trusted editors, or a random panel of unassociated readers, could approve/edit headlines. Further, votes without having visited the target article could be deweighted.

Another dimension is vote-gaming, including outright fraud. Increasing the cost of votes, with captchas or even registration/poll fees, might help.

Actually making vote-mobs explicit, as in the 'Uberfact' idea [+], I also think has potential. That is, rather than making recommendations/correlations for every user account, let users create their own named cliques. (Let the founders of such cliques have ultimate say on who else can join as voters.) Then, every article gets a different score per clique, and you can view the 'top' items from any clique's (or combination of cliques') perspective, whether you share their views or not.

Some cliques could be ideological, others by membership in some external community, others purely automated (perhaps mirroring some external scoring, like PageRank or Digg-score). Then you channel a lot of the gaming into a more productive and filterable expression of passion; the default view of the site might degrade, but even slightly savvy users can extract a elite, useful view.

[+] http://unqualified-reservations.blogspot.com/2007/08/uberfac...

I don't think you would want to as the site gets bigger. I think the solution is to give the end user a way to filter this stuff out.

take digg for example if you filter out all the stuff you are not interested in or just look at what your friends are digging it works.

digg, reddit and any other socially influenced news site will always "devolve" into the mainstream and people living closer to the edge will always seek out more focused and intellectually stimulating sources for news and discussion.

the networks that understand social influence and are brave enough to see where it can take them are the ones who will survive. Not the ones looking to control and guide the process with 'editors'.

When I first saw the headline, I thought it read "How to prevent social news sites from developing."

'Cuz that might work, too.

Shouldn't it be the highest common denominator? In the metaphor, I prefer a number compatible with me first, and a higher one second. The highest common denominator is the highest number compatible with everyone.