[EM] ignoring "strength of opinion"

[Abd ul-Rahman Lomax]

At 06:31 AM 12/3/2005, James Gilmour wrote:
>So you think that just because I feel more strongly than you do in 
>my liking for A and my dislike for B, just because I
>shout about it more loudly than you do, and just because I mark my 
>ballot paper with bigger numbers than you do, my view
>of A and B should have more effect on the outcome than your view?

Mr. Gilmour is sloppy about "bigger," and does not seem to have 
actually considered what a Range ballot would look like.

"Loudly?" We are talking about a voting method which allows voters to 
*rate* candidates. Ratings are necessarily smaller or larger, unless 
you reduce them to binary, which apparently Mr. Gilmour prefers.

If you have a ballot which asks you to rate the candidates, and you 
are informed that the candidate with the highest average rating will 
be elected, and you want to make your vote count with maximum 
strength, then you will vote the maximum rating on the ballot for 
that candidate. The ballot might even have language making this 
explicit, with 0 being equated to "least preferred" and the maximum 
rating being equated to "most preferred."

In Range, each voter's rating of a candidate has equal strength in 
determining the average rating of that candidate. However, in terms 
of how the vote resolves pairwise, and suppose that we have Range10, 
i.e., ratings of 0-9 possible, then if a voter ranks A as 5 and B 
as  6, the voter's vote will have less impact on the A/B pairwise 
election than one who ranks A as 0 and B as 9.

If you care strongly about the pairwise election between A and B (and 
assume that A is a favorite and B is not, but is a front-runner), you 
may vote 9 for A and 0 for B. But if you vote 0 for B, you do run the 
risk of helping to defeat B in a pairwise election with another 
candidate. Range actually encourages honest ratings, Mr. Smith has 
that right. I suspect it could be proven that the optimum rating *is* 
the expected utility of that candidate winning. For one's vote to 
have the maximum effect, if the voter desires this -- yes, perhaps, 
most will -- one need only ensure that the maximum rating is given to 
at least one candidate and the minimum rating to at least one. Every 
other rating is intermediate. If you don't like Range, then simply 
vote Approval. I.e., binary Range. Maximum = Approval and Minimum = 
Disapproval.

>If we are going to weight the effect of our respective contributions 
>to determining the outcome on nothing more than how
>strongly we say we feel about the respective merits of the 
>candidates, we really will open a Pandora's box.

Range is simply a voting method. It isn't Pandora's box. It would not 
fill the world with woe.

Fundamentally, Mr. Gilmour does not want to *allow* voters to express 
less than maximum-strength votes. I'd assume that full-strength 
voting would be the norm. I.e., normally, everyone would be voting as 
"loudly" as everyone else. But if someone wants to whisper, those who 
think they know better than that someone want to prevent it....

this is more of the same old same old.

>I have no problem with different voters expressing their preferences 
>with different weightings: for example, for
>candidates A, B & C:  voter1: 1, 2, 3;  voter2: 1, 99, 100;  voter3: 
>1, 999, 1000.  The voting system can take these
>different weightings into account in "allocating" each voter's vote, 
>but each voter must make the same contribution to
>determining the outcome as every other voter, no more and no 
>less.  If that doesn't fit with social choice theory -
>tough!

This is really weird. Let's first look at what a *different* 
contribution to the outcome would mean. If we think that class A of 
voters is wiser than class B, then we simply count the votes of class 
A twice in determining the final votes.

But if we aren't going to do that, and we aren't (except in the 
special case of proxy voting, where voters may indeed have weighted 
votes, but only because they actually represent multiple voters), 
then we have two reasonable choices that I can see. We can normalize 
the votes. This was mentioned before, and Mr. Gilmour does not seem 
to realize that normalizing is what he might be asking for. However, 
normalization is substantially more complex to count and compile, and 
if the ballot is clear, the general agreement of those studying Range 
voting seems to be that it is not necessary. I described above how a 
ballot might make it clear that, to have maximum impact, the most 
preferred candidate should be rated with the maximum rating.

However, if we don't normalize, it is *still* true that each voter 
has the same effect on the final votes. The votes are simply 
totalled, or averaged, which amounts to the same thing. No vote is 
weighted more than another. However, a voter who votes "3" for a 
candidate when the ballot allows up to 100, say, is essentially 
voting against that candidate. (It seems Mr. Gilmour may intend low 
numbers to indicate increased preference; that is not how Range is 
normally presented, though it could be.)

I don't think that Mr. Gilmour has given much thought to how Range 
would work and how voters would use it. Range is well-known in Olympic scoring.

Frankly, almost any election method would work if followed by a 
ratification vote.... Range voting, in general, could make a very 
good polling mechanism for suggesting who would be named in the 
ratification vote. If a majority of voters agree, after having seen 
the Range votes, then it is clear that the election is fair....