I’ve been wondering how to “fix it” but I have nothing concrete.
Letting go of the assumption that every user account’s votes should have the same weight would probably go a long way. I’m not saying such a measure is called for right now; I’m just bringing it up to get people used to the idea if things get worse.
Letting go of the assumption that karma means much above −3 would also go a long way. Karma is just here really to keep trolls away. If there are vast differences in Karma scores posted from around the same time, then maybe that means something. I know personally that the comments and posts I am most proud of are, generally speaking, my least upvoted ones.
To consider an example, this and this were posted around the same time, both to discussion. The former initially received vastly more karma than the second. But the former, while amusing, has virtually no content. The second is a well reasoned, well supported post. Did the former’s superior karma mean that it was a better article? Obviously not. That’s why the second was promoted and, once it was, eventually overtook the former.
Another obvious example is the sequences. Probably everyone here would agree that at least 75 of the best 100 posts on LW are from the sequences. But, for the most part, they sit at around 10-20 karma. Those that are outside that are the extraordinarily popular ones, which are linked to a lot, and sit at probably around 40 karma. This is not an accurate reflection of their quality versus other articles that I see around 10-40 karma.
I really try (but don’t always succeed) to vote karma based on “Is this comment/post at a higher or lower karma score than I think it should have?”. If everyone used this, then Karma scores might have some meaning relative to each other. But I don’t think many people use this strategy, and the result is that karma scores are skewed towards more read and funnier posts. Which generally tend to be shorter and less substantial.
Letting go of the assumption that karma means much above −3 would also go a long way. Karma is just here really to keep trolls away.
When a comment I make is not upvoted to at least +3, I give a moment’s consideration to the question of what I did wrong (and delete some of the comments that fail this test).
Some of your comments should be useful to the elite but not the masses. Such comments are only sometimes voted to +3. E.g., IIRC you regularly make decision theory comments that don’t go to +3, so it seems you don’t follow this rule even when talking about important things.
(It’s only semi-related, but who cares about the votes of the masses anyway? You’re here to talk to PCs and potential PCs, which is less than 1% of the LessWrong population. You’re beyond the point of rationality where you have to worry about not caring about NPCs becoming a categorical rule followed by everyone. On that note, you should care about the opinion of the churchgoer more, and the LessWronger less. Peace out comrade.)
Would it be difficult (and useful) to change the voting system inherited from reddit and implement one where casting a vote would rate something on a scale from minus ten to ten, and then average all votes together?
Difficult? Probably not. Useful is debatable. I’m not sure that the Karma system is important enough to consider in much detail. I just don’t see much low hanging fruit there.
Letting go of the assumption that karma means much above −3 would also go a long way. Karma is just here really to keep trolls away
I wonder how hard it would be to build a LW addon (like the antikibbitzer) that replaced numeric readouts with a tier label (e.g. “A” >=10, “F” for <=-3, etc.), and how using that would affect my experience of LW.
I think that would be pretty awkward, since posts would start in the “C” range. I think most people here would consider getting a “C” bad. But tiers make for an interesting concept, if you move away from grades.
Use the fact that the date is also displayed to give a ”?” instead for posts less than a day old, or have thresholds that move a bit for the first couple days.
Sure; the specific tier thresholds are secondary and might even be user-definable parameters, so people don’t need to know what tier they’re in on my screen, if knowing that would make them feel bad.
I would second that. On the other hand, how would you decide what weight to give to someone’s vote? Newcomers vs older members? Low vs high karma? I’m not sure a function of both these variables would be sufficient to determine meaningful voting weights (that is, I’m not sure such a simple mechanism would be able to intelligently steer more karma towards good quality posts even if they were hidden, obscure or too subtle).
What if the site just defaulted to a random sort order, so different people are presented with different comments first? That would still tend to bias in favor of older comments getting high presentation rank more. I’m not sure that’s such a bad thing, though.
Letting go of the assumption that every user account’s votes should have the same weight would probably go a long way. I’m not saying such a measure is called for right now; I’m just bringing it up to get people used to the idea if things get worse.
Letting go of the assumption that karma means much above −3 would also go a long way. Karma is just here really to keep trolls away. If there are vast differences in Karma scores posted from around the same time, then maybe that means something. I know personally that the comments and posts I am most proud of are, generally speaking, my least upvoted ones.
To consider an example, this and this were posted around the same time, both to discussion. The former initially received vastly more karma than the second. But the former, while amusing, has virtually no content. The second is a well reasoned, well supported post. Did the former’s superior karma mean that it was a better article? Obviously not. That’s why the second was promoted and, once it was, eventually overtook the former.
Another obvious example is the sequences. Probably everyone here would agree that at least 75 of the best 100 posts on LW are from the sequences. But, for the most part, they sit at around 10-20 karma. Those that are outside that are the extraordinarily popular ones, which are linked to a lot, and sit at probably around 40 karma. This is not an accurate reflection of their quality versus other articles that I see around 10-40 karma.
I really try (but don’t always succeed) to vote karma based on “Is this comment/post at a higher or lower karma score than I think it should have?”. If everyone used this, then Karma scores might have some meaning relative to each other. But I don’t think many people use this strategy, and the result is that karma scores are skewed towards more read and funnier posts. Which generally tend to be shorter and less substantial.
When a comment I make is not upvoted to at least +3, I give a moment’s consideration to the question of what I did wrong (and delete some of the comments that fail this test).
Some of your comments should be useful to the elite but not the masses. Such comments are only sometimes voted to +3. E.g., IIRC you regularly make decision theory comments that don’t go to +3, so it seems you don’t follow this rule even when talking about important things.
(It’s only semi-related, but who cares about the votes of the masses anyway? You’re here to talk to PCs and potential PCs, which is less than 1% of the LessWrong population. You’re beyond the point of rationality where you have to worry about not caring about NPCs becoming a categorical rule followed by everyone. On that note, you should care about the opinion of the churchgoer more, and the LessWronger less. Peace out comrade.)
Would it be difficult (and useful) to change the voting system inherited from reddit and implement one where casting a vote would rate something on a scale from minus ten to ten, and then average all votes together?
The biggest problem wouldn’t be technical, it would be the lower usability and the increased focus put on karma.
Also, averaging would be bad, showing the median vote would reduce the appeal of always putting +10/-10.
Difficult? Probably not. Useful is debatable. I’m not sure that the Karma system is important enough to consider in much detail. I just don’t see much low hanging fruit there.
So do I.
I wonder how hard it would be to build a LW addon (like the antikibbitzer) that replaced numeric readouts with a tier label (e.g. “A” >=10, “F” for <=-3, etc.), and how using that would affect my experience of LW.
I think that would be pretty awkward, since posts would start in the “C” range. I think most people here would consider getting a “C” bad. But tiers make for an interesting concept, if you move away from grades.
Use the fact that the date is also displayed to give a ”?” instead for posts less than a day old, or have thresholds that move a bit for the first couple days.
Sure; the specific tier thresholds are secondary and might even be user-definable parameters, so people don’t need to know what tier they’re in on my screen, if knowing that would make them feel bad.
I would second that. On the other hand, how would you decide what weight to give to someone’s vote? Newcomers vs older members? Low vs high karma? I’m not sure a function of both these variables would be sufficient to determine meaningful voting weights (that is, I’m not sure such a simple mechanism would be able to intelligently steer more karma towards good quality posts even if they were hidden, obscure or too subtle).
What if the site just defaulted to a random sort order, so different people are presented with different comments first? That would still tend to bias in favor of older comments getting high presentation rank more. I’m not sure that’s such a bad thing, though.