iterated Condorcet checks out

Mike Ossipoff dfb at bbs.cruzio.com
Mon Aug 12 22:50:50 PDT 1996


Steve Eppley writes:
> 
> Mike O wrote:
> -snip-
> >Of course, on the other hand, Iterated Condorcet isn't appreciably
> >more complicated. Its iteration rule isn't complicated. So I'm
> >certainly not saying that I oppose Smith//Iterated-Condorcet as a
> >public proposal. 
> 
> I think I have a long way to go before I can be convinced that the
> Iterated Condorcet idea is sound.  It may behave too much like 
> Approval and ignore too much of the electorate's real preferences.

Well that's why it should be a flexible ballot option instead of
an automatic thing. It's somewhat like Bucklin, and in fact my
search for the method's possible problems was based on that fact.

It's true that with any new method, it's always best to be cautious,
because sometimes there later turns out to be problems that take away
the advantages. But, as I was saying, the more difficult a
strategic vulnerability is to find & use, the less of a problem
it would be in actual elections.

But it needn't behave like Approval except where the voter
chose to use the method's special option. Even then, it's
different from Bucklin (the rank-balloting stepwise version
of Approval), because where Bucklin _automatically_ has each
of us give a vote to our next choice till someone gets a majority,
Iterated-Condorcet, even if we choose to use the stepwise
1st choice promotion option, only gives 1st choice status to
our next choice if someone worse will otherwise win.

While realizing the value of caution, I still say that
Iterated-Condorcet seems likely to thwart order-reversal
& thereby eliminate need for defensive strategy,  in a
deviate electorate, at least for all practical purposes.

Anyway, I'm not claiming that methods with refinements for
devious electorates should be candidates for the 1st 
method recommendation to ER. It seems reasonable to just report
to ER about the vote that we completed, and then send to ER
individual & consensus conclusions about the proposed methods.

Sure someone could discuss Iterated Condorcet, &/or other
similarly iterated methods, in their individual report on
the topic, but they could also choose to send that afterwards
instead--after the relatively simple comparison of the more
familiar proposed methods. That's how I'll do it myself:
Anything I send about devious electorate refinements would
be sent after the completion of the current project to
compare the proposed methods for ER.

Refinements like that iteration are a relatively new topic, and
I don't object to holding off on posting to ER about it, at
least for the immediate time-being.

> 
> Would Smith// serve any purpose, though?  Wouldn't the iterations 
> eliminate the candidates not in Smith, anyway?

I can't say for sure at this time, but my intuitive impression is
that compliance with the Smith Criterion is a separate thing,
and would still require the use of the Smith set. But I'm not
even sure enough about that to call it a "conjecture". These
rankings modified during the iterations, to rank candidates equally
, might not always change the fact that a candidate not in the Smith
set might have the fewest voters ranking some same candidate over
him.

> 
> ---Steve     (Steve Eppley    seppley at alumni.caltech.edu)
> .-

p.s. By the way, as Steve suggested in his letter that introduced
Iterative-Condorcet, a whole family of methods is implied, based
on adding that iteration to various methods. Possibilities to
check out could include iterated versions of Plurality, MPV,
Bucklin, etc. Those are the 1st ones that I'd check out.

Of course (at least it seems to me) MPV, for that purpose,
would have to be the "IR-1" version, where equally 1st-ranked
alternatives all get a whole vote from the ballot which they
all share the top position on. Similarly, an Iterated-Plurality
would have to treat those equal 1st choices like Approval.

I have no idea whether iterated versions of Plurality, MPV
or Bucklin would be good, but they're worth checking out.
But IR-1 will encounter some resistance, due to a misperception,
even among some electoral reform organizataion leaders, that
it violates "1-person-1-vote" in some meaningful or important
sense.

Still, if Plurality, MPV or Bucklin could be made into 1st
rate methods by iteration, then methods already liked
by many people who aren't into sw reform would become good
methods. Well, I've never known MPV advocates to accept
improvements in MPV. It's too early to predict that
iteration can make any of those methods 1st rate (maybe
you already know a reason why it wouldn't), and I'm
just agreeing that these possibilities seem worth checking
out.


Mike



 



> 


-- 



More information about the Election-Methods mailing list