[EM] Poll Ballots, from poll-topics poll

[MIKE OSSIPOFF]

>MIKE OSSIPOFF wrote:
>
> > 1. I should specify that if the method that you've designated returns
> >    a tie, then, instead of using Random Ballot to solve that tie, your
> >    ballot simply gives an Automatic Approval vote to each of the winners
> >    that your designated method has returned, and to everyone whom
> >    you've ranked higher than at least one of those winners.
>
>That rule works OK for cases where I rated the two tied candidates
>closely.
>
>What if I vote A>B>C>D, and there is a tie in the method I designate
>between A and D? Voter's Choice would end up executing a poor
>Approval voting strategy on my behalf. Random is somewhat better in
>that I get at least a 50/50 chance of having an A-only vote placed.

That's true. I didn't know what would be best for that. I figured that
you might need even the worse one. But if it's tied with something
you like better, then of course it makes more sense to help only
the one you like better, if there's a good chance that those 2
will do similarly in the final Approval count.

>
>Maybe a better method would be to pick the midpoint between my
>A and D ratings (or whichever two candidates tied), and approve those
>candidates I rate at or higher than this midpoint.
>
>I wonder if this method might also encourage voters to rate the
>candidates more sincerely? Say I rate A=100, B=70, C=50, and
>D=0. Then a tie between A and D would select A, B, and C,
>and a tie between A and C would select A only. A tie between
>B and C would select A and B. And so on.
>
>Another possibility -- Take the mean of all the candidates I ranked
>higher than the lowest tied candidate, and choose the candidates that
>are above the mean.

There's also a case for saying that all the winners chosen by your
designated method are probably popular enough so that they all
have a good chance of being close rivals in the final Approval count.
If so, maybe just give a final Approval vote to your favorite of those
winners by your designated method, and for everything you prefer to
that.

For your Approval votes in the final Approval count, extend your
approval set down to...

1. To your most preferred of the winners chosen by your designated
   method.
2. To the candidates above the median or midrange of the winners by
   your designated method,
3. To the candidates above the mean of the candidates that you
   ranked higher than your least preferred of the winners chosen by
   your designated method.
4. Apply Richard's 1st few-voters Approval strategy approximation
   to the winners chosen by the method that you designated.
5. Apply Richard's 2nd few-voters Approval strategy approximation
   to the winners chosen by the method that you designated.
   (Get your estimate for the number of votes cast by other voters
   by applying Richard's 1st Approval strategy approximation to all
   the ballots, based on their expressed ratings of the candidates).

Richard, do you have a clear favorite among these? Maybe this calls
for a procedural vote.

The changes that I suggested for the next poll seemed uncontroversial,
like including a ballot section for pure merit, and using a better
Random Ballot procedure. But this question of how to deal with ties
seems like more of a difficult decision. Let's get some opinions on
which of those 5 alternatives (or some other alternative?) are better.
If there doesn't seem to be consensus, we could vote on it. If we
vote on it, my inclination would be to use Cloneproof SSD, but
if not everyone considers Cloneproof SSD best for this poll, then we
should use Voter's Choice--and hope that none of the designated methods
produce more than 1 winner. But maybe there will be a consensus of
opinions expressed, at first or soon later.

#4 sounds best to me, if Richard's 2nd Approval strategy approximation
is better than his 1st one. Then #3 seems 2nd best.


Mike Ossipoff


_________________________________________________________________
Get your FREE download of MSN Explorer at http://explorer.msn.com