[EM] Poll conditionality options. Conditionality approaches. Sequential nested conditionality.

[MIKE OSSIPOFF]

First, let me say that, probably, GMAT or MMT conditionality could be used with an Approval options
system such as the one that I propose. 

I chose MTAOC conditionality because of its straighforwardness, and because it's
easy for it to be optional, even by candidate.

Poll conditionality options:

1.Instead of marking middle ratings to particular candidates "(conditional)", one could,
instead, just annotate one's ballot "(all middle ratings conditional)".

2. I spoke of the following options for designation of coalition-suitable candidates:

1. All above-AERLO candidates are coalition-suitable. If the ballot doesn't use AERLO,
then all of its top-rated or top-ranked candidates are coalition-suitable. I've suggested
that that be the default assumption, when the voter doesn't address the matter, and I
still suggest that (unless others disagree).

2. Only the ballot's initially top-rated candidates are coalition-suitable.

3. The voter could individually designate any particular candidate(s) as coalition-suitable.

I'd like to suggest an additional option:

4. The ballot's _current_top-rated or top-ranked candidates are coalition-suitable. Of course
a ballot's current top-rated candidates can differ before and after the application of AERLO.

I still suggest that #1 is the best default assumption, because any one of a person's 
AERLO protected candidates would surely be a good compromise to hir. That's why
those candidates were given AERLO protection.

The following isn't intended to apply to the poll that I propose.

Conditionality approaches:

1. Sequential nested conditionality:

This is just a tentative comment, but maybe a method could use nested conditionality, 
with a rank-ballot, instead of Bucklin's stepwise vote assignments and majority-requirement.

Starting with the last rank level as the "outer candidates", and based on (say)
MTAOC conditionality, counting your inner candidates (those ranked over your
outer candidates) as coalition-acceptable,
determine whether you withdraw your middle-rating to the outer candidates.
Then include the next higher rank of candidates among the outer candidates,
and make the same determination regarding your middle ratings to them.

With this being done simultaneously for all the ballots, I don't know if that simultaneous
application of this process would cause a prohibitive problem. Maybe not. If not,
this method could be a substitute for AOCBucklin.

If it works, I don't know what its merits would be, as compared to AOCBucklin.

Maybe it could work as an Approval election option, along with Approval, the
voting methods of AOC, MTA, MCA,  MTAOC and MCAOC.

2. Conditionality by top-2:

I don't know if this would work. 

It would be nice if it were possible, for conditionality, for the voter to indicate that
s/he wants to drop hir middle rating to a candidate if that candidate, and one of hir
top rated or AERLO-protected candidates are the top-two votegetters.

Of course the big problem with that is that its application could change who the
top-2 votegetters are. If the process is then repeated, with those as the top-2,
then of course the winner could change with each new application of that 
conditionality option, resulting in an unstable mess.

I don't know if there's a way to avoid that problem. If there were, it might be
a good substitute for conditionality by mutuality.

I should add that that consideration is the best basis for Condorcet: 

One would like to give, to below-favorite candidates, a vote that doesn't count
against anyone s/he ranks higher. That goal naturally leads to a rank count that
amounts to Condorcet--the method that elects the CW if there is one.

But all the Condorcet versions that I've heard of fail FBC and CD (they have
the Co-operation/Defection problem; they fail in the Approval bad-example).

But what if those two problems could be somehow fixed for Condorcet?

After all, Condorcet is what naturally follows from the desire to give to
below-favorite candidates a vote that doesn't count against anyone
higher-ranked.

These comments about fixing Condorcet in that way, and about top-2
conditionality are just preliminary tentative comments.

Mike Ossipoff





 


 		 	   		  
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.electorama.com/pipermail/election-methods-electorama.com/attachments/20120107/9245a1df/attachment.htm>