Bass
Initiate
Posts: 196
Reputation: 5.76
Rate Bass
I'm a llama!
|
 |
The CoV reputation system
« on: 2006-12-09 13:29:22 » |
 |
I was just wondering here about how exactly the sumucratic reputation system here works. I have viewed the Introduction to Meridion page and Im a still a bit confused, but it really interests me and its kind of frustrating.
I have a few questions on it.
Firstly what does the formula e^(ln(N)*2*(R-5)/(9-1)) mean exactly?
Whats is the difference (and reason) between a level 1 voting or a level 5 or 9 voting?
What does N or N*N stand for? (Im guessing here that N refers to all those who have voted)
I also still dont get on how reputation is calculated in the way it is.
Also there are a few things which I can see as potential problems, though it may just be my ignorance speaking here.
Sumucracies promote competition between members and with the system of influence and weighted opinion, it's clear to see why. Now doesn't this mean that someone who attains a high level of influence, and therefore has the ability to override the influence of other individuals could easily twist and manipulate things to work towards their own ambitions and goals? Would this person not also be the perfect candidate for a dictator-like presence later on, depending on if their influence gains great heights?
And my second here is, how is someone's influence truly rated? Is their somekind of board that determines each individuals influence via somekind of evaluation test or process? If so, then how exactly are these board members elected and who is to say that they possess the competence to judge if any particular person has a certain amount of influence?
Sorry about all the questions, but I'm just really interested and its really frustrating that I don't understand it more fully.
Thanks.
|
|
|
|
|
David Lucifer
Archon
Posts: 2642
Reputation: 8.53
Rate David Lucifer
Enlighten me.
 
|
|
Re:The CoV reputation system
« Reply #1 on: 2006-12-10 00:14:14 » |
|
Quote:| Firstly what does the formula e^(ln(N)*2*(R-5)/(9-1)) mean exactly? |
The formula for influence is derived from the following assumptions:
1. There are N members registered in the system.
2. Everyone's vote should count for something (nobody has zero influence).
3. A new (unknown) member (reputation 5) gets 1 vote.
4. Someone with a perfect score (reputation 9) gets N votes. This is so that influence scales with increasing membership.
5. Someone with a perfectly bad score (reputation 1) gets 1/N votes.
Quote:Sumucracies promote competition between members and with the system of influence and weighted opinion, it's clear to see why. Now doesn't this mean that someone who attains a high level of influence, and therefore has the ability to override the influence of other individuals could easily twist and manipulate things to work towards their own ambitions and goals? Would this person not also be the perfect candidate for a dictator-like presence later on, depending on if their influence gains great heights?
|
There isn't much danger of that because someone with high influence can't become a dictator unless no one else has a decent reputation, and a dictator wouldn't be able to maintain their high influence without risking their reputation.
Quote:| And my second here is, how is someone's influence truly rated? Is their somekind of board that determines each individuals influence via somekind of evaluation test or process? |
There is no such board. Reputation is calculated from everyone's vote.
|
|
|
|
|
|
|
Bass
Initiate
Posts: 196
Reputation: 5.76
Rate Bass
I'm a llama!
|
|
Re:The CoV reputation system
« Reply #3 on: 2006-12-12 12:47:56 » |
|
Quote from: David Lucifer on 2006-12-10 00:14:14 2. Everyone's vote should count for something (nobody has zero influence).
|
I don't quite understand this. What about when someone first joins Meridion and has no votes? Where does there non-zero influence come from? Do they start of with 0 influence or something like 0.01 influence?
Also.
If we look at integral on the reputation board he has 0 influence...
0 * (e^((2 * ln(131) * (5.00099 - 5)) / (9 - 1))) = 0
PS I appreaciate both of your posts. Thank-you very much.
|
|
|
|
|
David Lucifer
Archon
Posts: 2642
Reputation: 8.53
Rate David Lucifer
Enlighten me.
|
|
Re:The CoV reputation system
« Reply #4 on: 2006-12-17 04:26:13 » |
|
Someone who first joins Meridion has a reputation of 5 (or close to 5 depending on how they rate themselves). The influence depends on another factor (besides reputation) and that is an activity level. You can read about the activity level in this announcement. integral has an activity level of zero therefore an influence of zero.
|
|
|
|
|
Sheldor
Magister
Gender: 
Posts: 9
Reputation: 6.47
Rate Sheldor
|
|
Re:The CoV reputation system
« Reply #5 on: 2009-05-20 19:41:52 » |
|
OK, but what algorithm do you use to calculate average weighted by influences? Do you solve recursive equation
reputation(n+1) = A . influence(n)
influence(n) = activity * exp (2 log N ((reputation(n)-5)/(9-1)))
until it converges (if it does, ...) each time someone changes his votes?
Or you just recalculate reputation based on votes and influences once per time? (Sorry for technical questions, but it is really nice idea, so I'm really interested how it is done..)
|
|
|
|
|
Sheldor
Magister
Gender:
Posts: 9
Reputation: 6.47
Rate Sheldor
|
|
Re:The CoV reputation system
« Reply #6 on: 2009-05-20 20:01:07 » |
|
Btw. this great Meridion system inspired me to study PageRank algorithm used by Google. If one would really want to solve recursive equation I wrote above for case where influence function is not linear (if it is, then it is approximately PageRank algorithm itself), its interesting problem to generalize it in such way that it would always converge to unique solution. So thank you for inspiration!
P.S. by matrix A in my post above I meant matrix of votes, of course.
|
|
|
|
|
David Lucifer
Archon
Posts: 2642
Reputation: 8.53
Rate David Lucifer
Enlighten me.
|
|
Re:The CoV reputation system
« Reply #7 on: 2009-05-22 00:03:35 » |
|
Quote from: irigi on 2009-05-20 19:41:52
OK, but what algorithm do you use to calculate average weighted by influences? Do you solve recursive equation
reputation(n+1) = A . influence(n)
influence(n) = activity * exp (2 log N ((reputation(n)-5)/(9-1)))
until it converges (if it does, ...) each time someone changes his votes?
|
That's right, the calculation is repeated until it converges (and it always converges rapidly, less than 10 iterations iirc).
|
|
|
|
|
Sheldor
Magister
Gender:
Posts: 9
Reputation: 6.47
Rate Sheldor
|
|
Re:The CoV reputation system
« Reply #8 on: 2009-05-22 08:00:38 » |
|
Ok, thanks a lot for your reply .. It seems it almost always has only one solution, which is really surprising for me, since it is nonlinear function. Maybe it has something to do with fact that exponential influence dependence is one-to-one function. But still... I will keep playing with it. 
|
|
|
|
|
Author
Logged
