The problem is that a single upvote-downvote axis has to capture two different, sometimes orthogonal, judgments

I would like to see a HN-like site with a 2 part Karma system.

It would appear as three numbers A and S:N where A = Agree, S = Signal and N = Noise. S:N would not really be a ratio, and the Karma calculated from the three numbers could be any function, and wouldn't necessarily just consider S:N as a ratio.

But voting would effectively be a 4-choice radio box, so you could only pick one of the 4.

    - Agree     (+A +S)
    - Disagree  (-A +S)
    - Signal    (+S)
    - Noise     (+N)

Agree/Disagree will move the Agree number up or down and also add a point on the Signal number. Signal will up S. Noise will up N.

I suspect that those who try to game this system and censor comments by voting Noise as opposed to Signal will get overwhelmed by the Agree/Disagree voting on Signal.

I also suspect that Signal could just add to Karma, and Noise could just subtract from it. This might be able to exhibit the same dynamics as up/down voting on HN.

Another idea is to put the Noise vote button away from the up/down voting arrows and put it next to "Flag". The Signal control should still be next to Agree/Disagree, however. (Maybe a little star or sunburst? Wait, I know, a Sine Curve! Noise could have a square wave icon next to it.)

I proposed this before on HN, but I can't find it with SearchYC.

Signal/noise is another good way to express the 'valuable' axis -- terms with a long history in net discussions.

I would still like to be able to agree/disagree without giving a 'signal'/karma boost. A comment might be true, but trivial/superfluous/repetitive -- that's an agree-but-no-signal. (In the case of snark or offtopic trolling, I might even say I agree-but-that's-noise.) Similarly, something that's both a rule-transgression (noise) and egregiously wrong (such as dangerous misinformation) should get disagee-and-noise.

Noise voting should take precedence over agree voting.

Perhaps an even better idea would be to not have the Agree/Disagree voting add directly to signal, but for the score to have some affect on Karma and comment visibility through some function.

Also, this scheme might be to your liking: http://news.ycombinator.com/item?id=613520

It might be interesting for you to look at the rating system that's used on the site RSDN.ru (Russian Software Developer Network). The site has been around for quite some time and they've formulated a very balanced approach to rating.

On that site one has 7 buttons to rate a message: 1, 2, 3, '+', '-', '+1' and ':)'

1, 2 and 3 affect score (=karma) of the message and its author based on the level of the person who rates it (i.e. 2 from a 5th level person would add 5*2 = 10 etc)

'+' and '-' show that you agree/disagree with that comment, but do not affect the score.

'+1' adds 1 point to score independent of rater's level.

':)' shows that you find this comment funny.

I translated rules from this page http://rsdn.ru/forum/Info.aspx?name=info.forum.rating and as you may see, there are more rules that help to prevent excessive you-rate-me-I-rate-you behaviour, keep score up-to-date (your level depends only slightly on your all-times score but greatly depends on the marks that you've got during the last month) etc.

If people are interested, I can translate the rest of the rules.

Interested!

This has some of the same qualities as Slashdot, where you tag one of several reasons on your moderation.

EDIT: I just thought of a more user-directed mod/metamod system along the lines of Slashdot, but this way would be more Web 2.0. (As opposed to web 1.99a like Slashdot.)

Have a choice of pre-defined tags (about 10), each with a karma score attached to them. These tags would be long the lines of: "Funny" "Interesting" "Insightful" "Logical Fallacy" "Troll" Users could choose to either moderate a comment, which would be attaching 1 or 2 tags to a comment and thus altering the comment's visibility rating either-or meta-moderating, which would be voting on the fairness of moderation. So you could see what tags are attached to a comment and vote fair/unfair on each of them.

Now here's the rub: you get Karma by fairly moderating. (And lose it by unfairly moderating.) So while moderating changes a comment's visibility, only meta-moderating affects Karma.

Karma would be rewarded by the number of tags one could apply to a comment.

EDIT: More ideas. Users should be able to define their own tags, but only the admins would decide what point score should be attached to them!

A very interesting example, thanks!

Some of these ideas are are instances of the pattern: redirect more of what would otherwise be a cluttering one-word 'attaboy!' or 'haha!' or 'this doesn't belong' response into a single-click and running-tally.

I see that RSDN.ru has done that with ':)'; Facebook has done it with 'Like'; HN and others have done it with 'flag' (though this flagged tally is usually non-public).

I find it funny that you'd vote a comment "right" in order to indicate that you found it "wrong". ;-) I'd rather reverse them, so you mark evil comments as sinister...

I've struggled and gone back and forth on that, and it's somewhat arbitrary -- compared to the obvious up-down orientation for signal/noise votes. The fact that questionnaires and poll-charts are more likely to put 'yes/agree' on the left and 'no/disagree' on the right swayed my proposed orientation, but I could be flipped.

I also think rather than just a net tally, results could be shown as a teeny tiny inline horizontally stacked barchart -- a sparkline of sorts -- with the overhang of the leading sentiment in its vote-direction. For example, with left-agrees, this would be more agrees than disagrees:

  ++++++++
      ----

More disagrees than agrees:

  +++
  --------

But this sort of display might argue for right-agrees; number-lines and cartesian axes go positive to the right, and right-motion in some way signals acquiescence/continuation in the direction of the reviewed text.

edit: yeah, I'm pretty sure I agree too, I'm not sure why I wrote so much...

I think strict Up-Down voting ultimately regresses to a Boo/Yay judgement; despite whatever benevolent community coercion exists to color the vote along some more complex institutional axis. (In so far as we're talking about a free public forum... substitute virtually any conception of "free" you like. e.g. if this were a pay-for community we could probably implicitly afford to have a higher expectation for the "respect" for institutional rules.)

I think the concept of providing additional axis of voting could lead to very interesting community conversations. I think having a voting system that was usably different would drive adoption of the additional institutional rules (e.g. voting via the four cardinals would be interesting enough to be "forthwhile" somewhere on the web) .

something like: ( ) I liked this comment. YAY! ( ) I'd like this comment to get a response. YAY for conversation! ( ) I do not think this comment needs a response. BOO for conversation! ( ) I disliked this comment. BOO!

* Though I'm not sure what to make of the third one. These may not be great exemplars.

I think that this would make for a very interesting experiment -- have two up/down arrows next to each post -- one for "I Agree/Disagree" and one for "This is Valuable/Worthless".

After all, the single up/down does combine several dimensions into one judgment.

Agreevoted.

Maybe instead of cardinals, you could have:

(Up Arrow) "?", "!" (not necessarily mutually exclusive) (Down Arrow)

But what would be more interesting is to not force any meaning on them whatsoever. And wait and see if some sort of meaning emerges from the community.

Yes, exactly. I kind of assumed that adding a second voting axis was a nonstarter, but that would indeed provide more useful information about each comment.