Some of us are building a real solution to the industry's problem. karma, as demonstrated by reddit and HN, is subject to the hive mind phenomena. there is a way to integrate peer voting, but it most be up-down, to start, and it goes beyond that for a scalable solution. can't say more...
I always thought Advogato's trust metric[1] was really interesting. It was part of Raph Levine's (of libart and Ghostscript fame) PhD thesis, but he later switched topics. As I recall it worked fairly well, but Advogato was a small community and probably wasn't yet susceptible to most of the problems of simple voting.
ku5oshin also had a unique peer-review style system that seemed quite well thought out.
Slashdot had meta-voting, which was pretty neat.
It amazes me that after a decade of this, there is very little innovation. At least the communities back then were trying hard to find novel solutions. Instead of trying to figure out new mechanisms that are resistant to common abuses, everyone seems to focus on finding and punishing the cheaters. You can never catch them all, and it doesn't seem to scale well.
We need an open platform that allows easy experimentation and plugging of vote algorithms and balances. I've been working on building one, but it still has a ways to go. Even basic reddit/digg clones are scant to non-existent, so we've been working on covering most of that space.
Is this article lamenting a loss of quality, or the loss of the sense of intrigue and community associated with any small or secretive club? I'm not convinced quality has suffered, only that more people using them leads to more articles and more debate, destroying the sense that one is in the know on all the (smaller number of) news items from the day, and that their opinion is important (because there are less people to debate it).
While reading the post, I too was thinking of the Slashdot / K5 / Advogato examples.
It's weird to me that there are millions of professional nerds for whom this trifecta is not a critical historical influence. Or even in their mind. At all.
Anyway; the cycle of blooming and decay is as old as people. I don't think there is a "solution". Just gotta accept it.
To accept something less than potential is a cop-out.
Solutions materialize, but it's never an all-at-once event. Printers evolved substantially from Gutenberg's such that we have personal high-volume color laserjets sitting near us with WiFi connections and AES encryption.
Regarding karma, this is one place where I think HN's implementation gets it right (or at least more right than others). You can see your own karma, and the votes of things that you've submitted (I'm talking about comments not stories here), but you have to go look to see anyone else's, and you can't see the karma they've gotten from their comments (except as a relative number versus your own comment at the same level within a thread).
There are surely still karma whores (I'll admit to liking seeing the little number next to my name go up) but it's not quite as in-your-face as somewhere like Reddit.
Personally, I'm not such a fan of the lack of transparency on HN, because it makes the manipulations of the moderators very opaque and shady. My previous account got slowbanned despite the karma accumulated over years.
If you're going to run a place like that, overrule the community and keep them in the dark, you might as well ditch karma altogether (or make it 100% invisible and just use it as an internal aid). Otherwise you're just encouraging behavior you clearly don't want. I'm not arguing the right or wrong about the way HN is run, I just don't see the point in having karma at all here.
Transparency is key and the structural goal of a news platform should be building trust corporately and incentivizing individuals to build TRUST...not karma.
Agreed with onemorepassword. The influence of the "persuasive centrists" the OP notes means that comments that aren't largely predictable will often be downvoted, discouraging different points of view and forming a boring hivemind. (Here come the downvotes.)
Re: gaming the system, voting rings and other tit-for-tat behaviors are very much alive and well down here in the Bay Area. Without getting into specifics, many groups here have strong if unwritten "upvote your friends" policies encouraging promotion of each other's commercial interests, irrespective of the quality of the articles. It's rather like special-interest politics, where the damage done to the community is large but dispersed (from upvoting trash), whereas the benefits are localized and perceptible.
My social network is also the best source of news I have, but most of the information exchange takes place in private, often offline -- added plus that face-to-face is the only way to get access to a lot of quasi- or truly-confidential information.
One of the things that lends itself to abuse of things such as karma is a lack of cost. It costs nothing to vote something up or down, therefore we associate little value with giving karma (as you're not losing anything), whereas the accumulation of karma typically requires some effort, so the cost of acquisition is higher.
I agree with this post but I'm more optimistic than the author that a solution will be found.
Let me use HN for examples, as that's the site I'm most familiar with.
The author correctly identifies that HN does a poor job of selecting new items to display. Technically, HN exploits too much (gives existing high ranking stories too much front page time) and doesn't explore enough (give new stories a chance). I also expect that the population who reads the "new" list behaves differently to that who read exclusively the front page.
We can fix both problems! We know of algorithms to balance exploration and exploitation -- they're known as bandit algorithms. Instead of displaying a fixed front page to everyone, choose the stories to show to balance exploration and exploitation. This way new stories get a fairer chance and the population observing both is the same.
There are some technical issues but they are solvable. I work on this stuff in my day job (Myna -- mynaweb.com -- go sign up now! ;-) and we have an algorithm for handling preference data. It would need some adjustment but it's doable.
Next up is voting rings and so on. Also solvable -- stop using votes as the only input signal. Observe what people actually do -- do they actually click the link? Come back and comment? These are strong indications of interest. How much text do they write? It's just a few words it's more likely to be a spammy "ata boy!" comment. How long do they spend on the link (maybe we can track this by using an iframe?) and do they scroll? The point is there are a whole heap more signals that people give to indicate real interest. Start using these and it becomes much harder to forge interest.
This area is known as implicit feedback, and again there is a decent body of work available.
I have noticed that when the comment count surpasses the up votes the article is usually very confrontational but isn't necessarily authoritative or insightful on the subject matter.
