One of the most interesting things about this site is the karma scoring, and that it reflects (to a greater degree than you see elsewhere) an objective assessment of the merits of an argument.

[Edit^6: the proposal in this post is related to the Kibitzer system, but this post discusses adding information, while that system concentrates on taking information away. Special thanks for matt's comment and to Vincentyu for being the first to point to prior discussion.  A related issue is discussed here (2009) with reference to a wikipedia, and on which Eliezer said "I may end up linking this from the About page when it comes time to explain suggested voting policies"). Data: It took me ~2 days of effort to obtain get linked to this information (09 June 2012 11:29PM -> 11 June 2012 10:28:26PM).]

Suppose a controversial post/comment has six up votes and three down votes. Right now we only see the net result: 3 points, but when the voting is mixed we're losing important information. If it's reasonably easy to implement, could we please show up and down tallies separately? E.g show "3 points (+6,-3)", at least when the voting is mixed? I think the negative votes are the single most important thing. In particular, I want to know about negative votes I receive and where I receive them, because those are the posts where I need to think carefully.

Example: here's a welcome post by syzygy, which relates to Eliezer's post about Politics as the Mind Killer. I know that it's controversial, because I can sort by controversial and it shows up high on the welcome post thread (neat feature!), but I can't tell how many down votes it has. Does syzygy commit a fallacy? (I don't mean to pick on you, sorry about that; I liked your post.)

Of course this change wouldn't fix everything. If a post has "-1 points (+0,-1)", that doesn't mean only one person read it and disapproved; maybe 100s read it and thought it was bad, but saw that it already had -1 net and considered that sufficiently punitive. This is pretty good; we don't want to spend all our time fiddling with scores.

I mean if we wanted to get fancy and use Bayesian inspired scoring, we could let everyone who wishes assign a score (say from -5 to 5) and report posterior summaries of the scores. Or, more importantly if we value objective scoring, we could identify posts that are controversial and we could have the system randomly select users with respectable karma, and assign them to give their score on the post. Such a score would be valid in a way that the current "convenience" scores are not. Additionally, posts could be scored on multiple axes: soundness of argument, potential impact, innovation, whether we agree with the normative basis of a judgement, etc....

But I'm not arguing for a complicated change, just a simple less wrong one.

Other than feasibility concerns, or maybe aesthetics, the strongest argument I can see against this proposal is that we might embarrass or shame users. Can any one give an example where that might be a concern? I figure that since we already show negative scores, users have gotten over most of that inhibition, but I'm new here.

Another possible criticism is that it's a non-issue: almost all posts are all plus or all minus, so it's not worth the effort. I disagree with this one because I think the posts where we have mixed judgements are the most important ones to get right.

EDIT: Wouldn't it be nice to know how many down votes this post has?

New Comment
66 comments, sorted by Click to highlight new comments since:

I think I speak for the silent majority when I say "wait, how can I speak for a silent majority?"

This made me laugh, but I think I may be missing something too. Is there a special reason for inertia here? [or is it just the usual reasons]

[This comment is no longer endorsed by its author]Reply

Eh, I just am fine with the way things are - I think a total score conveys the important information while maintaining a drama-muffling veil of anonymity even for the crowd itself. So yes, you were supposed to take away the impression that I think a silent majority of people are just fine with how the site is.

Thanks! (Speaking the silent majority is indeed tricky)

Here's a special reason witzvo: it's distracting us. You're bustling with newbie excitement. We've been doing this for a while and it works and we've got more important things to work on.

[This comment is no longer endorsed by its author]Reply

Thanks! That discussion also mentioned the issues list. So now I can at least get a peek at what's needed to effect a change. Would submitting a patch do it? Well, now I know more about where to look.

I also found this comment interesting:

I want to vote for and against the same post sometimes. Not too useful, but might help for ordering by controversial.

I agree with the sentiment. I think the useful low hanging fruit would be to allow you to select a section of a post with your cursor and up/down vote that section. This, and other good proposals, certainly complicate the UI, though. Another proposal there was to require a comment with a vote. This is simpler to implement but adds user-friction and clutter, so I'm not sure about the tradeoff.

Is there more I'm missing? I tried some searches but couldn't find the right phrases.

I think the useful low hanging fruit would be to allow you to select a section of a post with your cursor and up/down vote that section.

