[EM] Blake takes the low road...

[MIKE OSSIPOFF]

I guess Blake wanted to lower the level of the campaign discussion.

[Because of my mode of reply, there are no ">" marks, and so I
state who's talking]

Blake wrote:

From:  Blake Cretney <bcretney at p...>
Date:  Sat Apr 14, 2001 4:55pm
Subject:  Re: [EM] Some brief campaign argument

[...]

This list has had problems in the past.  For example, the whole issue
of the misinterpretation of Condorcet's words.  As far as I can tell,
the claim of Condorcet's method using Mike's "defeat-support" was
generally spread by this list.

I reply:

Are you still on that? I've agreed that Condorcet's interest in
probabilities suggests that he'd have used margins, but the fact
remains that, according to French usage in his time, defeat-support
better matches the words _in his drop-weakest method definition_,
though not elsewhere.

Now that I've agreed that, from his interest in probabilities, he'd
have likely used margins, can't that issue be laid to rest??

You yourself have discounted and criticized Condorcet's wordings and
proposals. So why do you continue to use them?

Contrary to your claim that all of Condorcet's wordings refer to
Ranked Pairs, the fact is that if you're right that Condorcet wanted
to solve all cycles in his drop-weakest method, then we get
something that drops weakest defeats till there are no cycles. I don't
think you want to propose that. So can we get off the issue of what
Condorcet meant???

Can we just agree that Condorcet wasn't as specific as we'd have liked
, and leave it at that? I'm more interested in what methods meet
criteria that measure for standards that I consider important, and
I suggest that you address that subject, and drop the issue of what
Condorcet meant.

Blake continues:

You still see this on a few web pages.
The important lesson here, is that you shouldn't believe everything
you read on this list

I reply:

Thank you for clarifying that lesson for us, Blake. You mean we shouldn't 
believe everything that Craig Carey & Don Davison say?
An astounding new suggestion.

We also shouldn't believe that the definitions at your website
are correct, but that's another subject.

Blake continues:

Realizing that Condorcet specified margins

I reply:

...but not in the wording of his drop-weakest proposal, according to
a French dictionary of Condorcet's time. But can we drop the issue of
what Condorcet meant?

Blake continues:

puts Mike's
"defeat-support" in a new light.  It suggests that since it is so new,
and peculiar to this list, it really is possible that we might rid
ourselves of it.

I reply:

I suppose it's possible to rid ourselves of anything we want to be rid of. 
And for you, Blake, newness is reason enough to want to be rid
of something. I sometimes point out that following tradition isn't always 
the best way to achieve progress :-)

Defeat support is peculiar to this list because the intention of
meeting criteria that measure for getting rid of the lesser-of-2-evils
problem is pretty much (but not entirely) peculiar to this list.
Yes, the IRVies claim that IRV gets rid of what they call "the
lesser-of-2-evils dynamic". But I'm talking about serious criteria-based
efforts to get rid of that problem.

Blake continues:

I expect that 20 years from now, people will still
be talking about Condorcet, but will have forgotten about
"defeat-support".

I reply:

I have no idea if that's so, but I question Blake's belief that
that claim of his has any relevance to the question of whether wv
is better than m.

I'm not doing this because I believe Condorcet(wv) will become
popular and get enacted. I'm doing this because I intend to do my
part, regardless of whether or not the work will succeed. I'm not
saying Condorcet(wv) won't succeed. Only that I don't know if it will,
and whether or not it will has nothing to do with my intention to do
my part.

But don't let Blake's personal prediction distract you from the
question of whether margins methods do as well by criteria as wv
methods. Margins method fail all the defensive strategy criteria.

I've asked Blake if he has other criteria by which he measures
compliance with the standard of getting rid of the lesser-of-2-evils
problem. And the standard of majority rule in multicandidate elections.
No reply. Maybe Blake isn't interested in those standards. Apparently
not, because margins methods fail those standards.

Blake continues:

That would be a good thing, because having so many
Condorcet completion methods weakens the cause in general.

I reply:

In other words, Blake is saying: "In order to get rid of all this
dissension, let's all just adopt my proposal."

I'm curious what kind of a "cause" is furthered by margins. Not
majority rule or getting rid of the lesser-of-2-evils problem.
Blake is sounding a lot like the IRVies here. His "the cause" apparently
is the replacement of existing voting systems with new ones, and he
doesn't want to talk about such divisive things as which of those
proposed replacements are any good, and which aren't.