HN and reddit are really social curating dashboards with a commenting system. both are better than self-described curating sites like storify.
a major problem is the overwhelming flow of links from the entire internet because of the third-party content structure.
exploration platforms like stumbleupon send garbage traffic, so that's not the answer, either. you don't want traffic for traffic's sake, you want real people that read and contribute.
news is a special species of information:
1) all "open" information in the sense that it is intended for the general community.
2) it is time-sensitive
3) it is geographically relevant
the current internet is structured for historical or aged-content. that's why you see all of the indexing of third-party content instead of an integrated platform that standardizes the news format and can personalize content for anyone around the world.
you have to look at the production side of the equation, not just the product, if you want to solve the news problem. considering that all of the VCs have given up and the legacy publishing institutions are approaching the problem from the perspective of adapting their structure, no one you know of is actually working on the news problem.
it's a multi-billion market opportunity with very little competition. the potential financial gains are exciting, but at the end of the day ( and the beginning), everyone wants an information edge. I want the information edge!
i've got years invested in the problem and a public beta platform to iterate up from. it will be obvious in hindsight!
Solution for the one-vote problem could be a box at the top of the front page with "Which (if any) of these five new stories are front-page worthy?" Does potentially provide a lot of visibility to crappy posts though.
21 comments
[ 3.0 ms ] story [ 60.7 ms ] threadthe best factor is paid subscriptions.
ku5oshin also had a unique peer-review style system that seemed quite well thought out.
Slashdot had meta-voting, which was pretty neat.
It amazes me that after a decade of this, there is very little innovation. At least the communities back then were trying hard to find novel solutions. Instead of trying to figure out new mechanisms that are resistant to common abuses, everyone seems to focus on finding and punishing the cheaters. You can never catch them all, and it doesn't seem to scale well.
[1] http://advogato.org/trust-metric.html
For the curious: http://github.com/sjuxax/raggregate .
Is this article lamenting a loss of quality, or the loss of the sense of intrigue and community associated with any small or secretive club? I'm not convinced quality has suffered, only that more people using them leads to more articles and more debate, destroying the sense that one is in the know on all the (smaller number of) news items from the day, and that their opinion is important (because there are less people to debate it).
It's weird to me that there are millions of professional nerds for whom this trifecta is not a critical historical influence. Or even in their mind. At all.
Anyway; the cycle of blooming and decay is as old as people. I don't think there is a "solution". Just gotta accept it.
Solutions materialize, but it's never an all-at-once event. Printers evolved substantially from Gutenberg's such that we have personal high-volume color laserjets sitting near us with WiFi connections and AES encryption.
There are surely still karma whores (I'll admit to liking seeing the little number next to my name go up) but it's not quite as in-your-face as somewhere like Reddit.
If you're going to run a place like that, overrule the community and keep them in the dark, you might as well ditch karma altogether (or make it 100% invisible and just use it as an internal aid). Otherwise you're just encouraging behavior you clearly don't want. I'm not arguing the right or wrong about the way HN is run, I just don't see the point in having karma at all here.
The hivemind doesn't mind if you argue against it in the abstract.
My social network is also the best source of news I have, but most of the information exchange takes place in private, often offline -- added plus that face-to-face is the only way to get access to a lot of quasi- or truly-confidential information.
Let me use HN for examples, as that's the site I'm most familiar with.
The author correctly identifies that HN does a poor job of selecting new items to display. Technically, HN exploits too much (gives existing high ranking stories too much front page time) and doesn't explore enough (give new stories a chance). I also expect that the population who reads the "new" list behaves differently to that who read exclusively the front page.
We can fix both problems! We know of algorithms to balance exploration and exploitation -- they're known as bandit algorithms. Instead of displaying a fixed front page to everyone, choose the stories to show to balance exploration and exploitation. This way new stories get a fairer chance and the population observing both is the same.
There are some technical issues but they are solvable. I work on this stuff in my day job (Myna -- mynaweb.com -- go sign up now! ;-) and we have an algorithm for handling preference data. It would need some adjustment but it's doable.
Next up is voting rings and so on. Also solvable -- stop using votes as the only input signal. Observe what people actually do -- do they actually click the link? Come back and comment? These are strong indications of interest. How much text do they write? It's just a few words it's more likely to be a spammy "ata boy!" comment. How long do they spend on the link (maybe we can track this by using an iframe?) and do they scroll? The point is there are a whole heap more signals that people give to indicate real interest. Start using these and it becomes much harder to forge interest.
This area is known as implicit feedback, and again there is a decent body of work available.
a major problem is the overwhelming flow of links from the entire internet because of the third-party content structure.
exploration platforms like stumbleupon send garbage traffic, so that's not the answer, either. you don't want traffic for traffic's sake, you want real people that read and contribute.
news is a special species of information:
1) all "open" information in the sense that it is intended for the general community. 2) it is time-sensitive 3) it is geographically relevant
the current internet is structured for historical or aged-content. that's why you see all of the indexing of third-party content instead of an integrated platform that standardizes the news format and can personalize content for anyone around the world.
you have to look at the production side of the equation, not just the product, if you want to solve the news problem. considering that all of the VCs have given up and the legacy publishing institutions are approaching the problem from the perspective of adapting their structure, no one you know of is actually working on the news problem.
it's a multi-billion market opportunity with very little competition. the potential financial gains are exciting, but at the end of the day ( and the beginning), everyone wants an information edge. I want the information edge!
i've got years invested in the problem and a public beta platform to iterate up from. it will be obvious in hindsight!
:)