[EM] Re: Approval Completed Condorcet

[Forest Simmons]

As you say, voted ballots like this would be rare.  Here's why:

In the zero information case, it would mean that either the voters were
stubbornly ignoring reasonable strategy OR it would mean that all
candidates in the Smith set are extremely low utility.

In the partial information case, it would mean that voters are stubbornly
ignoring reasonable strategy OR that the partial information was more
disinformation than information, leading all the voters to believe that
their favorites were ahead by comfortable margins. 

(Like in Lake Wobegone, where all the children are above average.)

Forest

On Fri, 22 Feb 2002, [iso-8859-1] Alex Small wrote:

> I just realized a situation where Approval Completed Condorcet could cause
> a problem.  It's rare, but here it is:
> 
> Suppose that every voter is stubborn and only indicates approval for his #1
> choice.  If there are 6 or more candidates it's entirely possible that the
> Smith set could consist of 3 candidates who received NO first place
> rankings.  Consider the following:
> 
> 33%:  D>A>B>C>E>F
> 33%:  E>B>C>A>F>D
> 33%:  F>C>A>B>D>E
> 
> D, E, and F lose all pairwise contests against A, B, and C.  Meanwhile, A,
> B, and C are cyclic.  Unfortunately, none of them received any approval.
> How then to resolve it?
> 
> Granted, this is an incredibly unlikely situation.  But, if Florida ever
> holds elections via this method we should be prepared for the possibility
> that it will happen ;)
> 
> I'm in a hurry, but off the top of my head I can't imagine this occuring if
> we implemented Approval Completed Condorcet via Cardinal Ratings.  If you
> had, say, a scale of 0 to 5, everybody in the Smith set would have received
> at least a little bit of approval (at least some ratings of 1), so the
> situation above would seem to be impossible. (but I've learned that it's
> best not to say that a complication is impossible when discussing election
> methods, at least not without thinking for a while).
> 
> Alex Small
> 
>