[EM] Of course non-extreme rating is only for two particular situations

Michael Ossipoff email9648742 at gmail.com
Mon Sep 10 13:40:08 PDT 2012

When I said that sincere rating is really subjective intuitive SFR,
that statement needs a little qualification.

For one thing, subjective intuitive non-middle rating could, instead
or also, be the result of uncertainty about whether or not a candidate
is qualified for approval (even if there weren't or isn't a Chicken
Dilemma (CD)).

The only reasons for not rating a candidate at an extreme, max or min, are:

1. There isn't a Chicken Dilemma because:
 .....a) There isn't a Chicken Dilemma if B isn't an
approval-qualifying candidate.
......b) The B voters genuinely don't prefer A to C.

(I'm not saying that there couldn't be more reasons for there not
being a Chicken Dilemma)

2. It isn't certain whether or not the candidate in question (called
"B") would qualify for approval, in the Approval voting system. Maybe
you don't have enough perception of B's (mayble marginal) merit, or
the winnabilities, outcome-expectation, or pair-tie probabilities, to
judge whether B should be approved. In that instance, of course you
might give B a score consistent with the probability that s/he should
be approved. Or, if the method is Approval, you might approve B with
the corresponding probability.

Of course one reason why you might not know if B qualifies for
Approval could be that it's a u/a election, and it isn't obvious
whether B is among the acceptables or the unacceptables. Maybe the
unacceptables are a lot worse than B, but B is a lot worse than the
acceptables. By various different considerations. So, though you know
of an unacceptable set, and an acceptable set, you don't know in which
one B should be classified. Give hir fractional rating or (in
Approval) probabilistic approval.

[end of list]

Of course, if you give a non-extreme rating when there isn't a reason
to, then you're being exploitably sincere. So I shouldn't have implied
that sincerity can never be exploited.

People probably have a tendency to give non-extreme ratings even when
it isn't optimal to do so. Do that and you can be had.

But, when there's reason to rate non-extreme, I still say that natural
subjective intuitive non-extreme rating, sincere non-extreme rating,
is likely a fair approximation to good SFR, or (when there isn't a
Chicken Dilemma) approximates well the probability that the candidate
qualifies for approval.

My non-extreme (3/10) rating of MJ is justified mostly by uncertainty
about whether MJ qualifies for approval, because the 0-rated
alternatives are so much worse than MJ, and MJ is so much worse than
Score, Approval and Emocracy (a Score version).

Additionally, for there to be a Chicken Dilemma, first MJ would have
to qualify for approval; and, second, the MJ preferrers would have to
prefer Score and Approval to my 0-rated alternatives, and
strategically 0-rate Score and Approval. That's a lot to suppose.
There's nothing wrong with guessing that they might do that, but I
don't even know if they don't prefer Score and Approval to unimproved
Condorcet and IRV. Anyway, so the existence of a Chicken Dilemma is
far from certain. But one thing more for sure is the uncertainty about
whether MJ qualifies for approval.

Mike Ossipoff

More information about the Election-Methods mailing list