[EM] MJ has SFR, and more fraud-secure count than Score (but not as good as Approval)

[Michael Ossipoff]

SFR:

Because MJ has more than 2 rating positions, I expected that if it had
SFR it would be as in Score, nonprobabilistic. I tried that and it
didn't work.

When Jameson suggested that probabilistic SFR would work in MJ, he
emphasized that Approval's SFR metthod had to work in MJ, if you vote
at the extremes. Obviously that wasn't so, due to the different
count-method.

Anyway, when Jameson said that, I tried probabilistic SFR with MJ,
but, as Jameson suggested, I tried it by Approvals SFR method, and
also, for Approval's SFR goal. That didn't work, because MJ isn't
Approval.

Even though MJ has more than 2 rating positions, its SFR is probabilistic.

Say it's the Approval bad-example, and you're an A voter:

With some probability, greater than .5, chosen based on your estimate
of the faction-sizes (that probability might be unity), give B 0
rating instead of some good nonzero rating  (max? rating also chosen
based on the estimated faction-sizes, and maybe on how little you like
B).

With some probability (maybe unity. maybe same as or different from
abovementioned probability) , greater than .5, chosen based on your
estimate of the faction sizes, give C some small nonzero rating (maybe
just 1 increment over min, chosen based on your estimate of
faction-sizes, and on how little you like B).

If the A faction is large enough, and the B faction small enough, then
you'll give B a 0 majority, and therefore a 0 MJ score. If the A
faction is large enough, and the B faction small enough, you'll give C
a nonzero MJ score. Thereby, if the A faction is large enough, you'll
give B a lower MJ score than C.

If the B voters do the same, then C is sure to get a small nonzero
rating. Whichever faction is the larger one will be more able to get a
nonzero MJ score. Whichever faction is smaller will be more likely to
get a 0 MJ score.

I don't know the details. It doesn't seem as simple as Score/Approval
SFR.  Maybe, of MJ and Score/Approval, one of those gives more helpful
SFR than the other. I don't know. And, if so, I don't know which.

Count-Fraud Security:

As Jameson described it, yes, MJ seems more count-fraud-secure than
Score. But not as good as Approval in that regard.

And that fact accentuates Score's count labor. Because Score is
simpler and more easily-counted than the rank methods, there's a
tendency to consider it count-fraud-secure. But adding all those
numbers from 0 to 10, or 0 to 100 obviously provides much better
count-fraud opportunity than incrementing the candidates' counts when
they get an Approval.

And, as Jameson suggested, with A-F MJ, incrementing a candidate's 5
counts, for the ratings, is also more fraud-secure than Score's
additions.

Another lesson from this is that 0-100 Score is much a less-good idea
than 0-10 Score.

But Approval is very much better than Score (or MJ). And maybe that's
the most important conclusion from this.

Yes, Score (unlike Approval or MJ) gives easy nonprobabilistic SFR,
but making the count easy and fraud-secure is a lot more important
than making SFR easier--Is it really too much trouble to draw a
numbered piece of paper from a bag?

I'd previously considered Score to be more desirable than Approval
because of its easier SFR, and its better expressiveness.

But count-fraud-security is paramount.

Still, I've noticed that Score's better expressivity results in better
voting. The stark choice that Approval calls for can make people make
the wrong choice, and not really vote in their own best interest. If
people aren't very good at making that judgement, then it's much
better to let them give a fraction of max.

On the other hand, they can do that probabilistically in Approval. A
little less convenient, sure, but hardly prohibitively so. A little
more for the voter to do--What's wrong with that? What's wrong with
having to do a little more, be more involved with making, on your own,
the fraction of max that Score offers you readymade?

So, on balance, I much prefer Approval to Score, for the above reasons.

And MJ?

Comparing MJ to Score, .

They both have advantages. MJ wins in the fraud-security category.
Score wins in the elegant simplicity category.
Explaining about the median is more likely to lose your audience than
explaining about adding up scores. But what I really detest about MJ
is its need for elaborate bylaws. And that would likely be ruinous to
its public acceptance

I think Score's SFR is easier, because it needn't be probabilistic. In
fact, I feel that Approval's probabilistic SFR is easier than that of
MJ.

But MJ's count-fraud-security could easily be the deciding factor, in
choosing between Score and MJ, in spite of MJ's much more difficult
SFR.  ...assuming that you could get people to accept MJ's elaborate
bylaws, and its use of the median.

Bottom line:

Approval is by far better than Score or MJ.

Mike Ossipoff