[EM] If B>2/3, vote A=B>C

[MIKE OSSIPOFF]

EM list:

Now I've redone the problem & checked carefully, and, with
0-info, with Condorcet, one should vote B equal to A, in
1st place, if B is greater than 2/3.

With Margins it's always advantageous to express the AB & BC
preferences, with 0-info.

All that's assuming that we ignore the possibility that, even
with 0-info, strategies could be influenced by how others are
likely to vote, even though we don't know exactly how many people
have particular preferences.

***

Blake brought up that 0-info incentive as a criticism of Condorcet,
but if someone wants to talk about a method's utility-maximizing
strategy, and only talks about it under one particular set of
conditions, that seem a little less than honest.

Any criticisms based on utility-maximizing strategy should
cover more than just one kind of conditions. Not just 0-info,
but also conditions where voters believe that they know
the other voters' preferences utilities, or conditions where
they have probability information about how many voters have
what preferences & utilities. For instance someone could say
that A supporters have certain utility ratings for the other
2 candidates, and likewise for B & C voters, and that the voters
have estimates about the numbers of A, B, & C voters that have
, in the voters' opinion, some degree of uncertainty, or maybe
none.

No doubt that would be the kind of problem of interacting
strategies that game theory covers.

Anyway, a discussion of utility-maximizing strategies isn't
complete when only 0-info is considered, and any attempt at
criticism based on that incomplete discussion has a rather
half-assed nature.

I'm not saying it would be easy to study & describe strategy
under all those conditions, only that one should do so before
picking the easiest part to discuss, and using that by itself.

We've heard the argument that 0-info has a smooth gradation into
probability-info, which grades into putative certainty. Sure,
but how people vote with probability info, or, as they believe it,
near certain info, in our elections here, is quite different
from how they'd vote with 0-info. It does make a difference which
conditions obtain, and we don't have 0-info. Nader voters who
believe Nader won't win vote for Gore, for the most part. If
they didn't have that winnability estimate, they'd vote for Nader.
So not only is an only-0-info discussion incomplete, but it's
very much inapplicable to existing political elections.

Mike Ossipoff

______________________________________________________
Get Your Private, Free Email at http://www.hotmail.com