Blake continues:

By voting
resoundingly against "defeat-support" we can help speed it to
obscurity.

I reply:

The issue of which way to measure defeat support actually would  be
considered obscure by most people. But Blake needs to give a better
reason to vote against defeat-support, if that's what he wants to
ask you to do.

Blake continues:

Of course, Mike claims that "defeat-support" is actually superior to
margins.  I expect him to give voluminous accounts of why this is.

I reply:

Where have you been. I've told why defeat-support gives better results
than margins. I've told it within the last few days, in this thread,
and I've told it numerous times on EM in the past.

Blake continues:

Please tell me if you find any of it convincing.  I can't possibly
respond to every argument Mike makes

You can't respond to any of them. Blake would like to portray this
as a situation where I've written so many different arguments for
defeat-support, and against margins, that there just isn't time to
reply to them all.

No, I've used two arguments. Defeat support methods comply with criteria
that measure for the standards of majority rule and getting rid of
the lesser-of-2-evils problem. And defeat-support doesn't overrule
as many voters as margins methods. That's 2 arguments.

I've repeatedly asked Blake how _he_ would prefer to measure for the
standards of getting rid of the lesser-of-2-evils problem and the
majority rule standard.

, but I suspect that I can respond
to those people find sensible.

I reply:

Are you going to post your replies to EM, or are you going to send
them by individual e-mail, so that they won't be refuted?

Blake continues:

Putting aside the issue of margins vs. "defeat-support", I think there
are good reasons to choose Ranked Pairs over the other Condorcet
contenders.  Here are some of the main ones.

1.  Simplicity.  The method is simpler and more straight-forward than
most of the other Condorcet completion methods.

I reply:

Your simple definition isn't complete. The justification for Ranked Pairs 
isn't nearly as obvious as that of SSD.

Yes, Ranked Pairs can be defined more briefly than can SSD (even
when defined completely), but people will wonder why we start by
dropping the _strongest_ defeat that's the weakest defeat in a cycle.

Blake continues:

2.  Quality.  The method is just good!

I reply:

Ranked Pairs doesn't do as well by social utility as does SSD.
At least when both methers measure defeats in the same way.
Ranked Pairs' lower SU may be the result of overruling more voters.

Admittedly margins methods have slightly better SU than wv methods,
when sincere voting is assumed, at least when (unrealisitcally) there's
no truncation.

Blake continues:

Some of these other methods
(like Mike's SD method), you basically have to say, I know this method
has problems, but I don't view them as too serious.

I reply:

SD's nonmonotonicity isn't a desirable feature, but IRV has a worse
nonmonotonicity problem, and, as I said, SD has the great advantage
of being the simplest Condorcet version that dominates IRV in terms
of criteria. I'd rather beat IRV with SD than lose to it with Ranked Pairs. 
Your simple Ranked Pairs definition isn't complete.

Blake continues:

You have to say,
there's a better method, but I don't think you're smart enough to
understand it.

I reply:

I don't know whom you've been talking to, but I don't tell someone
that they aren't smart enough to understand a complicated count rule--
_they_ tell _me_ when they don't understand it. SD's great simplicity
certainly counts in its favor.

Blake continues:

That's a real defect if you want to actually sell
people on the method.

I reply:

What, simplicity is a defect?

Blake continues:

As well, some of these methods can only be justified by quite peculiar
reasoning.  For example, last I read, Mike was claiming that Ranked
Pairs might find the best over-all ranking, but SSD finds the best
winner.  But what kind of a best ranking do you have if the best
winner isn't at the top?

I reply:

If I said that RP finds the best ranking while picking a less-than-best
winner, of course I spoke incorrectly. A ranking that is topped by
other than the best winner can't be the best ranking. But did I really
make that statement? Can you post the archived message in which I
said that?

What I said is that Ranked Pairs meets some special criteria that are
about consistency in the output ranking. The resulting output ranking
is _not_ the best ranking if you consider the winner important.
What I said was that if someone is less interested in the winner than
in those special output ranking consistency criteria, then they might
prefer Ranked Pairs. I wouldn't consider RP's output ranking the best
if it starts with a poor winner. But I just mean that, for someone
who considers those output ranking criteria more important, that
person might prefer RP.

