Consider the low-visibility Main an accidental feature, post in Discussion instead, and let the content speak for itself and get promoted to Main on merits.
It’s not clear to me this is karma-optimizing, because I believe that every vote you get while in discussion is not amplified when the article is moved to Main. So if you get 20 upvotes in discussion, and then it moves to Main and you get another 20 upvotes, then you’ve lost 180 karma vs. beginning in Main and getting the same number of upvotes. I don’t think starting in discussion results in getting almost as many people to upvote it while it’s in Main as upvoted it in Discussion.
Sure, but I assume that Kaj’s goal is maximizing his readership, not his karma.
Drucker/Peters/Leboeuf:
What gets measured gets done.
The system should be aligned so that maximizing for what is visible maximizes for what is desirable. This is in large part because optimization systems which target what is visible are much more reliable than optimization systems which target what is invisible. (How would Kaj have noticed that a surprisingly low number of people were reading the post, except by that it had a surprisingly low karma?)
Personally, my goal is much more in the “maximize karma” camp than the “maximize readership” camp. The main reason for that besides the increased visibility of karma (both to myself and others) is that karma has sign.
With 2 buttons: Upvote + Brain (I just learned something—use sparingly)
With three buttons: Eye (Worth seeing, make it more visible) + brain, + thumbs up (agreement)
With four buttons: Eye + brain + Heart (I wish to signal support ) + Checkmark (This is probably correct)
...and so on. To save space, you could double-click to indicate the reverse (make this less visible, this is incorrect, i condemn this). Visibility wise, eyes would function as upvotes. There’s diminishing returns on additional buttons, but I really doubt that the optimal number is one button set. I’m guessing optimal probably sits around 2, possibly 3.
Wouldn’t you really love to sort posts by the number of people who say they personally benefited or learned from them, rather than karma? Wouldn’t it be nice if opinion-popularity and visibility weren’t interchangeable? And if someone was up-voted for making a good argument which was nevertheless controversial, wouldn’t it be nice to know how many people actually agree? And heart/checkmark solve the whole “agree denotatively, disagree connotationally” thing, which I think happens much more frequently than people realize, and checkmark counts are good ways to assign confidence.
I understand the vilibility->optimization effect and I, too, feel the pull. However, I often make commenting or posting decisions which I expect to be karma-suboptimal, and so do many others here. Or at least that’s what I think I do. Much more often now than when I first started participating here. I assume that Kaj, with over 30k karma, would care even less about maximizing karma.
I assume that Kaj, with over 30k karma, would care even less about maximizing karma.
I certainly can’t speak for Kaj, but I’d hesitate to assume that people with a lot of karma care less about getting more of it. After all, caring a lot about karma is likely to increase someone’s chances of getting a lot of karma in the first place.
However, I often make commenting or posting decisions which I expect to be karma-suboptimal, and so do many others here.
As do I- I try to keep my upvote percentage on recent comments high, and so sometimes will avoid controversial subjects (unless it’s one of the controversial subjects that I have decided to always discuss, in which case I still try to comport myself in a way that minimizes downvotes), even though a +3-2 comment would result in an additional point of total karma.
But the primary impact karma-maximization has on me is it urgifies generating valuable content for LessWrong. I finished my Decision Analysis sequence as quickly as I did because doing so put me at the top of the Top Contributors, 30 Days list (which was only 5 people back in 2011), and if I had delayed the first post would have slipped to more than 30 days ago, and I wouldn’t have had enough to leap over lukeprog. I kept notes and wrote book reviews when before I would have just read books. And so on.
Huh, I guess some of us are rather more competitive than others… I treat the 30-day karma list about the same way I treat the points earned by my favorite sports team: something to enjoy when it’s up, but not something to base my decisions on.
Huh, I guess some of us are rather more competitive than others...
I tend to only use competitiveness as an instrumental goal, and so it only shows up sometimes.
I treat the 30-day karma list about the same way I treat the points earned by my favorite sports team: something to enjoy when it’s up, but not something to base my decisions on.
I think this should change when you are on your favorite sports team, or you’re not playing sports correctly :P
Karmawhoring, while fun for a while, tends to lead one into… non-optimal directions.
Isn’t that true by pejorative label?
I would consider the size of the replies thread (which is visible, though not as obvious as karma) to be a better metric of readership and impact.
Disagree. As an example, consider two posts I’ve written about Judea Pearl’s work, Causality: A Chapter by Chapter Review, and Understanding Simpson’s Paradox. Both of them have the same number of comments- 19. (I’ll give you a moment to savor the irony of doing a same-X comparison on a post whose primary discussion was about reverse regression.) I got 52 times as much karma for the first post, and it was probably more than fifty times as much work to generate, and probably more than fifty times the value of the second post.
