[EM] Discover Magazine article

[MIKE OSSIPOFF]

On Thu, 19 Oct 2000, Bart Ingles wrote:
 > My thought while reading it was, "Gee, I'm glad to see someone found a
 > legitimate use for Borda!"

Rob L. replied:

I think that was the most interesting part.  I wonder how the various
Condorcet methods would hold up in handwriting recognition, etc.  It may
be interesting to track down these organizations and see if they've tried
it.  If Borda does better, it may provide some evidence that "symmetry" is
an important criterion.

I reply:

I expect that, where strategy & LO2E aren't considerations, as
when the voters are Utopian, or when the "voters" aren't human,
then Borda is the ideal rank-count.

It's the one to use when you know how several alternatives rank
according to several different considerations. Each consideration
is considered a "voter". If you can assign numerical ratings to
the alternatives, for each consideration, that's even better. If
not, then use rankings & count them by Borda.


An example: Steve said that, when Tideman(wv) and BeatpathWinner
give different results, the Tideman(wv) winner pairbeats the
BeatpathWinner winner in the vast majority of cases. So if the
people could choose between those winners, they'd choose the
Tideman winner, and so the use of Tideman is more democratic.

Markus replied that there could be a cycle of such relationships
among methods, and that that invalidates the idea.

But that's an IRVie argument against Condorcet. The same solution
can be used. If you have a whole arrow-graph diagram of methods
that beat eachother in terms of winner comparisons (whose winner
beats whose winner what percent of the time), then you could use
Condorcet. But Condorcet is mostly for getting rid of stratgy
problems, LO2E. As I said, when "voters" aren't human or alive,
there's probably no need to use other than Borda. Borda can be used when
all we have is pairwise defeats & their magnitudes. In this
case, method M's Borda score is determined by determining, for each
other method, the percentage of the cases where its winner pairbeats
that other method's winner. That percentage is summed over all the
other methods.

I mention that just as an example of a no-humans count where
Borda would likely be the thing to use. Strategy-freeness & majority rule
probably have little meaning in a no-homans count.

Of course, with Tideman & BeatpathWinner, we don't need that, because
it's just 2 methods we're considering as the best. And even if there
were more (limited to those for which there isn't a nonmonotonicity
example), Tideman would probably be "Condorcet winner" anyway, so
we wouldn't need a circular tiebreaker. Or would we want
to award the win to the "Condorcet winner" or the Borda winner, when
they differ? Maybe, in this no-humans count, we'd choose the
Borda winner.

Mike Ossipoff



_________________________________________________________________________
Get Your Private, Free E-mail from MSN Hotmail at http://www.hotmail.com.

Share information about yourself, create your own public profile at 
http://profiles.msn.com.