News for hackers. Stuff that matters.
Re:Ask HN: More types of flagging? by gojomo on Sunday December 20, 2009 @08:44PM (#1006958)
My theory is that a separate agree/disagree fights groupthink because it moves popularity/agreement out of the reward/penalty dimension.

People want to register their opinion -- and sometimes that opinion is just yes/no. It's good for a site to offer a low-effort, low-visual-pollution way to capture that -- single click votes/favorites/likes work well for that. (It's better than lots of 'me too' or 'I disagree' or 'my thoughts exactly' micro-comments.)

But, if those signals are mixed with a sense of righteousness/transgression -- which is inevitable with leaderboards and display rules whereby 'high-rated' comments move up, and 'low-rated' comments fade from view -- then people may withhold unpopular but important viewpoints, or be tempted to race to be the first to post a banal but crowd-pleasing viewpoint.

My theory could be wrong. Some people might care so much about agreement that they obsess over that score, and knowing exactly how many people agree/disagree would then cause even more synchronization-of-publicly-expressed-views. But I think this crowd is sophisticated enough to draw (and make use of) the distinction between a bad comment and a controversial minority viewpoint.

And I would like to be in a place where someone who advances a controversial minority viewpoint, but does so in an articulate, thought-provoking, civil manner, could be on the leaderboard even if every individual post of theirs has more net disagreement than agreement.

(Which brings up a related point: it would be interesting to report an agree/disagree axis as two totals, not just the net difference. '101 agree, 100 disagree' is more meaningful than a net score of '+1'. A sparkline bar graph or tick-series could work really well for this, though it might not need to appear on every comment, or appear until requested.)

Re:Ask HN: More types of flagging? by Perceval on Sunday December 20, 2009 @09:27PM (#1007016)
Ranking +101/-100 comments higher than +1/-0 comments is something that they implemented at reddit a couple months ago. They use an algorithm called a "Wilson Score confidence interval for a Bernoulli parameter."

Here are some relevant links:

http://blog.reddit.com/2009/10/reddits-new-comment-sorting-s...

http://www.evanmiller.org/how-not-to-sort-by-average-rating....

http://blog.linkibol.com/post/How-to-Build-a-Popularity-Algo...

Digg also has a way to rank comments by "controversy", so the comments that are most contested by up and down votes are placed at the top: http://about.digg.com/blog/new-comments-system-released