'... Simple answer: yes... build a matrix of friends/enemies/etc and iterate over it until you have absolute values ...'
Slashdot tried a model similar to this. The result you still got polarisation of views, short snipes without any real value. I don't think it's simple. Why? You can't change "human behaviour". Anything you do there will be some person trying to find a loophole. Joshua Schachter (delicious fame) went into great lengths in his talk on itconversation about this.
Especially just with an algorythm approach. Though one thing I can think of that did more than anything for me was a points filter (meaning you can simply not view someones comments below a certain threshold ) . A more broader approach has better chances of encouraging (enforcing) good behaviour.
One successful approach I have seen, can be found in the perlmonks (a perl programming) site where users gain karma through numerous means (time on the system, posts and moderation) enforced by a top down 'Benovolent dictatorship' and in effect a hierarchy.
You join the site as a novice & work your way up earning rights through encouraged & enforced behaviour. For instance you have to be a certain level (say level 5) before you can front-page a post. Another level and you can join groups effecting the design of the site. This type of conditioning means for instance users don't have power beyond their maturity. If they do things out-of-line (say for instance random down-modding or excessive referencing) they can be pulled in before the inevitable kitten fights (al-la slashdot) begins.
In time, as a user reaches a certain level (of maturity) certain rights are allowed. So the combination of brute user moderation and a more subtle framework encourages finding solutions to problems rather than /. type brawls.
I don't think mere code alone solves undesirable behaviour in systems. And it would be a good area for developers with a 'psychology', 'group dynamics' background to explore and add more knowledge.
The system I pointed out is free from those flaws, for precisely one reason: karma in that sort of system is always relative.
That is, everyone moderates according to their own tastes, and nobody is 'better' than anyone else. It all sums to zero, the only filtering is by individual preference.
'The system I pointed out is free from those flaws, for precisely one reason: karma in that sort of system is always relative '
Can you give me some sites that use this (aside from skype)? Can you explain how say person 1, rates person 2 and how the rating is viewed from person 2's point of view? ... or say person 3 observing?
person 1 allocates one or many positive or negative points to person 2.
person 2 shows person 1 as having higher/lower karma, person 1 shows person 2 as having higher/lower karma.
Person 3 has no viewpoint unless they are in the system. There is no 'absolute' objective viewpoint without taking a stance.
A user with no ratings of anyone sees everyone with the same karma.
All four of the major extended interactions work too. (friend of a friend, enemy of a friend, enemy of an enemy, friend of an enemy)
Another property is that it doesn't matter how many points you use. You may rate someone up or down as much as you desire, or make up your own rating system, whatever works for you.
'... Person 3 has no viewpoint unless they are in the system ...'
So there is effectively no visible karma? - unless person 3 intersects person 1 and person 2?
That's a pretty good system. How do you measure (unless you have admin on the system) person 1 compared to person 2 and person 3? From the way I see this would it work on a system where a (visible) reward is given out for effort? (I define effort as creating links + comment).
The reason I ask is that how do you encourage users here to add content and create a leader board? I know the constraints are bit tight, but if you solve this it would be worth recommending to features.
Well, one way is to pick a viewpoint and use that. But really, the important thing is, you see that you are valued by the people you respect. As long as you're a user in the system, it's obvious whether you have good or bad karma.
Slashdot tried a model similar to this. The result you still got polarisation of views, short snipes without any real value. I don't think it's simple. Why? You can't change "human behaviour". Anything you do there will be some person trying to find a loophole. Joshua Schachter (delicious fame) went into great lengths in his talk on itconversation about this.
Especially just with an algorythm approach. Though one thing I can think of that did more than anything for me was a points filter (meaning you can simply not view someones comments below a certain threshold ) . A more broader approach has better chances of encouraging (enforcing) good behaviour.
One successful approach I have seen, can be found in the perlmonks (a perl programming) site where users gain karma through numerous means (time on the system, posts and moderation) enforced by a top down 'Benovolent dictatorship' and in effect a hierarchy.
You join the site as a novice & work your way up earning rights through encouraged & enforced behaviour. For instance you have to be a certain level (say level 5) before you can front-page a post. Another level and you can join groups effecting the design of the site. This type of conditioning means for instance users don't have power beyond their maturity. If they do things out-of-line (say for instance random down-modding or excessive referencing) they can be pulled in before the inevitable kitten fights (al-la slashdot) begins.
In time, as a user reaches a certain level (of maturity) certain rights are allowed. So the combination of brute user moderation and a more subtle framework encourages finding solutions to problems rather than /. type brawls.
I don't think mere code alone solves undesirable behaviour in systems. And it would be a good area for developers with a 'psychology', 'group dynamics' background to explore and add more knowledge.