[EM] Re: Strong FBC, strategy, thresholds

[Kevin Venzke]

Chris,

 --- Chris Benham <chrisbenham at bigpond.com> a écrit : > Kevin,
> Are you sure that the method  "Have an approval cutoff in the rankings. 
> If there is no CW, elect the approval winner" factors "ALL  of the 
> strategy away from the rankings or ratings of the candidates"? Might it 
> not be possible that a group of voters believes that its  favoured 
> candidate is not the CW but might be the Approval winner, and therfore  
> might be able to gain by insincerely order-reversing so as create a 
> circular tie ?

Yes, it's conceivable.  I imagine this kind of strategy would be pretty risky, 
though.  Also, I wonder if it could still be said that Strong FBC is satisfied.

I think Forest's method does away with strategy in the rankings by using it as
a tiebreaker for the "pair approval" ballot, which is guaranteed to produce a tie.
So all you have to do is come up with a second ballot which ties most of the
time, always results in just two candidates tying, and is clone-proof.

Harder than it sounds.  Perhaps you could do two-iteration approval, where the
approval winner is the first tie-participant, and then you go back and count only
the approval votes of those who didn't approve the winner, and thereby pick the
second participant.  Use the rankings to break the tie.

Doesn't sound like a promising system, but I do believe it would take the strategy
out of the rankings.  The rankings seem a bit superfluous here,
especially as the second approval tie participant isn't all that legitimate.

> You wrote: "I'm increasingly fond of the idea of having Condorcet where 
> no candidate may win who isn't  a first choice of 20% or so (arbitary 
> threshhold)."
> I dislike arbitary features in general and I hate arbitary thresholds 
> especially those not related to the number of candidates. What if you 
> have a  "first choice threshold" of 20% , and there are 6 or more 
> candidates and NONE of them makes the threshold ?

Of course those are valid complaints, but I think you're missing what I was
going for with this.  I see this as a hybrid of Condorcet and Approval, but
without the approval cutoff.   Because of the threshold, voting only your 
favorite in first place is akin to bullet-voting in an Approval election.
The idea is to convince voters to raise some compromises into first place.
Why?  Because the first rank is the only one we *know* has actual support.

Condorcet election:
40: ADBC
35: BDCA
25: CDBA

55 D>A, 65 D>B, 75 D>C.

D wins outright, but we have no idea whether D has any actual support.  All
we know is that if the electorate were able to vote like a legislature (what
Condorcet simulates), and had to elect *some* candidate, D would win.

The method I described above would elect A.  Perhaps a less arbitrary method of ruling
out D would be to count first-place preferences double.  This would cause both A and
B to defeat D (A>D 80, B>D 70, C>D only 50).

I suggested something like that awhile ago, except all not-last rankings were
to count double.  That also avoids cutoff insertion.

Kevin Venzke
stepjak at yahoo.fr


___________________________________________________________
Do You Yahoo!? -- Une adresse @yahoo.fr gratuite et en français !
Yahoo! Mail : http://fr.mail.yahoo.com