(Fwd) re: Mike's criticism of Bruce

Steve Eppley seppley at alumni.caltech.edu
Mon Feb 3 18:28:36 PST 1997


Mike inadvertently addressed this message to me, and asked me to 
forward it to the EM list.  --Steve

------- Forwarded Message Follows -------
Subject:       re: Mike's criticism of Bruce
To:            Steve Eppley <seppley at alumni.caltech.edu>
Date:          Wed, 29 Jan 97 4:53:11 PST
From:          Mike Ossipoff <dfb at bbs.cruzio.com>

Steve Eppley writes:
> Mike O. accused Bruce A. of dishonesty.  I don't recall Bruce 

But I don't really regard as an accusation, as much as a theory
to explain things that would otherwise be difficult to explain.

> making any serious factual errors, other than broken promises. 
> (In particular, the promise that he would defend Regular-Champion's
> result in any scenario... he has been silent on its "rich party" 
> failure and its "truncation" failure scenarios).
> 
> To the best of my recollection, Bruce hasn't engaged in Disorderly
> Discussion; he simply ceased claiming (in EM, at least) that
> Regular-Champion is a better method than Smith//Condorcet.  

But didn't Bruce recently, in a conversation or other communication
to Rob L., propose Fishburn//Copeland//Condorcet?. After everything
that we've said about the merit comparison of those methods,
after all the arguments regarding standards, principles &
criteria, & my statements about Smith//Condorcet's complete
domination of other methods, including the Copeland versions,
when it comes to criteria--after all that, Bruce, in a _private_
communication to Rob L., advocates Fishburn//Copeland//Condorcet,
a Copeland version that will share the problems of any Copeland
version.

And is the use of private communications with individuals 
a practical way to work on 1 person while avoiding open
discussion?

And Bruce's objection to LO2E-2, in which he said that
LO2E-2 can require that _every_ overlapping majority
must be able to simultaneously ensure the defeat of whomever
each of those majorities wants to defeat--that was an objection
that Bruce e-mailed to me in individual discussion before we
were on this list. I answered the objection. No reply from Bruce.
Then, later, Bruce recycles & re-uses the same objection, without
any acknowledgement that it has already been answered. These
are examples of disorderly discussion.

Sure, admittedly, Bruce doesn't do it as often as Don & Demorep
(who do it continuously). But it raises the question: Why?

> If Mike is aware, however, that Bruce has continued to advocate
> Regular-Champion in other forums without addressing the criticisms 
> posted in EM, I'd consider that a more significant allegation of 
> Disorderly Discussion, and with someone of Bruce's intellect
> an indication of possible intellectual dishonesty.

Not Regular-Champion, but another Copeland version, 
Fishburn//Copeland//Condorcet. Same thing. Using Fishburn
instead of Smith doesn't make a great difference in public
elections, and using Condorcet after Copeland doesn't help
except when Copeland returns a tie. So really, whether
the proposal is Regular-Champion or whether that proposal
is given a paint-job & made into Fishburn//Copeland//Condorcet,
it's effectively the same thing with the same problems.

So yes, we do have the kind of disorderly discussion you
mention, though the advocacy was in a private communication
with 1 member of the list, rather than open to the scrutiny
of the other list members.

> I appreciate a number of the things Bruce has contributed.  My
> favorite:  With a small modification of Arrow's "Independence
> from Irrelevant Alternatives" axiom (i.e., the members of the Smith
> set shall all be considered relevant), it's not impossible for
> voting methods to satisfy all the modified-Arrow criteria. 

Sure, but why does there need to be that intermediate meaning
for irrelevance? If a candidate is powerful enough so that when
he joins the race it results in the defeat of the otherwise-winner,
can that candidate be called irrelevant in a meaningful sense?
The irrelevant candidate, as I said, is if your Uncle Elmer,
a complete unheard-of, declared himself a Presidential candidate.
He wouldn't take away Clinton's victory.

So IIAC seems like something to throw out rather than
something to re-write--except for its relevance to
the No Spoilers Criterion, and, if I understood you
correctly, its relevance to the Gibbard-Satterthwaite
theorem. But those things involve IIAC in its original
form, not a re-written form.

But Bruce of course has made useful contributions. For instance,
when he pointed out that the Independence from Irrelevant Alternatives
Criterion is equivalent to the No Spoilers Criterion. And of
course when he told me about the Schwartz set. These things
don't bear on my theory about motives, but I don't deny the
contributions.

By the way, speaking of Arrow, if Arrow showed that we can't
have IIAC, Pareto, & No-Dictator, then might we be able
to get IIAC if we abandoned Pareto? Do we need Pareto?
If Pareto is violated only under certain hopeless conditions,
then losing Pareto wouldn't mean losing GMC. I have no idea
how big a violation of Pareto would be needed in order to
get IIAC. Maybe it would be of a kind that _would_ necessarily
lose GMC too.

Just a suggestion of a possible approach to a more ideal method
(if one doesn't mind losing Pareto & if other important criteria
aren't lost with it).

***

For the benefit of anyone who hasn't been on this list a
long time, let me define IIAC & Pareto:

IIAC: Removing a non-winning candidate from an election
shouldn't change who wins, when the same rankings are
re-counted.

Pareto: Never elect someone over whom a particular other
candidate is ranked by everyone.

***

Achieving IIAC would achieve a kind of idealness that we've
sought. Losing Pareto could be unimportant if it only
happened in situations that wouldn't affect GMC.

I mean, if, with a 100,000,000 voters, does it really
matter whether 100,000,000 or 100,000,000 - 1 of them
vote A over B? Isn't B pretty soundly beaten either way?

GMC: Never elect a majority-rejected candidate (a candidate
over whom someone else is ranked by a majority) unless
every candidate in the set from the method is to choose
is majority-rejected.

Mike



More information about the Election-Methods mailing list