[EM] Re: Truncation, defeat strength

[Kevin Venzke]

Forest,

This is a good idea, but I doubt a great method will be devised which requires
complete strict rankings.  That method would either have to pass strong FBC
(seems impossible, I think?), or fail strong and weak FBC (seems very
undesirable).

> If we could come up with a satisfactory convention on filling in
> indifferent ballot preferences, then the margins/winning votes controversy
> would become moot.

Hmm, that's assuming the method needs a measurement of defeat strength.
For RP(wv), let's say, if I anticipate a cycle, I might want to tie some 
candidates to avoid it.  So having the method undo my strategic decision 
just to sidestep the defeat strength issue seems undesirable.


I wonder, has anyone considered RP(approval); that is, measure defeat
strength as the approval votes received by the victor?  This wouldn't
always elect the approval winner, as he could potentially be beaten by
anyone who beat him pairwise...  The practical difference is just that we're
loathe to lock in the victories of unpopular candidates.

If voters complain when a strong WV victory is overruled, it can be pointed
out that it's their own fault for not approving the victor.

I'm not sure how often or why RP(approval), SSD(approval), or "While there is
no CW, eliminate the Approval loser" would much differ.

Kevin Venzke
stepjak at yahoo.fr




 --- Forest Simmons <fsimmons at pcc.edu> a écrit : 
> On Thu, 4 Sep 2003, Forest Simmons wrote:
> 
> > Suppose that somebody comes up with a great method that requires complete
> > rankings, perhaps a de-cloned version of the Kemeny Order.
> >
> > Here's a suggestion for a convention on filling in the truncated ballots:
> >
> > Automatically rank the truncated candidates according to the candidate
> > card published by the highest ranked candidate on the ballot.
> >
> > So if a ballot comes in as A>B>C=D, and candidate A's published preference
> > order is A>D>C>B, then the ballot is completed as A>B>D>C.
> >
> > A more complicated example:
> >
> > Let's say a ballot comes in as A=B>C=D, and the published preferences of
> > candidates A and B are A>B>C>D and B>A>C>D, respectively.
> >
> > Then it is clear that C=D should be replaced with C>D, but how about the
> > A=B?
> >
> > We consult C's published preference order C>B>A>D, to break the tie.
> >
> > The result is B>A>C>D.
> >


___________________________________________________________
Do You Yahoo!? -- Une adresse @yahoo.fr gratuite et en français !
Yahoo! Mail : http://fr.mail.yahoo.com