Blake continues:

If we only wanted to rank half the
candidates, would we expect to use yet another method?  That's the
kind of reasoning that SSD is based on.

I reply:

Oh, is that what SSD arguments are based on? :-) Where did I say
that in my arguments for SSD? Rank as many or as few candidates as
you want. SSD is still more clearly motivated & justified than RP.
Can you show that RP no longer does worse than SSD in SU when
some short rankings are voted? In the case of Ranked-Pairs(margins),
I'd expect further worsening of SU when there's truncation, due
to that method's greater vulnerability to truncation.

In fact, truncation of course is what especially brings out the
failures of your margins measure (though this part of your posting
is about RP vs SSD rather than about wv vs m).

I don't think there's a great merit difference between SSD & RP
(though there's a big merit difference between defeat-support & margins).

But SSD is a better public proposal because of its more obvious motivation & 
justification.

Blake continues:

Do we want to have to sell
this off the list?

I reply:

You mean using a different method when we only rank half the candidates?
Has anyone but you suggested doing that?

Blake continues:

3.  Pedigree.  Of the Condorcet completion methods listed, only Ranked
Pairs has been written up in academic publications (PC is arguable).
Mike has nothing but contempt for academics, but at least economists
have some training in math.

I reply:

Would you like to explain SSD's faults that I'm unaware of due
to lack of mathematical training? I don't value the writing of most
academic authors on voting systems. I've told why. So now Blake is
appealing to authority--We should copy the academic publications.
But is there really a need for copying just to be copying?

Blake's comments above are a good example of his tendency to criticize
an individual as part of an effort to discredit a proposal that is
different from his proposal. I'm sorry that Blake wanted to drag
the level of these discussions down to his level in that way.

I don't find the merit difference between SSD and RP something to get
at all worked up about, much less something that could motivate
personal attack such as Blakes.

Blake, if there's anything wrong with SSD, Cloneproof SSD, or
the standards or criteria by which I compare voting systems, due to
a lack of mathematical training on my part, then what is it?
Can you show mathematically that Ranked-Pairs(m) is better than
SSD in terms of the standards and considerations that are important
to people?

Or is it just that if I had more mathematical training then like you
I'd copy the standards in the journals instead of advocating the
use of the standards that are important to voters?

If so, then your training seems to have only negative value.

This poll has had the unintended effect of bringing out the worst
in someone whose worst is none too good. I'm glad that Blake takes
the poll seriously. I'm sorry that as a result he displays the worst
of his character.

Blake continues:

The fact is that Ranked Pairs is better
established, and has been better studied than any of the other
Condorcet completion methods.  Some people don't see much difference
between Condorcet completion methods, and in this case, it makes sense
to go with a published method.

I reply:

Maybe it's you who don't see much difference between Condorcet
completion methods. Some people haven't heard about differences in
Condorcet completion methods. Some people haven't heard that there's
anything better than IRV, or better than Plurality. I hope you're not
going to use that as a reason to advocate Plurality, IRV, or a
criterion-failing Condorcet completion method.

Blake continues:

Of course, Ranked Pairs has a disadvantage on this list, because
unlike most other methods, Ranked Pairs' inventor isn't a list member,
and so won't be voting ;-)

I reply:

So now we're claiming that people who prefer other methods than
Blakes do so because they invented them?

I'm told that Markus invented Cloneproof SSD. My liking for that method
isn't affected by finding out that someone else may have "invented" it.

Blake continues:

This election is a real opportunity.  I hope that people take it upon
themselves to get informed.  If you have any questions, email me, or
post to the list.

I reply:

Blake, are you going to resort to off-list e-mail in order to
make your statements safe from rebuttal?

Blake quotes:

Ranked Pairs gives the ranking of the options that always reflects
the majority preference between any two options, except in order to
reflect majority preferences with greater margins.
(B. T. Zavist & T. Tideman, "Complete independence  of clones in the
ranked pairs rule", Social choice and welfare, vol 6, 167-173, 1989)

I reply:

If that were really the goal, then we'd say:

Drop the weakest defeat that's in a cycle. Repeat till there are
no cycles.

Instead, RP says (in its brief but complete wording):

Drop the strongest defeat that's the weakest defeat in a cycle. Repeat
till there are no cycles.

[end of definition]


Mike Ossipoff






_________________________________________________________________
Get your FREE download of MSN Explorer at http://explorer.msn.com