[EM] 3-slot Condorcet//Top Ratings

[Jameson Quinn]

2016-09-24 21:08 GMT-04:00 C.Benham <cbenham at adam.com.au>:

> Jameson,
>
> You claim  to be concerned about the "chicken dilemma" scenario and you've
> typed that you think a method
> should "tend to respect Majority Condorcet". while making it plain that
> you aren't ready to die in a ditch for
> FBC compliance (and presumably that also applies to Later-no-Help).
>
> Also you say that practical reform proposals should be able to be easily
> explained and justified to the non-expert
> "man in the street" (although the method you currently advocate, U/P, only
> just  fills that bill if you omit some
> "fine print" that contains some complicated arbitrary rules about default
> ratings).
>

The current version of U/P does not have that complexity. Here is the full
statement:

For each candidate, you can downvote (rate "unacceptable"); upvote (rate
"preferred"); or neither. All candidates downvoted by a majority of those
voting in the election, or with fewer than half the upvotes of the
most-upvoted candidate, are eliminated, unless that would eliminate
everyone. Of those remaining, the most-upvoted wins.


>
> C: As Mike Ossipoff recently explained, U/P   (along with IBBIFA, MTA,
> MCA, MJ)  fails the  Chicken Dilemma criterion.  And U/P fails the
> Condorcet criterion.
>

Hmm... I just realized that Mike's messages have been inexplicably going
into my spam folder for some time now (less than a year, more than a
month). I'm really sorry about that! It was never my intention! I've now
gone back and read most of the recent messages and there's plenty of good
stuff there.

I do not believe that the CD criterion is the best measure of which methods
deal well with CD situations. The CD criterion requires a method to be
punitive, meaning that it will react very badly (elect the Condorcet loser)
in certain center-squeeze scenarios:

35: A>B
25: B (honest is B>A)
40: C>B

CD criterion says that if A wins under honest voting, then B, who is the
honest and voted Condorcet winner, cannot win under strategic voting. I
recognize that A winning under honest voting is probably a failure of a
voting system, but it is not a failure that makes a system particularly bad
at chicken dilemmas. And B winning under strategic voting ameliorates,
rather than compounding, that failure.

What makes me think that U/P is a good system for dealing with CD, even
though it doesn't pass the CD? Basically, the idea is that chicken strategy
never puts ally1 below ally2 without putting ally1 below enemy; thus, it is
never successful without being dangerous. That means that, in order for it
to successfully change the election in a pure CD scenario (one where the C
voters have no A/B preference), the smaller B group must hope that they can
mobilize a supermajority of strategic B voters while keeping a
supermajority of A voters honest. I think that that's much less feasible
than if you dropped either of the "super"s.


>
> So as something that better fits your stated aims, I suggest simply
> 3-slot Condorcet//Top Ratings:
>
> *Voters give the candidates one of 3 ratings (say Top, Middle, Bottom).
> Default rating is bottom-most.
> Inferring ranking from these ratings, any candidate that pairwise beats
> all the others wins.
> Otherwise the candidate with highest Top Ratings score wins.*
>
> Smith//Top Ratings would be technically a bit better, but  the "Smith
> set"  part  would probably make the method harder to explain.
>
> I think that this is quite a good system.

The reason that I did not include any "pairwise" in U/P's definition is
that I think that it's tough to present pairwise results in a simple way.
Basically imagine a situation where A won, and somebody asks why didn't B
win. You don't want to have to mention any candidate but A and B in the
answer.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.electorama.com/pipermail/election-methods-electorama.com/attachments/20160925/ad189856/attachment.htm>