[EM] proposal: weighted pairwise comparison

[Dgamble997 at aol.com]

James you wrote in part:

>Let's match some names to the numbers, for fun...
>45 Bush 100 > Lieberman 10 > Dean 0
>10 Lieberman 100 > Bush 90 > Dean 0
>5  Lieberman 100 > Dean 90 > Bush 0
>40 Dean 100 > Lieberman 10 > Bush 0
>
>    I imagine Joe Lieberman as someone who is not particularly liked by
>either the Democrats or the Republicans. (I'm immensely glad that he
>didn't win the primary, and I find it amusing how poorly he did after all
>his "electable Democrat" bullshit.) 
>    I think Lieberman would be a very crappy president. (And I think that he
>was the kiss of death on Gore's ticket.) Still, if I was convinced that I
>could only chose between Lieberman and Bush, I'd hold my nose and vote for
>Lieberman. At least his environmental record is okay...

Some people when faced with a choice between bad and worse will hold their 
noses and choose bad in order to defeat worse. Other people will refuse to make 
a choice between bad and worse.

Take the following example:

The results of the first round of the 1969 French Presidential election were:

Pompidou (Gaullist, centre right) 44.0%
Poher (Centrist)  23.4%
Duclos (Communist) 21.5%
Other Left (3)  10.1%
Other (1) 1.1%

The French Communist party advised its supporters not to vote in the second 
round run-off between Pompidou and Poher even though Poher was to the left of 
Pompidou.

The reason:

" We refuse to choose between cholera and the plague"

Roland Leroy (member of PCF central committee in late 60's/ 70's) 

The result- Pompidou won, turnout fell 10% between the rounds and 5% of 
people won voted in the second round spoilt their ballot papers.

Some people will vote bad to defeat worse, many others will not.

In my original post I was not intending to repeat low utility Condorcet 
winner arguments or high utility IRV loser arguments. I was trying to make the 
point that systems based solely on ranking can produce results that look bad when 
ratings information is taken into account.

Example 1

45 A 100>B 70>C 0
10 B 100>A 70>C 0
5 B 100>C 70> A 0
40 C100>B 70> A 0

Example 2

45 A 100>B 10>C 0
10 B 100>A 90>C 0
5 B 100>C 90> A 0
40 C100>B 10> A 0

For the examples above zero information Approval voting gives the best 
results- B wins in example 1 and A wins in example 2.

Of course it's easy to construct examples in which zero information Approval 
also produces bad results:

55 A 100> B 70> C 0
10 B 100> A 70> C 0
35 C 100> B 60> A 0

Result:

A  65
B  100
C  35

What I'm interested in at the moment is going beyond IRV, Condorcet and 
Approval towards a single winner system that gives high utility, generally 
preferred winners.

David Gamble





-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.electorama.com/pipermail/election-methods-electorama.com/attachments/20040612/ad2f58c5/attachment-0003.htm>