What would that actually do? (This most definitely is not low hanging fruit.)

Another proposal there was to require a comment with a vote.

Bad idea. For starters, it means there is no way to downvote trolls without starting flame wars.

[-]gjm30

it means there is no way to downvote trolls without starting flame wars.

Not if the mechanism for including a comment with your vote is separate from the usual commenting mechanism: e.g., you can enter up to 140 characters (or 80 or 72 or some other traditional number) to describe your reason for voting as you did, and then hovering over the upvote or downvote count shows a selection of these reasons. (All of them, if there aren't too many.) No provision for replying to these, other than by editing the comment that was voted on or adding an extra comment.

I don't know whether this would actually be a good thing on balance, but it doesn't seem like it need lead to flame wars.

I can imagine a vote-tagging mechanism that doesn't lead to interaction, but my intuition is that this requires that the vote-tags be standardized. The moment the tag becomes an actual comment expressing a novel thought by an individual, the usual conversational dynamics allow for interaction with the person addressed. (If the system makes this awkward by requiring it to be implemented as a comment that doesn't obviously descend from the thing it's replying to, that won't stop people from responding to it, it will just make people annoyed by the awkwardness.)

The moment the tag becomes an actual comment expressing a novel thought by an individual, the usual conversational dynamics allow for interaction with the person addressed.

interesting point, but I'm unsure whether that makes it a bad thing.

Anyway, the basis my preference for just letting you highlight some text before you click (above) is that (1) it's low effort which increases participation (2) it pinpoints the cause of the appreciation/disapproval (3) it's more flexible than a fixed set of tags.

I think the useful low hanging fruit would be to allow you to select a section of a post with your cursor and up/down vote that section.

What would that actually do? (This most definitely is not low hanging fruit.)

I think an implementation is easier than you might expect (but I'll try to do more to prove it). I haven't looked at the code on the site yet, but here's my guess: when you click up, for example, javascript is triggered to send an AJAX message to the server with the particulars (who clicked?, on what?). If the event handling code is changed so that it queries for the current selection if any, and adds that information to the message it sends the server, and if the server logs that additional information in its database, then the biggest part is already implemented.

To make this useful the information needs to come back out somehow, and making a nice shiny implementation of that involves more work, I admit. A bare bones implementation would be an additional icon on a comment (next to permalink, say). Click the icon and it gives you details about comments on the post. E.g. it dumps the time ordered (anonymous?) list of up's and down's together with anything that was selected at the time.

Edit: maybe you just meant "can you tell me how that would function as a social dynamic?" Sorry if so; let me know. see also my rant

Would submitting a patch do it? Well, now I know more about where to look.

Considering the ample number of both upvotes and noncommittal comments on your post, it seems that many would like this feature implemented but are unable/unwilling to do it themselves. Personally submitting a patch is probably the only way this is going to get implemented in the near future, and I'm sure this would be appreciated by many (including me). You might want to look at this page on the Less Wrong GitHub wiki.

Is there more I'm missing? I tried some searches but couldn't find the right phrases.

I feel like there were more off-hand comments about this, but this is likely the highest-voted post on this topic.

Thanks. To really do it, I'll probably need a test server. I'll see what I can do and report back by next weekend, hopefully, unless somebody has something like that sitting around.

So instead of people making obnoxious noises whenever they get a negative score they feel is unwarrranted, they'll complain about whenever there's even a single person that downvotes them, even if there a dozen other people upvoting them?

No.

I reject the implied general principle that when some people might behave badly in response to some information, concealing that information from everyone is justified.

I reject the implied general principle that when some people might behave badly in response to some information, concealing that information from everyone is justified.

Well phrased. I did not mean to imply that general principle. [c.f. Kibitzing scripts mentioned here somewhere to which I take it the point is personalization.] I meant that I've enough information in hand to know that I don't know what's "right" until I think more. My de facto principle would have been more along the lines of:

If you're in the woods, and everything gets quiet suddenly and you don't know why, maybe you should shut up and think.

But I'm not even endorsing that until I think more at this point.

