[EM] Russ's criteriai postings

[MIKE OSSIPOFF]

I'd said:

"Approval would pass CC if CC were defined votes-only, as you
yourself said."

Nonsense. Any reasonable definition of the Condorcet Criterion...

I reply:

Nonsense. I'm not interested in Russ's subjective judgement about what is 
"reasonable". Anyway, regrettably the discussion hadn't been only about 
reasonable criteria, by my own subjective judgement.

Russ continues:

...either
assumes that ranking is allowed or says so explicitly.

I reply:

Russ, you had at your website for years a CC definition that didn't "assume 
that ranking is allowed or [say] so explicitly", stupid.

You've just recently discoverd that that CC definition wasn't "reasonable"? 
Yes, you concidentally discovered that right after I told you that you no 
longer had premission to have it at your website.

My preference version of CC doesn't say or "assume" that rankings are 
allowed. It makes no mention of methods' rules, and it applies to all 
methods.

The difficult thing about replying to Russ is that first one has to guess 
what he's trying to say:

What does it mean for a criterion to say that ranking is allowed? Maybe Russ 
is trying to say that the criterion stipulates in its premise that it only 
applies to rank methods. Or maybe he means that its requirement requires 
that the method be a rank method, otherwise the method fails. Who knows 
which he means? Who cares?

And, when a "reasonable" criterion doesn't say that, it "assumes" it, Russ 
tells us. What does it mean for a criterion to assume something that it 
doesn't say?? People assume things, but I've never heard of a criterion 
assuming something unsaid. But maybe Russ does. And if Russ thinks that 
people assume that reasonable CC definitions mean something that they don't 
say, presumably Russ knows that by using ESP.

Russ continues:

CC applies to
ordinal methods only. Duh!

I reply:

"Duh" is right. Blake's CC applies to ordinal methods only, because it 
explicitly says so. Other CC definitions apply to all methods, though not 
necessarily meaningfully. My preference CC applies meaningfully to all 
methods. What I mean by applying to all methods un-meaningfully will be 
clarified after a few paragraphs.

The preference definitions of CC that we sometimes find, which don't 
stipulate how preference constrains voting apply to all methods too, and all 
methods fail them. That of course is not the intent of those criteria 
definitions.

Some criterion definitions, including some CC definitions, speak of voters 
ranking candidates. For instance such a CC definition would say that if, for 
every Y, more people rank X over Y than Y over X, then X should win. Does 
that apply to all methods? That could depend on what the criterion's writer 
meant by the verb "rank" If "ranks X over Y" means "votes X over Y", then it 
applies to all methods, and Plurality passes. If "rank X over Y" means "Vote 
X over Y on a rank ballot", then there could be disagreement about whether 
it applies to all methods:

For a nonrank method, there's no such a thing as a failure example with that 
criterion, because there's no example that complies with the criterion's 
premise. Some would argue that that means that the criterion doesn't apply 
to nonrank methods, since it can't test them. A ham sandwich would pass that 
criterion, if it weren't for the fact that voting system criteria explicitly 
apply only to voting systems.

But I suggest that that criterion applies to all methods, and Pluraoity 
passes it nonetheless. The way to show that a method passes a criterion is 
to show that there can't be a failure example. There's no reason why we 
should make an exception to that simple rule just to contrive to say that 
Plurality doesn't pass that criterion. By the simple usual meaning of the 
term, Plurality passes that criterion...meaninglessly, like a ham sandwich.

I criticize criteria that Plurality passes in the same way that a ham 
sandwich would.

Of course, if the criterion had specified explicitly in its wording that it 
doesn't apply to nonrank methods, then we're saved the trouble of even 
asking if there's a Plurality failure example, and we can say that the 
criterion doesn't apply to Plurality. That's the case with Blake's CC 
(unless he's changed it since I last checked).

Russ continues:

Even with that assumption, a preference-style CC is *not* equivalent to
the standard ("votes-only") CC *unless* you stipulate that the actual
votes *are* the voter's true preferences.

I reply:

That's part of what makes it a "preference-style criterion", stupid. That's 
common knowledge. You had such a criterion at your website for years. A CC 
that says:

If there's a CW, and everyone votes sincerely, then the CW should win.

Of course someone might say that Russ means that _the votes-only CC_ should 
"stipulate that the actual votes *are* the voter's true preferences". Of 
course then it wouldn't be a votes-only criterion, since it would mention 
preference. But who can guess whether that's what Russ means?

Russ continues:

But if you do that, you might
as well just use the "votes-only" definition of CC.

I reply:

Here, it's anyone's guess what Russ means. You might as well use the 
votes-only definition of CC if it's equivalent to a preference version of 
CC? Well, if the criterion applies only to rank methods, as Russ spoke of 
earlier, then it obviously is not true that you might as well use that 
instead of the universally-applicable preference CC. A criterion that 
explicitly applies only to rank methods is not equivalent to one that 
applies to all methods. Preference CC compares Plurality to the 
pairwise-count methods, because it compares every pair of methods.

Now, if the criterion explicitly said that nonrank methods fail, then that 
might be equivalent to preference CC. But nonrank methods fail by fiat, 
while they fail preference CC for exactly the reason why we would expect 
them to. A rules criterion doesn't have convincing justification. "Nonrank 
methods fail because I say so", rather than by some actual failing that the 
criterion brings out.

Preference CC is the only CC that isn't a rules criterion. Likewise the 
preference versions of a number of other criteria.

Russ continues:

This is really just common sense, folks

I reply:

Oh, is that what it is :-)

In an immediately subsequent posting, Russ corrected himself:

He'd said:

Nonsense. Any reasonable definition of the Condorcet Criterion either 
assumes that ranking is allowed or says so explicitly. CC applies to ordinal 
methods only. Duh!

Then he added:

After thinking about it a bit more, I realized that my statement, "CC
applies to ordinal methods only," could be misconstrued by legalistic
pedants. Let me forstall their bleatings with a dose of common sense.
Strictly speaking, CC applies to non-ordinal methods too, but they fail
by *definition*.

I reply:

Russ is all confused.

A nonrank methods fails a criterion by definition if the critrerion says 
"Nonrank methods fail this criterion", or "Methods should have 
rank-balloting", etc. If the criterion says that it doesn't apply to nonrank 
methods, then nonrank neither pass nor fail.

Russ continues:

If you never went to Harvard, then is it true that you "failed to
graduate from Harvard"? Well, yes and no. You didn't graduate, but how
could be said to "fail" if you didn't even try? The same applies to
non-ordinal methods. You can say they "fail" CC, but then they didn't
even try.

I reply:

Russ isn't being quite clear with us about what criteria he's referring to. 
If the criterion is one that says explicitly that it doesn't apply to 
nonrank methods, then nonrank methods don't pass or fail.
That isn't the same as failing. But it's true that their non-failure isn't 
anything to be proud of. They're getting off on a technicality,escaping 
evaluation by that criterion. That's one reason why I don't like criteria 
that don't apply universally and uniformly to all methods.

I told James that replying to him is like being a highschool teacher in 
highschool for problem students. But replying to Russ isn't like that.  
Replying to Russ is more like trying to talk with a babbling, drooling 
person whose meaning can't even be determined.


Mike Ossipoff

_________________________________________________________________
On the road to retirement? Check out MSN Life Events for advice on how to 
get there! http://lifeevents.msn.com/category.aspx?cid=Retirement