[EM] Matthijs van Duin criterion

[Matthijs van Duin]

On Tue, Mar 06, 2007 at 08:50:36PM -0500, Warren Smith wrote:
>Subject: [EM] Matthijs van Duin criterion

I'm flattered, but would anyone outside the Netherlands know how to 
pronounce that? ;-)


>Equivalent criterion: removing a candidate pairwise defeated by current 
>winner, cannot alter the winner.

As you pointed out, this is not equivalent.  It's the *new* winner who 
must not defeat the removed candidate...

An intuitive notion is that the new winner didn't win before because he 
was being "suppressed" by the removed candidate, who used to defeat (or 
at least tie) the new winner.

Violation of this feels awkward:  if the new winner also defeated the 
removed candidate, then why couldn't he win before?  How was the removed 
candidate preventing this?


Like I mentioned before, my criterion implies LIIA, implies Smith, 
implies Condorcet and Mutual Majority, so complying with my (on the 
surface rather simple-sounding) criterion, you get lots of goodies as a 
package deal.


>are there any pure rank-order methods that obey this criterion?

That's an interesting question.  I've been thinking about that too, but 
haven't come up with an answer yet...

-- 
Matthijs van Duin  --  May the Forth be with you!