News for hackers. Stuff that matters.
Re:Ask HN: More types of flagging? by pg on Sunday December 20, 2009 @05:11PM (#1006677)
Yes, that is a related problem. The floor used to be -8, but as the number of users increased (and maybe the median user got meaner) there started to be lots of comparatively earnest posts that got downvoted to -8. So I raised the cutoff to -4, but that still seems unsatisfactory. Voting still feels like it's being done by a mob.

I tried not displaying points on comments a while ago, in order to solve this problem, but users complained that this made HN harder to read because they couldn't pick out the good comments. I only gave it two days' trial though. Maybe I should have given it more time. Or maybe there is some other solution.

Anyone have any ideas?

Re:Ask HN: More types of flagging? by dschobel on Monday December 21, 2009 @01:13AM (#1007299)
The feedback of the voting system is still essential, it's just that you can't really trust the voters anymore. It's gotten ridiculous to watch the votes swing from extreme to extreme on one comment based on the follow-up discussion where people openly advocate for more or less points for the OP.

The only way to fix this I can see is either some form of meta-moderation where you hold people accountable for their votes (ala slashdot), or the system I'd personally rather see, where rather than a full-democracy the users pick whose votes they see on comments. I'd have my pool of 20-30 users whom I trust and I could see their scores vs the throng's.

Just as you hand-picked some users as moderators who you feel represent what you want HN to be, let us pick who we want to serve as our content filters.

Re:Ask HN: More types of flagging? by pg on Monday December 21, 2009 @01:43AM (#1007325)
I agree with you about the voting swings. I think that, more than the extreme numbers, may be why voting is starting to seem such a mob thing lately.

It's a very interesting idea to have users see only the votes of other users they choose. One problem though is that it could easily lead to leaking who voted for what, which would make a lot of users (including me) uncomfortable. E.g. your pool is your 7 coworkers, and user x. One day you know all your coworkers are on a plane, and that you're therefore seeing user x's votes. Another problem is that it could be a lot more expensive to generate pages.

Re:Ask HN: More types of flagging? by asdflkj on Monday December 21, 2009 @11:54AM (#1008113)
You could fix the deanonymization problem this way: when you pick a trusted user, you can see not only his votes, but all his trusted users' votes, too. Maybe weigh the votes differently depending on how "distant" they are from you.
Re:Ask HN: More types of flagging? by dschobel on Monday December 21, 2009 @04:39PM (#1008653)
This is a great idea.

My original thinking was to suggest a combination of enforcing a minimum number of trusted users and then reporting a weighted average of your trusted users + the mass' rating but your solution is conceptually nicer since it adds the same noise to the vote but while extending the idea of valuing a particular user's judgement.

The big technical issue will be computing these "trusted ratings" for every comment. The trust graph could be done asynchronously so that wouldn't be an issue.

Re:Ask HN: More types of flagging? by pg on Monday December 21, 2009 @01:20PM (#1008271)
Hmm. Good idea.
Re:Ask HN: More types of flagging? by b3eck on Monday December 21, 2009 @05:22PM (#1008726)
During the experiment of not displaying points on comments, was the voting less volatile, with smaller voting swings? If it was, then hiding the points until a few minutes or hours after the comment is posted (i.e. until the volatility becomes reasonable) could replicate the effect. If not, then maybe the voting swings are inherent as a statistical property of any sufficiently large community, in which case you may want to display a confidence interval of the voting score instead of the score itself.
Re:Ask HN: More types of flagging? by emil0r on Sunday December 20, 2009 @08:42PM (#1006953)
A time-delay on down-votes perhaps? More of a psychological effect than a technical solution.

Not showing any points above 5 (or 10 or whatever), while still having an internal counter might also help.

A trust system maybe. "I trust this user. So anything he put down as interesting I want marked as interesting."

Re:Ask HN: More types of flagging? by derefr on Sunday December 20, 2009 @06:14PM (#1006756)
It's a bandwagon problem. When people see a large number (positive or negative), they want to push it larger. To snap people out of this heuristic and make them actually judge the comment, I'd suggest making people do a small amount of "work" if the number is already large in either direction (and they're trying to push it larger.) Perhaps make them enter a CAPTCHA of lg(abs(score)) characters?
Re:Ask HN: More types of flagging? by tdedecko on Sunday December 20, 2009 @09:17PM (#1006995)
You could try not acknowledging downvotes on a new item until the number of downvotes reaches a threshold. For example the first downvote on an item would have no impact on the score or the placement of the item on the page. If the threshold is 2 then, the item would be affected after 2 downvotes. At that point the normal result for downvoting would occur. This would require two people to independently interpret the quality of an item. Though, I am only considering the case of an item being only downvoted.

Another idea I have is to require a reason for the downvote. Reasons could include: disagreement, uncivil, or offtopic. The score could be affected according to the reason. This could make downvoting more thoughtful and scores could be impacted less for disagreement.

Last idea: Instead of hiding all scores. Only hide the score for items with a score less than or equal to 1. These scores have less relevance and I think people will be less critical of comments if they don't see a negative score attached.

Re:Ask HN: More types of flagging? by awwx on Sunday December 20, 2009 @05:47PM (#1006721)
Allow people to turn on seeing the points if they want to, but give more weight to votes by people who aren't choosing to see the points?
Re:Ask HN: More types of flagging? by BRadmin on Monday December 21, 2009 @02:33AM (#1007382)
Do comments follow the same list algorithm as news posts, using time & points? This seems to make sense for news which is contextualized with time, but comments... why not remove the point display on them, but have them still listed by highest (invisible) point totals.

People would really need to evaluate each comment accordingly, knowing that each vote could potentially change a parents position on the page. This could entice people to use votes to steer the conversation towards the intellectually gratifying that they want further discussed, as opposed to just things they casually 'agree' with.

Re:Ask HN: More types of flagging? by Confusion on Monday December 21, 2009 @04:06AM (#1007458)
Voting still feels like it's being done by a mob.

There will always be a majority opinion, which gets expressed through votes. As HN gets above a certain threshold of active users, the majority opinion will be indistinguishable from 'mob voting', unless users can be trained to not vote when the vote is already high in 'their' direction.

I don't think a 'honey pot' agree/disagree voting axis will work, unless it is visible to everyone. Otherwise, these users will express their opinion with the current up/down votes.

Re:Ask HN: More types of flagging? by weaksauce on Sunday December 20, 2009 @05:31PM (#1006702)
Have you thought about showing the aggregate number of flags too? This might be interesting if you have multiple types of flags.

Another possibility is to make the flag threshold higher but make the flags non anonymous(That might create a hostile environment though, e.g. tit-for-tat retaliation but I would hope this environment is mature enough for a frank discussion.)

Re:Ask HN: More types of flagging? by icey on Sunday December 20, 2009 @10:21PM (#1007082)
Reddit's solution of hiding a thread once the top post went below a certain threshold seems to work pretty well. Then you have to go through two actions to downvote.