Ok. Good. Another problematic aspect for the group dynamics to think through. Frankly, I'm beginning to think/worry inertia is right. (Waiting is)

Maybe it'd still be useful to have an experimental server where things are a bit more dynamic and we can explore new features. E.g. gmail/twitter/google+/amazon all in some ways do a better job of "promoting things to my attention" than this site does. (They lose in other more important ways, though) I have some other ambitious ideas written over here. I'm not saying these are ready for use, but maybe they're ready to prototype and refine.

I really, really don't want to see downvotes. Downvotes are much more salient than upvotes and it is already too easy to get fed up with moronic behavior as it is. Being able to tell that there are three asshats that downvoted my "+7 (10-3) karma" comment would just be irritating. I like having the 'noise' of fools or committed rivals hidden away by majority vote so I can pretend they don't exist.

Downvotes are much more salient than upvotes and it is already too easy to get fed up with moronic behavior as it is. Being able to tell that there are three asshats that downvoted my "+7 (10-3) karma" comment would just be irritating.

Why is this downvoted twice? No wonder he's concerned about it. Sheesh.

This is a completely legitimate concern that I hadn't thought of, especially not in the terms he expressed it. We're only bags of chemicals, after all {exaggeration}. It shouldn't surprise us that we have to pull an Odysseus once in a while (or be one of the sailors with wax in our ears; I'm still working on the metaphor here).

[Anyone have a better link for that story, the one about tying himself to the mast, and manipulating his men to get through the perils of sirens and whirlpools?]

I'll think about it wedrifid. Thanks.

This is a completely legitimate concern that I hadn't thought of, especially not in the terms he expressed it. We're only bags of chemicals, after all {exaggeration}. It shouldn't surprise us that we have to pull an Odysseus once in a while (or be one of the sailors with wax in our ears; I'm still working on the metaphor here).

That's a good analogy. I know that given either human psychology in general or my own psychology in particular I'm better off not having access to certain types of information. Yet if it is available - and particularly if it flashes up somewhere on the screen - it takes huge amounts of willpower to refrain from looking. So tying myself to the mast and issuing wax earplugs to the sailors sounds like a good idea!

[-]Jack20

I second this. Moreover, I don't want other people to see downvotes. It will lead to more people complaining and asking for justifications for downvotes. And it will prime people for faction-based, my-side vs. your-side thinking. You'll see more votes indicating agreement vs. disagreement instead of quality judgments. And it will probably lead to more cautious comments since people have a greater negative affect toward downvotes than upvotes.

Plus, I've yet to see an argument for showing upvotes and downvotes separately beyond "Aren't you curious?!". Information about controversy and divisiveness is already encoded in the kind of replies a post gets. A proponent of the change needs to specifically explain how it will change the judgments and behavior of commenters and posters for the better.

Finally, adding more of this kind of information adds to the addictive nature of posting here because now you're getting even more instanct and constant feedback. It's a way of optimizing Less Wrong as an addictive karmic roller-coaster ride but doesn't actually improve anyone's rationality.

Plus, I've yet to see an argument for showing upvotes and downvotes separately beyond "Aren't you curious?!"

I disagree with this part. (Edit: on reflection I think this argument is unfair, I was not clear in the "edit" I put at the end of the OP) Pertinent example was alluded to in the OP: I want to know whether it is sensible to work on a patch to effect this change. The current points on this post are 19. I happen to remember that the count used to be higher (not sure how much higher) so I think that the recent arguments have brought the total down some [fair enough]. This is all side information that I do not have direct access to except by memory. It tells me that the voting is mixed, but isn't very precise. Is it +19=(119, -100) [i.e. this is controversial and probably not worthwhile] or +19=(+35,-16) [seems more likely] or +19=(+19,0) [implausible given side information, but if it WAS true, would potentially lead me to different actions]. I do not consider this just curiosity.

It will lead to more people complaining and asking for justifications for downvotes

This calls for speculation, of course, but I grant it. I do not however see this as entirely bad, nor is it obvious to me that it is bad in net. I'm currently still disposed to think that this is a net good (at least for many users). Why?

  • Questioning votes is a way to get answers that can lead to self-improvement and not just to "faction-based" outcomes. Again this is a question of balance.
  • Votes (or something like them) need to mean something overall. This requires some policing.

Actually, don't you see how ironic it is that you're using this argument at this particular point in the thread? This post is subordinate to a comment I made that said:

Why is this downvoted twice? No wonder he's concerned about it. Sheesh.

