[EM] SciAm article

[Alex Small]

Well, I read their article last night.  They have some condition called
"neutrality" which requires two things:

1)  That a method not favor any candidate over another in its rules. 
(Pretty innocuous)
2)  (here I'll quote them precisely)  "The second requires that the
voters' choice between candidates A and B should not depend on their views
about some third candidate C."

>From what comes next, they seem to be using "neutrality" as a synonym for
IIAC.  However, they say that "true majority rule" always satisfies
neutrality.  We know that Condorcet (which they call "true majority rule")
flunks IIAC.  Maybe they mean that Condorcet only flunks IIA when a cycle
occurs.  Anyway, it would be nice if they had given a slightly better
definition of this criterion.  I don't expect a detailed and rigorous
technical definition in a Scientific American article.  But surely they
could have done a better job.

Their arguments for Borda as a cycle resolution method are pretty weak. 
Basically, they define the criterion of transitivity, which means there
shouldn't be any cycles.  They then point out that Borda satisfies it, and
Condorcet doesn't.  Um, yes, if we don't define a cycle resolution method
for Condorcet.  But a lot of other methods also pass transitivity.

I guess that if you only limit the discussion to Condorcet and Borda, then
Borda as a completion method makes perfect sense.  Then again, if we only
limit the discussion to Condorcet and ANY method, that other method will
make perfect sense as a resolution method.

Overall I think it was a pretty weak article.  OK, it started out good, in
outlining the basic problems of election methods.  But their
recommendations were pitched in a really lousy manner.



Alex