I'm an algorithm designer, so I don't know about all of these changes but I do have a mathematical solution to PG's vote value problem:
before edit: Value of vote = Log[base:TotalUserKarma](UserAvgKarmaPerComment)
after edit: Vote Value = (Log[base:TotalUserKarma](AvgUserKarma)) * (Log[base:(25)](ThreadTop50Rank))
another edit: Vote Value = (Log[base:Median(AvgUserKarma)](AvgUserKarma)) * (Log[base:(25)](ThreadTop50Rank + 1))
EDIT: added a +1 to prevent a 0 value for the #1 thread. (unless you actually want votes not to count for the number 1 thread, in that case you can just pull out that "+ 1")
This would reward contributors with higher relative average karma per comment. If you have 0 karma your vote isn't worth anything and if you do have karma but your average is only 1, once again your vote isn't worth anything. You have to consistently contribute useful material to have a say whether or not someone else's material is userful to the community.
This seems to be a decent reflection of social circles in real life.
Try it out on users you think contribute very little and see if it is effective. Naturally, I don't have that data so I can only speculate.
The problem with average karma is stat it strongly discourages people from posting replies to less popular or older messages. I've made a conscious choice to ignore my average karma. I dont think it is conducive to good conversation, as it tends to be heavily biased towards the first few posts.
I wonder if we should go the other way -- I always liked slashdot's system of a 5 point cap on any post. Is a 500 point post really enter than a 100 point post, or is it more likely it is seen more often?
This is true, but it is a problem that currently exists.
For orthogonality's sake, I would suggest that if most of your posts come from 1:1 conversations, post by post, then perhaps there is a better medium to communicate with that person than through a board like HN.
The people that I have those conversations with are people I only know through HN, and the discussions come out of what happened here. Those discussions would not have naturally happened in another medium.
As an example I give you this discussion. In this thread I have 2 posts, both with 1 karma right now. I do not expect them to go up much if at all. Where would you suggest that I should have gone for this conversation with you?
EDIT: I changed the evaluation to not make the avg karma relative to total karma but instead to the median value of average karma of all users. The downside to this is that Median(AvgKarma) would need to be calculated quite often throughout the day because I doubt pg wants to pay for the resources to run that everytime someone votes.
The problem here is that someone who consistently provides high-quality (or at least frequently-upvoted) comments / submissions will likely have less of a contribution than someone with significantly lower karma. For instance: you would have (slightly) more influence than someone with 3000 karma and an average of 5.
Then this will reduce the amount of impact they have on the community as a whole. They will be doing us a favor by starting out ruining their karma.
Also, to state the obvious, every member here was a new member at one time and I'm not ready to say that is "typical" behavior. It's just very memorable behavior.
The whole reason why I submitted this was because PG said it wasn't the comments that cause issues but that bad comments were getting upvoted too much.
Actually, karma is very good market indicator for any board and filter bubbles tend to be specific to filtering out certain types of content. This won't be filtering anything by content.
before edit: Value of vote = Log[base:TotalUserKarma](UserAvgKarmaPerComment)
after edit: Vote Value = (Log[base:TotalUserKarma](AvgUserKarma)) * (Log[base:(25)](ThreadTop50Rank))
another edit: Vote Value = (Log[base:Median(AvgUserKarma)](AvgUserKarma)) * (Log[base:(25)](ThreadTop50Rank + 1))
EDIT: added a +1 to prevent a 0 value for the #1 thread. (unless you actually want votes not to count for the number 1 thread, in that case you can just pull out that "+ 1")
This would reward contributors with higher relative average karma per comment. If you have 0 karma your vote isn't worth anything and if you do have karma but your average is only 1, once again your vote isn't worth anything. You have to consistently contribute useful material to have a say whether or not someone else's material is userful to the community.
This seems to be a decent reflection of social circles in real life.
Try it out on users you think contribute very little and see if it is effective. Naturally, I don't have that data so I can only speculate.