[EM] Voting, Grading, etc.

[MIKE OSSIPOFF]

Forest Simmons wrote:

>>With computers available it would be easy to incorporate ratings by having
>sliders to click and drag up and down under the name of each candidate
>with the percent of approval showing in a little box above the candidate.
>When all of the sliders are set like the voter wants them, he/she clicks
>on OK. The Computer recaps the totals, and asks, "are you sure?" etc.

Yes, I've believed for some time that Cardinal Ratings might be
more readily accepted than the relatively unfamiliar idea of Approval
voting. After all, people are used to rating things from 1-10 or
from 0-10, 1-100, etc.

The graphical rating method that you suggest hadn't occurred to me,
but it sounds like something that could be a popular proposal.

That, and  Cardinal Ratings & Approval in general, avoids the issue
of how rank ballots should be counted. Even though there are some very
good rank-counts, the involvement of the IRVies in single-winner reform
might just make it impossible to get an adequate rank-count adopted.
But there's only one way to count the ballots in Cardinal Ratings or
Approval: Add them up. So it could turn out that Approval or Cardinal
Ratings could be the best candidate against the woefully inadequate IRV.

Comparing Approval and Cardinal Ratings (CR), Approval of course is
easier to implement, since it uses the same ballots we're using now
for Plurality. But CR is much more famililar, and therefore might be
more winnable. I'd say that only conversations or polling could tell
for sure whether Approval's simplicity & cost-free-ness of implementation
would win out over CR's vastly greater familiarity.

In my comparison of Approval & Condorcet in my previous posting, I
forgot to mention Approval's criterion compliances: Approval
will never give anyone incentive or need to vote someone over their
favorite. Approval & CR (which is strategically the same as Approval)
are the only methods known to have that guarantee.

Also, Approval & CR meet WDSC. The criteria that I've been mentioning
are defined at the website:

http://www.electionmethods.org

When I speak of Approval, it goes without saying that what is said about
Approval can be applied to CR, if it's about strategy.

When voters have no information about eachother, then the utility-
expectation-maximizing strategy in Approval is to vote for all of the
candidates who are better (as judged by you) than the mean of all the
candidates. Vote for all the above-average candidates.

When information is available about the probability that the particular
pairs of candidates will be the 2 frontrunners, then vote for all
of the candidates whose "strategic value" is positive.

To calculate candidate i's strategic value, sum the following
expression over all j:

Pij(Ui-Uj)

...where Ui is i's utility as judged by you, and where Pij is the
probability that i & j will be the 2 frontrunners.

Incidentally, in Plurality the optimal strategy is to vote for the
candidate with highest strategic value.

So Approval's strategy isn't more complicated than that of Plurality.

Of course, for most people, we can just say that in Approval they
vote for whom they would in Plurality, plus everyone whom they like
more, including their favorite.

Mike Ossipoff

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