You gain karma points if you write something that people value, and you lose karma points if you write something that
people think is inappropriate.
Except people sometimes use karma in ways that do not conform to this story. For example I had a case of a single person downvoting an entire page (from the user view) worth of comments of mine in the space of about 1 minute. These comments were quite different topically from each other. What are the chances that this user read all these comments, found each of them objectionable, and then voted accordingly? Near zero. What more than likely happened is either:
(a) an emotional reaction
(b) some sort of game theoretic defection or response to a perceived defection (?)
(c) Someone really didn’t like something I said, and so decided a “greater punishment” was in order.
Karma is at best a noisy measure. But it’s worse than that because there are no enforced standards for how karma should be used. If I look at a post and see that some users voted it down, that gives me no information about what went wrong with the post. Not only do people have different interpretations/thresholds for when something should be downvoted, but also people downvote “globally,” regardless of merits of individual posts, based solely on user name.
The best interpretation for karma I can find is “a number that roughly correlates with how you [note: not your specific post!] are keeping all the coalitions on lesswrong happy.” Is this a useful number? Not really. I generally ignore karma ratings on my posts, and generally (not always) don’t vote myself.
My suggestion is to have clearly articulated community norms about when something ought to be upvoted or downvoted. Luke’s quote above is not clear enough, and as a result, the karma measure is far too noisy/meaningless, especially considering that you build other things on top of it, like the controversial −5 point rule for threads. The other suggestion is that the mods should make a distinction between what the karma system ideally should be, and what it is.
Except people sometimes use karma in ways that do not conform to this story. For example I had a case of a single person downvoting an entire page (from the user view) worth of comments of mine in the space of about 1 minute. These comments were quite different topically from each other. What are the chances that this user read all these comments, found each of them objectionable, and then voted accordingly? Near zero. What more than likely happened is either:
(a) an emotional reaction
(b) some sort of game theoretic defection or response to a perceived defection (?)
(c) Someone really didn’t like something I said, and so decided a “greater punishment” was in order.
Karma is at best a noisy measure. But it’s worse than that because there are no enforced standards for how karma should be used. If I look at a post and see that some users voted it down, that gives me no information about what went wrong with the post. Not only do people have different interpretations/thresholds for when something should be downvoted, but also people downvote “globally,” regardless of merits of individual posts, based solely on user name.
The best interpretation for karma I can find is “a number that roughly correlates with how you [note: not your specific post!] are keeping all the coalitions on lesswrong happy.” Is this a useful number? Not really. I generally ignore karma ratings on my posts, and generally (not always) don’t vote myself.
My suggestion is to have clearly articulated community norms about when something ought to be upvoted or downvoted. Luke’s quote above is not clear enough, and as a result, the karma measure is far too noisy/meaningless, especially considering that you build other things on top of it, like the controversial −5 point rule for threads. The other suggestion is that the mods should make a distinction between what the karma system ideally should be, and what it is.