Which was on behalf of wedrifid's point, which is largely your point. Wedrifid's post now has "+4" points. I think this represents a dynamic of social thinking that is unhealthy without policing. [I have something drafted about policing, but based on what I've learned in this post, it's not ready for mainstream consumption on this site. This is another pertinent kind of inhibition, really.]

I had some idea at the start that this issue was subtle. I thought that I had defended against criticism well "enough." I admit to being wrong about that. People have brought things to my attention that are changing my mind and I now think it is much more subtle. (which is really cool actually) I do however think it's important to get it right (well as "righter" as is feasible and not too distracting). I also think that there is scope for more radical changes to bring improvement (at least in parallel rationalist online experiments) and I don't want anyone to think that rejection of this proposal means that it isn't important to talk about how to make things better. (1) I've found it instructive (2) Wasn't the first sequence post about a "dojo"? This site is, as far as I can tell, the public dojo. It's important that it's pretty good (and it is; better than I thought; much better than I thought from the snippets of the sequences that I first tried to read which were sometimes absurd, taken out of context).

It's a way of optimizing Less Wrong as an addictive karmic roller-coaster ride but doesn't actually improve anyone's rationality.

That seems like a new and valid way to make wedrifid's point.

[-]Jack00

The current points on this post are 19. I happen to remember that the count used to be higher (not sure how much higher) so I think that the recent arguments have brought the total down some [fair enough]. This is all side information that I do not have direct access to except by memory. It tells me that the voting is mixed, but isn't very precise.

What decisions would having more precise information impact and how would it change your decision?

Questioning votes is a way to get answers that can lead to self-improvement and not just to "faction-based" outcomes. Again this is a question of balance.

Questioning net downvotes can to self-improvement. But individual votes are really noisy by themselves. People vote for all kinds of reasons, few of which need to impact what kind of comments. And unless you get a lot and they are unexpected it's unlikely that learning why someone downvoted you is worth the time, page space and distracting everyone else.

<Actually, don't you see how ironic it is that you're using this argument at this particular point in the thread? This post is subordinate to a comment I made that said:

Why is this downvoted twice? No wonder he's concerned about it. Sheesh.

Which was on behalf of wedrifid's point, which is largely your point. Wedrifid's post now has "+4" points. I think this represents a dynamic of social thinking that is unhealthy without policing. [I have something drafted about policing, but based on what I've learned in this post, it's not ready for mainstream consumption on this site. This is another pertinent kind of inhibition, really.]

Wedrifid can correct me if I'm wrong, but I really doubt he cared that much about the -2 (or if he did care he didn't want to care). It's hard to when you have 20,000 in total karma. Policing of any kind requires effort. Karma, like anything else, can be over-policed. And like other things that are not of great significance any policing is often more trouble than it is worth. The point of karma is a) a collective moderation tool (mainly this involves downvoting spammers and trolls), b) a loose indicator of status to encourage effortful and intelligent contributions.

Wasn't the first sequence post about a "dojo"? This site is, as far as I can tell, the public dojo.

Yeah, this is a good point. The problem is, it isn't just that. And that's part of why individual upvotes and downvotes in many cases don't signal much-- people want different things from the site and vote accordingly. If enough people see a comment that noise will get lost and you'll get a pretty good estimate of the quality of the comment. An individual comment that is net downvoted can be noise as well. I have plenty of negative comments, as does wedrifid. Much of the time I deserve them but sometimes it's just a matter of my having information community doesn't have and not communicating it well, different priors or talking about something others don't care about. Overtime that noise gets washed away (think of all the comments that are voted up way more than they are worth, i.e. this)

I'm not at all against discussing changes to the site structure to make it better. I think there are a lot of things that people have agreed would improve the site but are slow to come about for coding reasons. You might try looking through some of the older meta threads and see what was discussed then.

That seems like a new and valid way to make wedrifid's point.

I have an undeveloped intuition that building the site that way hurts the content of the site, not just me. But I'm not sure about that.

Plus, I've yet to see an argument for showing upvotes and downvotes separately beyond "Aren't you curious?!"

Come to think of it if I had this information I'd probably use kibitzer to just calculate total upvotes and display only that in the place of 'karma', suppressing downvote information entirely.

