> The bottom line is that no active moderation system, no matter how many users are empowered to rate each other and each other's comments, can preserve community in the face of the multiple identity syndrome inherent to online forums.
This isn't true. Depending on what you want to design against you can use both social signals (downvotes) and a web-of-trust for singular identity and another one for dealing with spam / trolling and you can keep anonymity, you can keep good content, you might get some bubbling at the extremes, but you know what? I don't want to listen to Alex Jones idiots anyway, and with the proper graph analysis techniques you can filter out the Alex Joneses while keeping in the Bill Kristols.
Also, a poor mans web-of-trust is the invite only ban both model. If you invite me in, and I get banned you get banned too and anyone else you invited has their remaining invites revoked.
We have the science to tackle shitty online behaviour. The problem is that it destroys metrics for social networks. There are not 335 million people using Twitter a month, even if there are 335 million "active accounts".
The author makes a distinction between an online"community" and "society". He posits that trying to maintain a community with individual votes is foolish since it makes the votes an "agree/disagree" vs "this is a high quality/low quality".
We don't have the technology to do this. It relies on the users to make an effort to accurately rate quality instead of using them a blunt force tools, eg : rating something you don't agree with as "off topic"
"point me to" means "please give me a public access point for the service or code repository for such a service so I can inspect it and see how it works."
It's hard to service that request because social networks are closed source and most research into webs of trust is academic. Take email; there's an ad-hoc web-of-trust out there, but it's federated amongst a number of actors. I agree we need something open, but the incentives to make one don't exist.
What computer algorithm can be used to code up "...front page position should be determined by quality of conversation not voting"? Are we talking about sentiment analysis?
Arguably, the HN flamewar detector is a crude and limited algorithmic approach to exactly that problem.
It's obviously not a complete solution, as it seeks neither to address all quality dimensions nor to replace voting, instead being a factor alongside voting, but still...
Thanks for the excellent link. Quite a good read. From the article: [In order to prevent flamewars on Hacker News, articles with "too many" comments will get heavily penalized as "controversial"....] Ok, that isn't exactly even a slightly sophisticated method that could be extended to decide some comments are "high quality". In theory if you wanted to implement article ranking you need a method that ranks articles based on comment quality. Without a automated method you are really just using humans to do the ranking.
perhaps it could be devolved to the userbase - every user gets one vote per day that can be used to flag a conversation as interesting. If they flag a second discussion, the flag migrates to it (as if to say "actually, this is more interesting"). Display the top N articles in order of number of votes.
I'm not suggesting an upvote-downvote system though, but a per-user "flag" that can highlight 1 article per day. No downvote option, and the UI would highlight its intended use.
The upvote is an implicit down vote for other things.
I think you need two things. First
an upvote/downvote so that users can express their agree/disagree. This value wouldn't necessarily effect the visibility of the content. Second, a sort of budget system where users can trade earned karma to classify an items "quality".
I think this would lead to a system where you would spend your karma on the communities you care/know/like the most. If I were, for example, a blind apple hater, the upvote/dowmvote makes it easy to downvote apple content or upvotecother content to push it down the page. But I may be more inclined to save my karma for promting Android content instead.
"sentiment analysis" is flawed because it lacks broader context of the conversation. A negative/angry/abusive comment can still be a quality comment depending on the comments upstream.
Leaving off the problem of negative/angry/.... comments, is the current state of sentiment analysis or AI sophisticated enough to determine that a set of comments are higher quality?
Think about contentious issues like abortion, politics, vim vs emacs and the comments that could arise
"Those other people should die in a fire"
"But you should shoot them too, just in case"
These might be funny, light-hearted, sarcastic, and quality in some topics and angry, mean, toxic in others.
I'm sure it is being worked on but I don't see it happening anytime soon. It seems like it would require to much training and that the biases of the trained data would still produce incorrect outcomes.
19 comments
[ 2.9 ms ] story [ 55.8 ms ] thread> The bottom line is that no active moderation system, no matter how many users are empowered to rate each other and each other's comments, can preserve community in the face of the multiple identity syndrome inherent to online forums.
This isn't true. Depending on what you want to design against you can use both social signals (downvotes) and a web-of-trust for singular identity and another one for dealing with spam / trolling and you can keep anonymity, you can keep good content, you might get some bubbling at the extremes, but you know what? I don't want to listen to Alex Jones idiots anyway, and with the proper graph analysis techniques you can filter out the Alex Joneses while keeping in the Bill Kristols.
Also, a poor mans web-of-trust is the invite only ban both model. If you invite me in, and I get banned you get banned too and anyone else you invited has their remaining invites revoked.
We have the science to tackle shitty online behaviour. The problem is that it destroys metrics for social networks. There are not 335 million people using Twitter a month, even if there are 335 million "active accounts".
We don't have the technology to do this. It relies on the users to make an effort to accurately rate quality instead of using them a blunt force tools, eg : rating something you don't agree with as "off topic"
Can you point me to some extant services that leverage this science?
It's obviously not a complete solution, as it seeks neither to address all quality dimensions nor to replace voting, instead being a factor alongside voting, but still...
http://www.righto.com/2013/11/how-hacker-news-ranking-really...
Once they learn that "a" makes comments more visible and "b" makes them less, it just becomes an agree / disagree vote
I think you need two things. First an upvote/downvote so that users can express their agree/disagree. This value wouldn't necessarily effect the visibility of the content. Second, a sort of budget system where users can trade earned karma to classify an items "quality".
I think this would lead to a system where you would spend your karma on the communities you care/know/like the most. If I were, for example, a blind apple hater, the upvote/dowmvote makes it easy to downvote apple content or upvotecother content to push it down the page. But I may be more inclined to save my karma for promting Android content instead.
"Those other people should die in a fire"
"But you should shoot them too, just in case"
These might be funny, light-hearted, sarcastic, and quality in some topics and angry, mean, toxic in others.
I'm sure it is being worked on but I don't see it happening anytime soon. It seems like it would require to much training and that the biases of the trained data would still produce incorrect outcomes.