[EM] RV comments

[Abd ul-Rahman Lomax]

At 01:49 AM 7/21/2007, Michael Ossipoff wrote:
A silly thing to say. Every criterion is a measure of election 
success, because all criteria (at least the ones that are any good, 
the "results criteria") judge methods according to what they will or 
won't do. A criterion failure is the opposite of success. By a 
criterion, success is not having a failure.

Failing an election criterion does not define an unsuccessful 
election, except as defined by that criterion. Election criteria in 
general -- there is an exception -- are based on some assumed 
desirable characteristic, often quite abstract, and if it turns out 
that this assumption is incorrect, or at least not correct in all 
cases, then the criterion fails to accurately measure the success of 
an election.

I am assuming that there *is* such a thing as election success or 
failure, independent of all criteria, at least in some cases. That 
is, there are situations where most reasonable people, given full 
knowledge, would agree that a certain outcome was optimum and that 
other outcomes are suboptimal. There may be situations where 
reasonable people would disagree.

The most obvious example is the Majority Criterion. I'll define it a 
little more strictly than it is usually defined. If a majority of 
voters prefer a candidate, and are not forced by strategic 
considerations to suppress that preference, and accordingly express 
it, the preference of the majority must prevail. Approval does not 
satisfy this criterion as stated, nor does Range. Any 
Condorcet-compliant method does. It is a rank criterion; that is, it 
is independent of preference strength.

And it can fail spectacularly. That is, there are election scenarios 
where the majority preference is quite clearly the wrong choice, *and 
the voters will agree if asked*. If the voters had full knowledge of 
the views of other voters, they would change their preference based 
on what they know of the *preference strengths* of others. And this 
is what people *routinely* do in cooperative action.

This is not, as people like Ossipoff have asserted, "altruism," and I 
would not recommend that anyone change their vote in an ab initio 
Range poll, for example, based on opinion about the preferences of 
others. Feeding a Range poll distorted information about personal 
preferences will distort the outcome. *However*, subsequent process 
*after* that poll may involve shifted preferences.

And as I have written many times, I've seen this in action, where 
what was initially even a strong preference, by a majority, shifted 
based on knowledge of the overall preferences of the entire group. In 
this case the "Range poll" was an Approval one, and it appears that 
all voted sincerely, so the information was good. And that 
preferences were strong initially is based on what people said about 
them, with words like "over my dead body...." And my belief that they 
shifted was based on the *unanimous* ratification of the ultimate 
choice. Contrary to what they had said, those people did not die. At 
least not immediately!