Come to think of it if I had this information I'd probably use kibitzer

Aha! You didn't know either. It's not just me. I conclude that it is (probably) a failure of some aspect of the system (e.g. the newbie orientation system).

Question: how does your comment relate to what you quoted? [I think I'm having trouble disambiguating your "this"]

If the upvotes are shown separately or, equivalently, with upvotes and 'total votes' are shown, I would use scripts such that what I saw was just the upvotes.

Right. If you are empowered to change the default, it won't matter to you so much what the default is. (Fair paraphrase?) I too like the idea of letting us control the information we subject ourselves to (re: Siren's again) and it sounds like that's actually built in on some level. I have to go read how it works [but I don't have time right now; I keep saying that and talking instead. hmmm. bad. putting down the keyboard.]

[-]matt10

There's some weight behind this proposal. Consider modifing the Anti-Kibitzer (http://lesswrong.com/lw/1s/lesswrong_antikibitzer_hides_comment_authors_and/1hvk) to do what you want (or adding a ticket to request same - http://code.google.com/p/lesswrong/issues/list).

Consider modifing the Anti-Kibitzer ...

Thanks for mentioning these things. I didn't know the Anti-Kibitzer existed. Shouldn't this be basic information for newbies? Did I miss it? [more when I have time, maybe]

Gah! these older posts cited by matt may make everything discussed here irrelevant (as action items; not as discussion for it's own sake or some larger purpose perhaps). I think we have a serious information liquidity problem on this site.

I deserve some of the blame for not "doing scholarly due diligence" to find this post sooner, but come on, I'm a newb. (~30 days maybe, most of them idle; the last couple ridiculously active) and "prior work" is something I repeatedly asked for.

Hmmm. I was about to conclude there's a newbie orientation problem and a lack of information liquidity (I mean the property that what's known over here becomes known over there on the site). But then I got some data:

This is the 48th comment and it took ~2 days.

In the scheme of things, that's not that bad. I learned a lot. Hopefully I didn't waste too much of everyone else's time. I enjoyed the interaction. Rediscovering things can be instructive.

Thanks matt! I'm going to edit the initial part of the OP to reflect your information.

I'm making a swift change of judgement here. Please let me know if it's wrong.

There's some weight behind this proposal.

My proposal was essentially the proposal "keep the status quo, let upvotes hide downvotes, no more information please!". I take it you are considering the extended option "don't let me see my karma score below 0 at all", which is also an idea with potential.

Consider modifing the Anti-Kibitzer (http://lesswrong.com/lw/1s/lesswrong_antikibitzer_hides_comment_authors_and/1hvk) to do what you want (or adding a ticket to request same - http://code.google.com/p/lesswrong/issues/list).

I'm sure I'll be tempted to next time I endure an irritating number of downvotes for obnoxious seeming reasons. I've modified the anti-kbitzer script previously to hide my own karma and inbox notifications to reduce the addictive feedback mechanism.

Although it bothers me that this keeps coming up and adding noise to the site, upvoted for agreement, because I'd like to see it too.

Although it bothers me that this keeps coming up and adding noise to the site, upvoted for agreement, because I'd like to see it too.

Sorry to add noise. Vincent pointed out one old discussion, I take it there are others too? How could I find them?

Maybe if topics keep coming up, it's evidence that there are important issues to address that aren't being addressed (weak evidence in this case; you could also say we need to improve the guidance on the welcome page, or searching, or that I didn't do due dilligence, or ...)?

And if important issues aren't being addressed, maybe we need to know why and figure out how to mobilize solutions?

EDIT: Of course one problem with this discussion point is that there are many ways to do it, and there's no particularly clear best way, and just like politics (or syntax) everyone has an opinion. Someone has to decide and actually implement, and they have limited time. The decision doesn't have to be democratic or anything, I agree.

It's come up several times in comment threads, if I recall correctly, but the thread Vincent pointed to is the only actual post dedicated to it I believe.

My preference would be just two numbers: total number of votes N (up plus down) and total positive score P (up minus down). Advantages:

  • Two simple, direct measures of the two main things one might care about: how much interest a post generated, and how well received it was.

  • Keeps the current karma score as one of the two variables.

  • Doesn't include new things more difficult to calculate with such as ratios

  • No redundant information; number of ups and downs can be easily calculated if wanted as one half of the sum and of the difference of N and P.