Most of my technical posts get around that many comments- 0 to 50, say. Most of my nontechnical posts get many more comments, though, because most of my non-technical posts are things like Rationality Quotes threads (731 comments) or HPMOR discussion threads (953 comments). While some karma seems appropriate for those- since only one is up at a time, my having made one implies I put it up before anyone else, which increased the amount of time it existed- the amount of karma is roughly appropriate to the effort involved / value-added, whereas the number of comments is totally disproportionate to the value of my particular contribution.
In particular, the Understanding Simpson’s Paradox post highlights the weird distribution of comments. Oftentimes, threads are very short, but sometimes threads get very long- and generally, long threads have more heat than light involved, or have people slowly understanding each others’ positions rather than rapidly grasping them. Sometimes the slow way is necessary, but it seems unwise to say the slow way is preferred.
It’s a bit tricky because the points of view are different. The label is pejorative when applied to someone’s behavior externally and I am saying that even if you don’t care about labels applied to you by others, karmawhoring is unlikely to be a good strategy for yourself.
Disagree.
Well, we need to figure out what do we care about. You are saying that karma is better correlated with “effort involved / value-added” while I’m talking about “readership and impact”. I think it’s a whole separate discussion as to which particular metric LW should optimize for.
Website admins, by the way, should be able to produce number of unique views per post fairly easily.
The label is pejorative when applied to someone’s behavior externally and I am saying that even if you don’t care about labels applied to you by others, karmawhoring is unlikely to be a good strategy for yourself.
What I meant by that is that it seems unlikely to me that someone would identify “I did X because of Karma, and X is something I endorse” as karmawhoring. So, by definition, doing karmawhoring is unlikely to be a good strategy- like murder is guaranteed to be illegal, but killing is murkier.
You are saying that karma is better correlated with “effort involved / value-added” while I’m talking about “readership and impact”
To me, value-added is basically readership and impact, except with the readers giving some feedback on whether the impact was positive or negative. If you get a lot of people to read a random string of characters, and so you waste part of their day, this is a loss over those people not noticing a random string of characters that you generated.
That section was mostly the empirical claim that number of comments is a bad proxy for the value generated by the post, whether you use karma or readership or some other metric. I mean, if you want more comments in your posts, put in more typos (in order to not annoy your readers, have only one typo, and when someone comments with a fix, edit in a new typo), or instigate political fights in the comments.
Karmawhoring, while fun for a while, tends to lead one into… non-optimal directions.
Non optimal. It can’t be unless somehow max-karma happens to coincide with whatever other maxima is being searched for. But it is also unlikely to be net detrimental or a sufficient deviation from optimal to be worth focussing on as a problem.
I would consider the size of the replies thread (which is visible, though not as obvious as karma) to be a better metric of readership and impact.
That’s a better metric of controversy than impact. It is a reasonable indicator of readership.
Is it? Maybe the first few hundred points, I don’t know. Karma has no fungibility, and is a weak proxy for status (karma per post might be a better metric). What is it important for?
Consider the low-visibility Main an accidental feature, post in Discussion instead, and let the content speak for itself and get promoted to Main on merits.
It’s not clear to me this is karma-optimizing, because I believe that every vote you get while in discussion is not amplified when the article is moved to Main. So if you get 20 upvotes in discussion, and then it moves to Main and you get another 20 upvotes, then you’ve lost 180 karma vs. beginning in Main and getting the same number of upvotes. I don’t think starting in discussion results in getting almost as many people to upvote it while it’s in Main as upvoted it in Discussion.
Sure, but I assume that Kaj’s goal is maximizing his readership, not his karma.
Drucker/Peters/Leboeuf:
The system should be aligned so that maximizing for what is visible maximizes for what is desirable. This is in large part because optimization systems which target what is visible are much more reliable than optimization systems which target what is invisible. (How would Kaj have noticed that a surprisingly low number of people were reading the post, except by that it had a surprisingly low karma?)
Personally, my goal is much more in the “maximize karma” camp than the “maximize readership” camp. The main reason for that besides the increased visibility of karma (both to myself and others) is that karma has sign.
I’ve always wished there were more buttons.
With 2 buttons: Upvote + Brain (I just learned something—use sparingly)
With three buttons: Eye (Worth seeing, make it more visible) + brain, + thumbs up (agreement)
With four buttons: Eye + brain + Heart (I wish to signal support ) + Checkmark (This is probably correct)
...and so on. To save space, you could double-click to indicate the reverse (make this less visible, this is incorrect, i condemn this). Visibility wise, eyes would function as upvotes. There’s diminishing returns on additional buttons, but I really doubt that the optimal number is one button set. I’m guessing optimal probably sits around 2, possibly 3.
Wouldn’t you really love to sort posts by the number of people who say they personally benefited or learned from them, rather than karma? Wouldn’t it be nice if opinion-popularity and visibility weren’t interchangeable? And if someone was up-voted for making a good argument which was nevertheless controversial, wouldn’t it be nice to know how many people actually agree? And heart/checkmark solve the whole “agree denotatively, disagree connotationally” thing, which I think happens much more frequently than people realize, and checkmark counts are good ways to assign confidence.
I understand the vilibility->optimization effect and I, too, feel the pull. However, I often make commenting or posting decisions which I expect to be karma-suboptimal, and so do many others here. Or at least that’s what I think I do. Much more often now than when I first started participating here. I assume that Kaj, with over 30k karma, would care even less about maximizing karma.
I certainly can’t speak for Kaj, but I’d hesitate to assume that people with a lot of karma care less about getting more of it. After all, caring a lot about karma is likely to increase someone’s chances of getting a lot of karma in the first place.
As do I- I try to keep my upvote percentage on recent comments high, and so sometimes will avoid controversial subjects (unless it’s one of the controversial subjects that I have decided to always discuss, in which case I still try to comport myself in a way that minimizes downvotes), even though a +3-2 comment would result in an additional point of total karma.
But the primary impact karma-maximization has on me is it urgifies generating valuable content for LessWrong. I finished my Decision Analysis sequence as quickly as I did because doing so put me at the top of the Top Contributors, 30 Days list (which was only 5 people back in 2011), and if I had delayed the first post would have slipped to more than 30 days ago, and I wouldn’t have had enough to leap over lukeprog. I kept notes and wrote book reviews when before I would have just read books. And so on.
Huh, I guess some of us are rather more competitive than others… I treat the 30-day karma list about the same way I treat the points earned by my favorite sports team: something to enjoy when it’s up, but not something to base my decisions on.
I tend to only use competitiveness as an instrumental goal, and so it only shows up sometimes.
I think this should change when you are on your favorite sports team, or you’re not playing sports correctly :P
That’s my point, I am not playing for the team Karma, though I am a fan.
Karmawhoring, while fun for a while, tends to lead one into… non-optimal directions.
I would consider the size of the replies thread (which is visible, though not as obvious as karma) to be a better metric of readership and impact.
Isn’t that true by pejorative label?
Disagree. As an example, consider two posts I’ve written about Judea Pearl’s work, Causality: A Chapter by Chapter Review, and Understanding Simpson’s Paradox. Both of them have the same number of comments- 19. (I’ll give you a moment to savor the irony of doing a same-X comparison on a post whose primary discussion was about reverse regression.) I got 52 times as much karma for the first post, and it was probably more than fifty times as much work to generate, and probably more than fifty times the value of the second post.
Most of my technical posts get around that many comments- 0 to 50, say. Most of my nontechnical posts get many more comments, though, because most of my non-technical posts are things like Rationality Quotes threads (731 comments) or HPMOR discussion threads (953 comments). While some karma seems appropriate for those- since only one is up at a time, my having made one implies I put it up before anyone else, which increased the amount of time it existed- the amount of karma is roughly appropriate to the effort involved / value-added, whereas the number of comments is totally disproportionate to the value of my particular contribution.
In particular, the Understanding Simpson’s Paradox post highlights the weird distribution of comments. Oftentimes, threads are very short, but sometimes threads get very long- and generally, long threads have more heat than light involved, or have people slowly understanding each others’ positions rather than rapidly grasping them. Sometimes the slow way is necessary, but it seems unwise to say the slow way is preferred.
It’s a bit tricky because the points of view are different. The label is pejorative when applied to someone’s behavior externally and I am saying that even if you don’t care about labels applied to you by others, karmawhoring is unlikely to be a good strategy for yourself.
Well, we need to figure out what do we care about. You are saying that karma is better correlated with “effort involved / value-added” while I’m talking about “readership and impact”. I think it’s a whole separate discussion as to which particular metric LW should optimize for.
Website admins, by the way, should be able to produce number of unique views per post fairly easily.
What I meant by that is that it seems unlikely to me that someone would identify “I did X because of Karma, and X is something I endorse” as karmawhoring. So, by definition, doing karmawhoring is unlikely to be a good strategy- like murder is guaranteed to be illegal, but killing is murkier.
To me, value-added is basically readership and impact, except with the readers giving some feedback on whether the impact was positive or negative. If you get a lot of people to read a random string of characters, and so you waste part of their day, this is a loss over those people not noticing a random string of characters that you generated.
That section was mostly the empirical claim that number of comments is a bad proxy for the value generated by the post, whether you use karma or readership or some other metric. I mean, if you want more comments in your posts, put in more typos (in order to not annoy your readers, have only one typo, and when someone comments with a fix, edit in a new typo), or instigate political fights in the comments.
Non optimal. It can’t be unless somehow max-karma happens to coincide with whatever other maxima is being searched for. But it is also unlikely to be net detrimental or a sufficient deviation from optimal to be worth focussing on as a problem.
That’s a better metric of controversy than impact. It is a reasonable indicator of readership.
Karma is pretty important.
Is it? Maybe the first few hundred points, I don’t know. Karma has no fungibility, and is a weak proxy for status (karma per post might be a better metric). What is it important for?