[EM] SFR as f(Ub, p(config)). Sincere rating.

Michael Ossipoff email9648742 at gmail.com
Sat Sep 8 21:15:03 PDT 2012

It seems to me that the Chicken Dilemma must involve a candidate whom
you'd approve or give max rating to if there weren't a Chicken
Dilemma. ...if you didn't care if you were taken advantage of by the
other faction.

So the giving of SFR, it seems to me, depends on the candidate
qualifying for an approval.

That would justify the inclination for the giving of SFR, or the
amount of it, to be influenced by how much you like or dislike B.

If the information for such a calculation (or even a guess for it)
were available, then, the choice about SFR could be based on two

1. The probability that B qualifies for Approval, disregarding the
Chicken Dilemma.

2. The probabilities of the various values possessed by the various
quantities, in each of the various scenarios (each with its own SFR
formula) regarding what kind of faction-size information is available,
along with the probability of each kind of information-scenario.

I don't think that that's even guessable. Maybe it's just a matter of
subjective, intuitive feel. What I've occasionally done was to pick an
information-scenario, and estimates for its numbers, choosing the
scenario and corresponding formula that gives the SFR that intuitively
agrees with how I like the candidate. So, really, I was just giving
the candidate what I felt like giving hir, and justifying it with a
formula chosen for that purpose. Not really objective.

But the fact that the merit of the candidate justifiably influences
the matter of how much SFR is awarded, justifies something that a
Score advocate said at a website. He said that sincere rating can,
&/or often does, approximate good strategy. When I read that, I agreed
with it, because that has been my impression too.

If B has more points than A, and is the one who can beat C, if anyone
can, then give B a little boost, enough to close the gap between C,
and a B who has so many votes that he's overtaken A.   But not enough
to be likely to bring B up to more than A, if B is less popular. Of
course it might do that, but that's less likely if the SFR boost is
small, requiring a small gap between B and A.

But, as I said, that's a subjective judgement, and so how much you
like or dislike B will influence the amount of that boost, for the
reason I spoke of above.

One might as well just guess, award by feel, since the information
needed for a calculation would be just guesses anyway.

Giving to a candidate what you feel you should give hir can be called
sincere, even if you're doing for strategic purpose. In fact, I don't
really know what other meaning "sincere ratings" would have. When a
choice is that subjective, can you really say what is influencing it?
I don't think you can say that you're giving ratings that are sincere
in the sense that you know the candidates' utility for you, and be
sure that you've separated the considerations and are awarding based
entirely in proportion to utilities. Utilities are defined in terms of
probabilities--and a question about which of two lotteries you'd
choose. But, in a voting situation, knowing what you're inclined to
do, and knowing why, are two different things.

So, the statement about sincere rating maybe approximating good
strategy makes sense.

When it's said that strategic Score voting is at the extremes, that
comes with a condition: ...that there isn't any communication among
voters. But declaring how much you're going to give to a compromise
candidate whose voters regard your candidate as a rival is a form of
communication, and, because that is possible, strategic voting is no
longer only at the extremes.

Many of your ratings _will_ be extreme. SFR would just be for an
otherwise approval-qualifying candidate with whose voters there is a
Chicken Dilemma. ...Or sometimes when you don't know if a candidate
qualifies for approval, as you sometimes don't.

I used to say that Score has the disadvantage that strategic voters
will take advantage of sincere voters, as if sincere and strategic
voting were distinct and opposite. But the claim that I was making
wasn't correct, if sincere rating is really intuitive SFR, or at least
roughly approximates it, as that person at a website said, and as has
been my impression too.

I gave SFR to MJ in the minguo voting-systems poll, but rated all the
other voting-systems at the extremes. That poll uses 0-10 Score.

The inbetween rating for MJ was because, unlike unimproved Condorcet,
IRV and Plurality, MJ meets FBC. And, MJ doesn't have the
computation-intensive, fraud-vulnerable count that rank methods have.
 ...But MJ's tiebreaking bylaws make it a poor proposal, and its SFR
is more complicated and difficult than that of Approval and Score. For
those reasons, MJ seemed genuinely inbetween in merit, not obviously
approvable or unapprovable, not obviously in an acceptable or
unacceptable set. And, because preferrers of MJ could strategically
0-rate Approval and Score, I didn't want to co-operate with that, but
didn't want to forgo MJ as a possible alternative to the much worse
unimproved Condorcet or IRV..

If you equal top-rate a compromise in a poll, and then notice that the
compromise has acquired more points than your favorite, then maybe you
have been had. Someone else wasn't as co-operative as you were. It
happens often in polls. Just because you give their candidate a good
rating, doesn't mean they give one to your candidate.

Mike Ossipoff

More information about the Election-Methods mailing list