From my informal understanding of human psychology these changes will make karma much more strongly desired. I don’t think a spiral around karma is what we want; much more importance on karma, in fact, and we might see something analogous to what search-engine optimisation is doing to internet content.
Incidentally, what is search-engine optimization doing to internet content? I’ve noticed machine-generated non-content scoring pretty highly on searches, is that what you’re referring to? Or are there more subtle, pernicious effects I’m not entirely aware of?
I’m not sure, but keep in mind that karma score isn’t exactly ‘how good a post is’, it’s a proxy—and intelligences optimizing for a proxy measure can end up stomping on your supergoals.
Edit: There’s a new utility-function proposal—have your AI simulate each member of LW, and have them up/downvote its ideas!
I think it’s more of an effect of making them a limited resource at all than an effect of making them a resource that’s correlated with being approved of by a high-karma LWer.
Making upvotes a limited resource means that there will be fewer upvotes in total, and slower karma gains, and thus each point of karma that one gets will be more meaningful, and a stronger incentive to ‘do more like that’.
Kind of like how if you have an income of $10,000/week, $1 doesn’t mean much, but if your income is more along the lines of $500/week, $1 is much more significant.
There’s precedent for making changes with this effect. It used to be that you could vote on (and would automatically vote up) your own comments, and those points did not evaporate when new comments started to appear at 0 karma without the option for the poster to vote on them.
Fair enough. If a change in the karma system was worth doing, this issue is unlikely to tip things back in the other direction: it would have to be really borderline.
From my informal understanding of human psychology these changes will make karma much more strongly desired. I don’t think a spiral around karma is what we want; much more importance on karma, in fact, and we might see something analogous to what search-engine optimisation is doing to internet content.
Incidentally, what is search-engine optimization doing to internet content? I’ve noticed machine-generated non-content scoring pretty highly on searches, is that what you’re referring to? Or are there more subtle, pernicious effects I’m not entirely aware of?
This is a pretty good overview—in particular, the last paragraph under the heading “The Downward Spiral: Industrializing OBP Exploitation”.
Great, thank you! That pretty much matches my expectations, but the specifics were quite interesting.
Is content designed more specifically to be liked by others really a problem?
I’m not sure, but keep in mind that karma score isn’t exactly ‘how good a post is’, it’s a proxy—and intelligences optimizing for a proxy measure can end up stomping on your supergoals.
Edit: There’s a new utility-function proposal—have your AI simulate each member of LW, and have them up/downvote its ideas!
Which is why my proposal isn’t necessarily linked to Karma Score. You can just give everyone the same number of votes.
I think it’s more of an effect of making them a limited resource at all than an effect of making them a resource that’s correlated with being approved of by a high-karma LWer.
I don’t follow. How does making upvotes a limited resource make karma more strongly desired?
In exactly that way.
Abundant things are not valued. Scarce things are, or may be. This is why gold is usable as currency and rocks are not.
Making upvotes a limited resource means that there will be fewer upvotes in total, and slower karma gains, and thus each point of karma that one gets will be more meaningful, and a stronger incentive to ‘do more like that’.
Kind of like how if you have an income of $10,000/week, $1 doesn’t mean much, but if your income is more along the lines of $500/week, $1 is much more significant.
It would also skew total karma scores to users who posted heavily before the change.
There’s precedent for making changes with this effect. It used to be that you could vote on (and would automatically vote up) your own comments, and those points did not evaporate when new comments started to appear at 0 karma without the option for the poster to vote on them.
Fair enough. If a change in the karma system was worth doing, this issue is unlikely to tip things back in the other direction: it would have to be really borderline.
True.
Agreed. I shouldn’t have elaborated so much on implementation.