[EM] Three questions about Majority Judgment

[Michael Ossipoff]

2012/9/30 Michael Ossipoff <email9648742 at gmail.com>

 Jameson--

My three questions are asked below. But first:

I'd said:

First, thanks for putting in some good words for Approval, at EM yesterday.

But I note that, in your message at or to wikipedia, as part of your
 proposal of MJ there, you referred to Approval and Score as "inferior
 methods".

[endquote]

You replied:


No, I didn't. I said "inferior alternatives". Implicitly, that means
"... for this particular use case" (phrasing which I've now added
explicitly). And yes, I do believe that MJ is best for the particular
use case I was talking about there.


In fact, earlier in the piece, I explained that: "Please note that
this is not a one-size-fits-all solution. While Majority Judgment is a
good system overall, there are situations where I'd recommend others
even more highly. For US president, I'd recommend Approval; for US
Senate, SODA Voting; for most congressional and parliamentary systems,
a biproportional system such as PAL voting; in Robert's Rules
situations, approval with runoffs; and in loose internet voting, Score
Voting. Such flexibility is the spirit of the Declaration of Election
Method Reform Advocates, of which I am a signator."

[endquote]

Oops! Acknowledged.

But when you say that MJ is a better choice for wikipedia than Score
or Approval would be, shouldn't you support your claim more
specifically than you did? I mean, given MJ's much lmore elaborate
definition, you've got to offer something in return.

You said the old MJ saying, about how MJ prevents extreme-rating
stratgegizers from having too much effect. That's too vague.

Here are a few explanations of the meanings of expressions that I'll use:

When I say "you", I mean "You, and your same-voting faction"

When I speak of moving a candidate's median  "by 1 voter's worth", I
mean moving it a distance from hir previous median, on the
ratings-scale, to at or above the next voter's rating of hir.

"1 voters' worth per 2 persons" means "1 voter's worth for each 2
voters in your same-voting faction"

When I speak of "previous median", I mean the candidate's median just
before you (that means "you and your same-voting faction) cast the
last votes in the election.

My two questions:

1. Given that, in MJ and in Score, a voter can maximally help a
candidate by top-rating hir, and can maximally help one candidate over
another by oppositely extreme-rating them,

...and given that, in MJ that means that you're thereby raising that
top-rated candidate's median 1 voter's worth per 2 persons, or helping
Better against Worse by moving their medians respectively up and down
by 1 voter's worth per 2 persons:

...and given that giving an inbetween rating to a candidate would move
hir median just as much, in an unknown direction:

What would be the meaning and value of an inbetween rating? What
desirable effect would it have? When would it be a good idea? How
would you choose that rating?

Of course what it will do is: It will move that candidate's median 1
voter's worth per 2 persons, in the direction of your rating.

So, if you rate "sincerely", in proportion to your utility-estimates
of the candidates, then you're trying to move all of them to rating
positions corresponding to your estimated utilities.  Why would you
want to try to do that?

2. Because, no matter what some people like to assume, there's no
particular reason (unless you can give one) for rating candidates
proportional to their estimated utilities to you, then what should
motivate and guide inbetween ratings?

I suspect that, without a chicken dilemma,  MJ's strategy is identical
to that of Score. The same extreme-rating strategy as Score has.

The difference between the two methods' strategies would be in the
strategic fractional ratings (SFR), when there's a chicken dilemma.

Ok, let's look at SFR in MJ. When we discussed this a week or a few
weeks ago, I described an SFR that would work in MJ. It was rather
elaborate, and it was probabilistic.

Can you describe (specifically, in detail) a different, easier, SFR for MJ?

If you wanted to try non-probabilistic SFR in MJ, you want to raise
the median of Compromise. You want to raise it to above the median of
Worst. Obviously, to do that, you have to rate Compromise above
Worst's median. But the idea of SFR is that you don't want to help
Compromise beat Worst if Favorite is the one of {Favorite, Compromise}
who is more winnable. Now, you're rating Compromise higher than
Worst's median, because that's the only thing that can raise hir
median to above Worst's median.

But what if Favorite's previous median is higher than that of
Compromise, so that Favorite, among {Favorite, Compromise}, is the
rightful winner? The Compromise voters defect and bottom-rate Favorite
(as you expected them to). If you have to rate Compromise over Worst's
median, to help Compromise against Worst at all, then it would seem
that, if Favorite is the more winnable of {Favorite, Compromise}. and
therefore the one of {Favorite, Compromise} whom the Favorite and
Compromise voters should help beat Worst, then  you're being had if
you do so.

3. So when and how should inbetween ratings be used in MJ? How does
MJ's results with inbetween ratings in general, or SFR in particular,
make it better than Score?

Mike Ossipoff

















Then my question is: