[EM] Why the concept of "sincere" votes in Range is flawed.

[Paul Kislanko]

-----Original Message----- From: Abd ul-Rahman Lomax 

At 07:19 PM 12/1/2008, Paul Kislanko wrote:
>PS. This is what I don't like about approval. In my generalized
>voter-friendly ballot, Approval requires me to vote A=B=C=D... when I
really
>like A a lot better than the others. But that method doesn't have any way
>for me (the voter) to tell it that I do. So no matter how an approval count
>turns out, I'm likely to believe my vote didn't matter.

Of course. However, you haven't expressed the situation completely. 
Approval is quite like Plurality, it has two ranks, Top and Bottom. 
So with, say, eight candidates, you are ranking, in Plurality, 
A>B=C=D=E=F=G=H. In Approval, you may rank A=B=C=D>E=F=G=H. Please 
don't compare Approval with ranked systems that allow equal ranking, 
except to note that it is the simplest of them. Obviously, more ranks 
allows more expression.

--------------------

I probably did not make myself clear, but we're saying the same thing.

>Approval is quite like Plurality, it has two ranks, Top and Bottom. 

Plurality has top and { non-top }. Approval has { top } and { non-top }
where the {}s indicate that there can be more than one entity described by
the atribute. Plurality allows only one 'top' choice, so is qualitatively
different from Approval, which allows a subset of alternatives with order >1
as/in "top." (That I have to say "as top" vs "in {top}" is the qualitative
difference.)

>Please don't compare Approval with ranked systems that allow equal ranking,


I didn't compare any methods. I was careful to note that from the _voter's
perspective_ a ranked ballot that allows equal rankings can be counted by
any method. If I say A>B=C=D>E>F=G=H then if you want to count by Approval
you just use A=B=C=D>E=F=G=H. Note that Approval loses the information I
provided with A>B=... for my acceptable choices and the E>F=... for my least
objectionable dis-approved choice. 

That explains why _as a voter_ I worry about whether my vote expresses my
intent when the counting is done by Approval. Yes, I'd prefer any of {A B C
D} to any of {E F G H} but of the former A is my first choice and if none of
my approved-set wins I'd be less likely to emigrate because E won of my set
of non-approved alternatives. 

What I actually suggested was that ranked ballots with equal rankings
allowed can be counted by any method. If you allow truncation, you can even
infer from A>B=C=D>E that I prefer E to F=G=H where { F G H } lists the
alternatives that I most specifically DO NOT want to win. You can infer that
from my ballot, because I didn't include them as = to the first of
alternatives of which I do NOT approve.

The ranked ballot with equal ranks and truncation allowed is the most
general vote _collection_ method that makes it easy for the voter. You can
count those ballots using any method you want.

Except Range, of course. Since it requires an extra input from _the voter_
for each alternative (whether ranked or not? whether the voter has expended
the energy to contrive a rating for the alternatives she doesn't care
about?) it requires too much work for me, without even getting into the
strategic situations.