[EM] 0-info strategy clarification

[MIKE OSSIPOFF]

EM list--

I'd like to further clarify an answer that I made in my
most recent letter:

I'd previously said:

"If you don't [protect a certain preferred candidate by
sharing 1st place with him], for instance, one of them could
turn out to be a sincere CW who lost because of truncatiion, and because you 
didn't insincerely share 1st place with him."

Blake then said:

"But can you give an argument for why in a 0-knowledge situation,
falsely ranking 2 candidates as equal can maximize expected
utility, in margins? All you have done is suggested that in some
cases, equal ranking can help. But unless you know that
you are dealing with one of those cases, that isn't a good
reason to vote a particular way."

My answer was ok as far as it went, but it wasn't complete,
and left out what was most necessary to say in that reply:

Of course you don't know what situation you're dealing with.
It's a decision under uncertainty. Even with probabililty
information it would be a decision under uncertainty & we
wouldn't know what situation it is. But with 0-info, we
don't even have probability information. Either way, when
we don't know what situation it is, we calculate, based on
our specific probability information (of which we have none here)
the probability of each possible situation. One aspect of each
situation is what's the way in which we'd be able to affect
the result if there were a way in which we could. What pairwise
comparison is it that's the close one that would be affectable
by us if any of them is. Considering all the combinations of
pair-defeats, and, for each of those, the ways we could pick
a comparison to be the near-tied one (there are especially
unlikely to be two), we know that all of those possible situations
are equally likely.

Then, for each of those possible situations, we determine the
utility difference between the 2 outcomes that it's teetering
between, and which we could make go either way. With those
situation probabilities and utility differences, we can assign
utility expectation improvement values to each pair-ordering
we could vote, as compared to the utility of the outcome if we
didn't vote or if we voted the other way.

Having that, we could calculate the improvement of each possible
ranking of the candidates by us, compared to if we didn't vote.
So we pick the ranking, partial, complete, strict or position-
sharing, etc., that gives the most improvement over not voting,
in terms of its improvement of utility expectation.

***

I'm not saying that I've got all of that right, but it's got
to be something like that, doesn't it?

***

The point that I was trying to make was that of course we don't
know what the situation is, or that it's a situation where
we're saving a SCW from truncation. But we consider all the
situations that it might be, and how our utility would be
affected in each situation. So my argument doesn't depend on
our knowing that it's some particular kind of situation.

***

Mike Ossipoff

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