  • However, does not display explicitly the number of downvotes, for those who don't care to know it. Most of the time, I wouldn't. I might be too inclined to start wondering the reason for each of them, or worse, who did each of them.

Like this: "5 points / 7 votes"?
(Would mean 6 positive votes and 1 negative vote.)

Yes, although funny enough, seeing how it would actually look makes me less convinced that I want that and more attached to the current display (maybe it just triggers my status quo bias centers).

To me it indicates how showing derived facts instead of direct facts can be more confusing than helpful. I make an exception for showing the net point total because the current dynamics at least give some kind of rough notion of overall response and I don't want to introduce confusion, just add a bit more of the information back.

EDIT: the notion of direct I'm using here is not precise, which is part of the problem with this issue.

I might be too inclined to start wondering the reason for each of them

I'm new here, but isn't that exactly the point? To know where one's argument has a hole?

[Moreover, if people are dishing out minus votes for illegitimate reasons, this is a pernicious problem, no? That we should at least endeavor to have some mechanism to distinguish from legitimate disagreement. I agree we don't want to be too much distracted in the meantime.]

[[I'm deliberately punting on the "who did each of them" part of your concern, because that's more complicated and not what I'm proposing. Obviously it's important that the system knows who is who, on some level, so that bad actors can be discredited.]]

Wow, I'm such a newb. Did you know there's a "save" button at the end of posts? Well I didn't. And I don't know what it really means yet (but my weakly informed posterior is it's really useful). Where are all these things documented?!?

Tapping out for ~2 days to test my time-scale hypothesis (and, you know, sleep {mostly a joke, but I've become addicted, clearly})

What does that Save button do? I clicked it, but nothing happened.

Click either 'Main' or 'Discussion' (just below the site header), and a 'Saved' link will show up right underneath. That brings you to posts that you have saved.

Hmm. That really should have been Addicted. Anyway, Thanks Vincent! (I had no idea either bbleeker, and I think that's exactly the right question.)

Holy Crap! I had mail and I never knew. If, on the panel on the right, under your score, the letter icon is RED, you have mail. That is all here until the time is up.

It's not necessarily private mail. Responses to your comments (like this one) also appear there.

Good point. Also. To Whomever it might concern, I still haven't read my mail. I have a cognitive backlog to work through first. It may not be a bad thing that it wasn't obvious to me to click the Mail. Back to my backlog. Not back here yet. The only feature I want right now is a way to search my own posts for keywords so that I can go back to the right thing to Express something that I've only just figured out or that I think Needs to be said. But that's laziness mostly. I can get it done.

The way I do it is I use Wei Dai's incredibly useful script: http://www.ibiblio.org/weidai/lesswrong_user.php to view all of my comments at once, then use my browser's search functionality.

Sometimes I google for "lesswrong TheOtherDave keyword"

Thanks Again!

A lot of people say show sum of votes and ratio. I'd go for total counts of votes, and ratio. Total counts is more interesting to me than net positive/negative.

I figured that leaving the "points" alone would keep things simple, and leave the social dynamics of the site mostly unchanged. My proposal adds redundant information (i.e. up and down) to that. Mainly I wanted down, but directness argued for both. With "ratio" I worry that it will obscure things with small counts. E.g. a 85.7% post with seven votes leaves me wondering how many people didn't like it ( six up and 1 down). I'd rather interpret things myself.

Any of those would give you the others.

You realize that while this is true, it is not really the point, right?

Of course. I'm excellent at making true but useless statements.

Or do something more clever / opaque involving Bayesian weighting and probable "earned" score / controversy rating.

It'd be great if it would calculate a measure based off how the comment is to be appreciated by the kind of people whose contributions I myself appreciate. This may even mean counting as a positive downvotes by people whose votes are systematically correlated negatively with mine when it comes to appreciating the work of others.

I don't think this would even be all that hard, just data + math intensive.

Actually it would make a kind of fun programming/statistics project. Tempting!

(The contents of this post aren't relevant to OP at all or even germane to this thread except that they are a "reaction" to them)

having some metric for how controversial a post is would be useful... (other comments)

Thanks for this sequence of ideas. More... later.

(Bakkot) (The contents of this post are no longer relevant to OP except insofar as both discuss changes to the voting system.)

