[EM] My definition works fine, but ideally needs work.

[MIKE OSSIPOFF]

Yesterday Mr. Harper & Mr. Layton correctly pointed out that
my sincerity definition unfairly calls someone insincere if he
considers X & Y equal, but is forced to vote them unequal by
ordinary IRV or by Plurality.

I agreed then and agree now that that's true, because 1) he's
forced to; & 2) If he considers them equal then he really doesn't
care if he helps one against the other.

But it's also true that my definition, in its present uncorrected
form, works fine with criteria such as SFC, GSFC, and my
version of Condorcet's Criterion, etc.

That's because all it does is call someone insincere whe he isn't.
By so doing, it makes the example fail the criterion's premise, so
that the example can't be used as a failure example. No problem:
I write a failure example in which the people whose sincerity is
stipulated in the premise don't consider any candidates equal.
All of my failure examples are like that, in fact.

So, though it's true that something along the lines of Craig L.'s
sincerity definition would likely be the best way to fairly
define sincerity, and in fact Craig's definition itself may be
the best definition, it's also true that my less perfect definition
works just fine with criteria such as SFC, GSFC, & my Condorcet
Criterion version.

I'll likely be rewriting my sincerity definition before long,
maybe adopting an already-written definition. But when I said that
my criteria might be undefined, or that I might not want to apply
them to all methods, the answer to that "maybe" is: No, that isn't
so. The definition works fine with the criteria. The criteria still
apply to all methods, without having special rules for various
methods and without defining methods contrary to how they're
defined & used in actual implementations.

Mike Ossipoff
_____________________________________________________________________________________
Get more from the Web.  FREE MSN Explorer download : http://explorer.msn.com