[EM] Cycles in sincere individual preferences and application to vote-col...

[Kislanko at aol.com]

In a message dated 9/6/2004 1:48:49 PM Central Standard Time,  atarr at p
urdue.edu writes:

Paul  Kislanko wrote:

>The reason I wouldn't have chosen E over B, C, or D  on a ranked ballot 
>with A as an alternative is that B, C, D "trumped"  E on every issue that 
>was not the single one that A&E agreed  upon.

And yet, you already stated you would prefer E pairwise over any  of the 
three.  So why, oh why, would you rank E below them?  In  what meaningful 
sense do you rank any of those three over E?   Agreeing with them on more 
issues is irrelevant if the one issue you  support E on trumps all others.
I thought I'd explained that. Let me try again.
 
Issue 1 makes A>B>C>D and ties A=E, but that's only one issue.  Issues two 
through 10  makes B=C=D>E, so since A covers issue 1 and  issues 2-10 and E 
fails 2-10 and none of B,C,D fail issue 2/3, when A is in  the picture A>B>C>D>E. 
When A is NOT in the picture, my issue 1  puts E over B,C,D. (I guess I'd have 
to hold my nose, but that wouldn't mean I  wasn't sincere or 
logical/rational).
 



>But once A is out of the picture, there's one issue that E  trumps B, C, 
>and D on. And if A&E are both out of the picture than  my sincere ordering 
>of B, C, D could well change.

To which I  once again ask, why?  Why does the existence or absence of 
another  alternative change how you feel about B relative to C?  Note that  
even within a group, such reversal of preferences is not part of  
Condorcet's paradox.
See above. I've provided an example where one issue trumps all others  
because "I" am a radical pro-lifer. As long as there's somebody in the choices  that 
supports that and a lot of other of my issues, the one who supports attack  
weapons and capital punishment is lowest on my totem pole. But take out the one 
 who supports my position on those three issues, I'll move up the idiot who 
likes  AK-47s and capital punishment JUST because she also supports 
anti-abortion  legisletion. You can call me dumb for my original choice, but you can't 
accuse  me of being irrational if you want to dictate a voting method that 
ignores my  pairwise choices.
 



>  My original point was that you can;'t infer that I  prefer B>C from a 
> ballot that has A>B>C>D>E on it. If  you ask me which I prefer of B and C 
> (only) I might say C sincerely  because (in this example) C is the only 
> one that is both pro-gun  control and anti-capital punishment.
>
>When both of those are  covered by my first choice, I might rank C last 
>among B,C,D because of  something else, like fiscal policy.

At which point, I would accuse you  of being illogical.  Either you value 
the candidates' relative  stances on fiscal policies more, or you value 
their relative stances on  gun control/capital punishment more.  The 
presence of another  candidate who may be great or terrible on both of these 
issues should not  change your relative valuations of those issues.
No, but it would change my evaluation of the CANDIDATES. So to assume when  
all issues and all candidates are in play my ranked ballot means I ALWAYS 
prefer  C>D is unjustified. When comparing C and D based upon their stances on the  
issues, I might choose D>C, precise BECAUSE not all issues are in play.
 
The point is when comparing C and D the differentiator might be fiscal  
policies. The reason that wasn't my top priority when ranking all 5 is that I  had 
a different highest priority when all 5 are in play. 
 



>There's no reason to believe you can infer pair-wise wins  from a ranked 
>ballot voting method.

Well, that's what folks  have been doing since Condorcet's day.  You're 
welcome to question  it, and I'm welcome to disagree with you.

"Since Condercet's day" folks have been poking holes in his approach,  
beginning with his contemporary de Borda. The first example of a Condorcet  Winner I 
saw in literature was the alternative that was despised by 89 percent  of the 
voters. And we have Ranked Pairs, Beatpath, x-Quatas, etc. etc. etc. to  try 
to "fix" Condorcet-based methods, and they all give different results.
 
We need to axiomitize the study and stop saying "any input that is  illogical 
within the context of the method" is not worth worrying  about.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.electorama.com/pipermail/election-methods-electorama.com/attachments/20040906/1c545461/attachment-0003.htm>