[EM] RE: [Condorcet] Can we come to consensus?

[Juho Laatu]

Hello James,

Few remaining notes. No strong opinions but maybe some fun for 
interested readers.

On Sep 10, 2005, at 03:30, James Green-Armytage wrote:

> 	"Approval" and "consent" are synonyms.

"Consensus" may come from Latin (con-sensus) and be related to word 
"sense" (=> joint opinion or something like that). Don't know about its 
current use in English.

> I think that "consensus" has a meaning that is quite separate
> from "majority rule". E.g. 55 voters: A>B; 45 voters B>A. A is a 
> Condorcet
> winner, but in my opinion not a consensus candidate.

I accept that. I wouldn't call A a consensus candidate either. I think 
Approval has the same problem in similar (""extreme"":-) situations. 
The most "consensus like" candidate should be elected if there is no 
candidate that fully fulfils the criterion.

> 	I think that the degree of consensus for a candidate is measured by 
> the
> number of people who view that candidate as acceptable. Hence I think 
> that
> approval voting comes closer to being a "consensus" method than any
> Condorcet variant.

It seems to me that your use of word "consensus" is closer to "many 
voters agree and smile" (=candidate at least acceptable in Approval) 
while I used it as "many voters agree but may still be a bit unhappy" 
(=maybe forced to compromise in Condorcet).

> As long as you're doing pairwise counts, it seems to me quite
> unnecessary and inelegant to choose from outside of the top set.

I agree that choosing outside the top set is inelegant, but for me that 
inelegance is of aesthetic nature. Drawing a tree-like figure where 
Smith set would be on the top and other candidates below like a trunk 
of the tree would be elegant. But from election method point of view I 
prefer to study the applicability of each candidate for the job one by 
one, in which case the "local" pairwise comparison information should 
maybe play stronger role than what kind of (long, non-local) transitive 
candidate changing incentive paths there might be.

I also don't see forcing the results to be in linear order (in the 
style that we usually require the voters to use in their ballots) 
necessary since we know that the group opinions (=end results) may be 
cyclic anyway (by the law of nature).

(The branches and leaves of the tree also tend to hide the relations 
(that may be strong defeats) between the top set candidates. Maybe I 
should develop some new drawing techniques (well, I tried some years 
back). Having the required four candidates unfortunately easily leads 
to three-dimensional drawings. One technique that I found almost 
acceptable was to show each candidate one by one and how much they are 
ahead or behind each of their competitors (this technique is of course 
more closely related to the local pairwise comparisons than to 
illustrating the paths).)

BR, Juho