Sincerity, strategy

Mike Ositoff ntk at
Mon Oct 12 19:13:40 PDT 1998

The other main standard that people expressed was sincerity.
A method should encourage sincerity. Then surely it shouldn't
force insincerity.

IRO advocates claim that their method lets you rank sincerely
because your lower choices can't hurt your favorite. But that
can be said of Plurality too. So far so good; does that mean
that IRO & Plurality are better? No, because if you try to
protect your 2nd choice from defeat because you need it to beat
something worse, then that _can & often will_ defeat your
favorite, when you misjudge & help that 2nd choice when you
didn't mean to. That's really the whole strategy problem,
isn't it? In Plurality that happens when you vote Clinton ovre
your favorite by voting for Clinton instead of your favorite.
In IRO that happens when you insincerely vote Clinton in 1st
place so he won't be immediately eliminated.

Blake's sincerity standard is about indifferent voters, when
they know nothing about other voters. Are indifferent voters
really important? I've discussed several reasons why that
standard doesn't seem important, that interpretation of the
sincerity standard:

1. The only CW who could be defeated by insinsincere extension
would be a rather weak kind of CW whose election depended on
lots of indifference. It would be a circular tie, except taht
lots of people didn't give a damn. 

2. Since anyone who might be CW must be considered a serious
contender, the defeat of a CW by insincere extension, since
it requires indifference about him, requires lots of voters
who are indifferent between major contenders, or between a
serious contender, the CW, and less major candidates. How likely?
How important a problem if it happens? Defeat of a rather
indifferent CW?

3. That indifference, on the scale that would be needed, suggests
that the CW and the other candidate are virtually the same
in the opinion of lots of people. How big a problem if one
wins instead of the other, if they're that similar?

4. There won't be indifference, strictly speaking, by any
spatial interpretation of an election. I've talked about how
there won't be indifference with a 1-dimensional policy space.
But what if there are several dimensions? How likely is it that,
for many voters, 2 candidates will be exactly equally distant
to each one of those voters? Forget it.

So what we're really talking about is what really should have
been a circular tie anyway, with no genuine CW, just someone
who'd be BeatsAll winner if enough people truncate between
roughly equidistant candidates--but then decide not to truncate
after all. Again, big problem?


More information about the Election-Methods mailing list