[EM] Condorcet Criterion for plurality.

[MIKE OSSIPOFF]

Blake said:

>"MIKE " <nkklrp at hotmail.com>, on the subject of 'Re: [EM] Condorcet
>Criterion for plurality.', is quoted as:
>
> >In my previous posting about this, yesterday, I made it clear that
> >I'm not saying that Plurality can't be defined in Markus's
>contrafactual
> >way. I merely said that it's sloppy, shabby, & funky when a criterion
> >can't be made to apply to all methods without inventing a different
> >definition for some of them.
>
>Criteria and method definitions are just tools to help the analysis of
>methods.

And what, then, did you think that I believed they were?
Oh, ok, manifestos & contracts :-)

>A person has something they want to achieve in a method, so
>they write a criterion that they feel expresses the principle, but in a
>strict pass/fail form.  Then they must justify the relation between the
>principle and the criterion, and the desirability of the principle
>itself.

Except that the principle, which I call a standard, is completely
subjective. Either you agree with it or you don't. You don't
value LO2E, and I won't try to justify it to you.

As for relating the criterion to the standard, notice how people
are using the standard. If your criterion requires what they seem
to be talking about, then you have a good argument that your criterion
measures for their standard.
>
>I think we're in danger of treating a criterion as if it was something
>between a manifesto and a contract:  believing that like a manifesto, it
>must be worded to make the definition itself seem desirable,

Hopefully compliance with a criterion is desirable to people who
value the standard that it is argued to measure for. Would you prefer
criteria with which compliance is undesirable? Or not of interest?
Maybe you could write such criteria, but I can't guarantee that I'll
choose voting systems that comply with them.

>and like a
>contract, it must be written to prevent any possible misuse.
>
>If you find that a given definition of plurality

The "given definition of Plurality" that I use is the one that's
used by our "unscientific" election departments: The voter can mark 1
& only 1 candidate. The candidate with most marks wins.

>passes a particular
>definition of the Condorcet Criterion, or Independence from Irrelevant
>Alternatives, it doesn't make any sense to say GOTCHA, I've found a
>loop-hole.

Here's something that it makes sense to say: "The people use the
usual definitions of CC, are mistaken when theys say what complies
with it." It seems to me that the CC definition at
Blake's website is one that is met either by no methods, or by Plurality. If 
I'm correct about that, then instead of saying GOTCHA,
I've found  a loophole, I'll just say "Was that your intent?"
Maybe you knew about that, and maybe you'd just say that compliance
with CC doesn't count, or doesn't mean as much for some methods.
But that turns CC into something vague in its application. If something
meets CC, does Blake have a definite uniform rule to decide whether
or not that compliance means anything or counts? Or is it more of a
subjective judgement? I suggest that when you have to say that compliance 
with CC doesn't really count when Plurality complies, that
should tell you that you need a different CC definition.

A contract? Maybe you'd like to sue CC for breach of contract, because
it doesn't act as expected by many people who define it or cite it.

And for IIAC, my point was not that IIAC is no good because Plurality
meets it. My point was that IIAC means that Approval, and even
Plurality, are better than IRV, for all those people who bring up
Arrow's criteria.

>Criteria are only useful if they can be justified.

Another astounding pronouncement. Well I must admit that I'd have trouble
justifying a criterion for whose application we're expected to assume
that Plurality collects rankings.

>If you want to use
>different criteria in different circumstances, there must be some
>justification for your doing so, but it isn't necessarily "sloppy".

When, to apply criteria, you have one special rule for IRV, another
for Plurality, another for single-winner Cumulative, and maybe
another for Condorcet, that's worse than sloppy. There's no longer
any sense in which you can say that you're applying the same criterion
to all those methods. In your above-quoted paragraph maybe you mean
that it's ok to say that we have a set of different criteria:
IRV CC, Plurality CC, Approval CC, single-winner Cumulative CC, and
Condorcet CC.  :-) Well now, Tideman(m) passes Tideman(m) CC, but
Plurality fails Plurality CC. So what? Layton asked for uniformity,
and I have to agree with him that Markus's system (if it uses different
rules for different methods) is quite useless for compring voting
systems, because uniformity of criterion application rules is
necessary for comparing methods by criteria. As Layton said,
uniformity is needed. Uniformity of criterion application rules,
but not uniformity of balloting systems, when, in use, the methods
don't all use the same balloting system.

As I define them, criteria apply to all methods, without using different
application rules for different methods, and without assuming that the
methods are other than what they would be when actually in use.

Mike Ossipoff



_____________________________________________________________________________________
Get more from the Web.  FREE MSN Explorer download : http://explorer.msn.com