Interesting disclaimer technique: Is there an etiquette about disclaimers FAQ? [This time I mean the implication TheOtherDave.]

Is anyone familiar with an existing forum that does this? If so, how well does it work?

Stackoverflow, does not show plus and minus separately, I think, but they're a relevant data point. (I don't know a data point that fits your requirements). Their social dynamics seem pretty well engineered.

E.g. http://mathoverflow.net/questions/68100/using-bayes-theorem-gives-a-probability-1 I do not know how they handle down vote events in detail. Thanks for bringing up this avenue of attack. I recall that there is a "meta" stack overflow, that is probably the right place to go for further information on this. (I'm out of energy for today. I'm thinking that my energy tomorrow will be better spent building prototypes (not that I have time for that right now). Then we can have a more concrete discussion and issues of what needs to be done to this site right now won't be so immediate a concern.)

[-][anonymous]10

Stackoverflow shows up/down votes to users with at least 1k rep.

Here's a screenshot, http://techencoder.com/wp-content/uploads/2012/06/stackoverflow-up-down-votes.png

And here's the meta discussion, http://meta.stackoverflow.com/a/1007/131032

[This comment is no longer endorsed by its author]Reply

Proposal that doesn't require changing the site to be useful:

Would it be possible to obtain a data dump of a matrix of how users have scored posts? E.g. one row per user i (no id), one column per post j, and the i,j entry is +1, -1, 0 according to how user i scored post j (up, down, or no score)

Substantially better would be for the i,j entry to be (+1, -1, 0, NA) (up, down, no score but visited the post, no score b/c never visited the post).

Obvious objection: privacy concerns

Obvious response: prepare an algorithm and ship it to the Less Wrong server and run it there and only report back summaries that somebody trustworthy vouches for as sufficiently clean.

Examples like this: Netflix prize

Objectives: (1) post recommendation, (2) curation of a list of "best posts" with more refinement than just "posts with highest score"; e.g. best posts for different "kinds" of users, or for some sort of high-karma consensus user.

Aside: we can do better than a Netflix analysis. They phrased the problem wrong:

  • predicting how you would score a post given (1) your scores on posts you have read and (2) that you have also read the post we are "predicting" for
  • better: predicting how you would score a post given (1) your scores on posts you have read before some date and (2) that the post of interest is a "future" post that you may or may not actually read in the absence of an intervention that makes you read it (trickier; subtly different; needs a prior / model or better data much more)

Observation: the scoring data isn't public, but the who commented on what, effectively is.

Available alternatives: look what posts users you like wrote. Others?

Request: Please describe / link to examples of abusive use of voting by responding to this post. Are there multiple categories of abuse? Examples of bad reactions to criticism are also welcome.

One of the most interesting things about this site is the karma scoring, and that it reflects (to a greater degree than you see elsewhere) an objective assessment of the merits of an argument.

Ghmm. To me it provides information about community's beliefs. Here's an example . This had something like -7 +3 , first I assumed -2 was because of lack of clarity, edited it, and it got to -7 , after which I assumed LW simply dislikes the notion.

Ghmm. To me it provides information about community's beliefs.

Interesting. I think popularity clearly plays a role and that policing can be an important mechanism to correct that. (the example you cite has drifted up to -1 now). The example from this thread was at -2 before policing and is now at +5. Would that have happened without policing? I don't know.

I guess the bigger question is how to make the policing worth everyone's time. I think this requires a more clever mechanism than one-person, one-vote. E.g. a referee system calls for judgments on an issue from "experts" and not for a popularity contest. This is just the beginning of the sorts of changes that are needed, though.

As an aside, here are the results of my experiment in which I left this thread alone (for many weeks actually) and waited to see what would happen. The main result was your post (which is now at -2, sigh.). I think I'll up vote it to -1. I share your concern for the role of popularity in the voting. I don't think the voting is as rational as the majority wish it was (even in net and even after time, though less so perhaps esp. with policing). I'm still looking for a website which wouldn't reflect popularity in its voting, but would still allow for an organic (i.e. self-organizing) ranking system. I've drafted some plans but I doubt I'll take the time to realize them -- too many other things to do.

This might be a mistake. Effectively leveraging the masses without being vulnerable to their whimsy could be a powerful tool for humanity.