Re: SD2 versus Condorcet/CSSD

Mark parashakti108 at
Sun, 10 Apr 2005 20:23:25 -0000

--"denis bider" wrote:

>D: Mark, everyone. I am concerned about a property of SD2, 
specifically, that it gives a restricted number of experts the power 
to make decisions that impact everyone.

-M: The maximum number for SD2 currently is five. The idea here is to 
give enough directors for people to compare the directors with each 
other. Keeping the number small centralizes accountability and makes 
for quick and efficent decision making.

-M: In a large organization, such as a government, the directors 
would have a large body of advisors and non-voting representitves. 
The directors' job would be to listen to everyone with impartiality 
the same way that a tribunal of judges would. 

>D:I have no qualms with the suitability of PageRank and other 
algorithms for coming up with a reliable top list of experts. 
However, it is my hunch/belief/opinion that a group of experts cannot 
be any better at deciding what is best for people than the people 
themselves can, if an appropriate aggregation algorithm is chosen.

-M: These experts are the people. PageRank finds the *locus of 
consent* - the experts would be at the *center* of the people.

>D: I propose that Condorcet / CSSD could be such an algorithm. What 
does everyone think about the merits of the Condorcet / CSSD voting 
method, as compared to SD-2? Specifically: - approach 1 is to use the 
Condorcet / CSSD method as an aggregating algorithm in a poll that 
asks people about an important decision, affecting everyone, which is 
to be made;

-M: Condorcet is still in-degree based, therefore it is still "Joe 
Sixpack" based.
>D: - approach 2 is to use the PageRank algorithm to determine a 
restricted number of experts, and then delegate the decision to those 
experts, as per SD-2. Which of the two approaches is better? See also 
a description of the Condorcet voting methods here: It also includes a 
reference to a further description of CSSD. I look forward to any 
discussion. Kindest regards, denis

-M: Denis, I like your approach - you are proposing competing 
algorithms. Even though I don't like Condorcet because of its in-
degree basis, I was thinking that it could be combined with PageRank. 
This wouldn't be for director selection, this would be for peer-
ranking among those that know each other. People would rank-order 
those that they know the best. Those higher ranked would have higher 
initial PageRank. Since PageRank is a reiterive algorithm, these 
initial values would change during calculations.

-Mark, Seattle WA USA