[EM] RE : Re: Simmons' "solution" of voting system design puzzle is inadequate

[Abd ul-Rahman Lomax]

At 12:23 PM 1/21/2007, Kevin Venzke wrote:
>I find it puzzling, maybe even humorous, that you let these three
>paragraphs stand when by the time you finished writing this message,
>you had realized that clones are not generally thought to have to be
>exactly identical.

That's right. I let it stand.

>I find it a little less humorous that you criticized Chris for
>"inconsistency" when he was just noting that Range did not appear to
>satisfy *Warren's* definition of clone independence. I don't see Chris'
>personal position on ICC expressed at all here. If you felt that clones
>should be exactly identical candidates, the half sensible thing to do
>would have been to side with Chris and criticize Warren.

ICC is considered a valuable criterion because we intuit that there 
is something wrong with identical candidates shifting the result 
outside the set of identical candidates or toward it.

In Ranked methods, the obvious definition of "identical candidates" 
is that they occupy the same rank with respect to all other 
candidates. The definition was written for ranked methods, and that 
is a rational definition for ranked methods. Range, however, does not 
rank, it rates, and ratings have more precision, generally, than 
ranks. (At least with Range 100 they do.) This increased precision, 
combined with a definition written for ranked methods, makes it 
appear that Range does not satisfy ICC. However, if we look at the 
*purpose* of the criterion, it *does* satisfy that purpose.

We have seen similar problems with the Majority Criterion, for 
example. It is written, one might say carefully, to define the 
preference of the majority in such a way that Plurality satisfies it 
and Approval does not. Yet Approval allows the majority the same 
freedom that it has under plurality to elect its preference. All it 
has to do is to express its strict preference, and its favorite will 
win. So why does Approval not satisfy the MC? As with this situation 
with ICC, the definition is written in such a way as to allow Range 
to fail while certain ranked methods will satisfy it.

It is far more useful, in terms of what actually is needed in 
elections, to consider a clone as identically rated, within the 
resolution of the system.

It was this difference between the ordinary meaning of clone and the 
technical meaning as used in ICC that caused my confusion. And it was 
a useful confusion, as far as I can see. It exposed certain issues, 
as I find my mistakes often do.