[EM] Brief MJ comment. ITC & conditional methods.

[MIKE OSSIPOFF]

I don't oppose MJ. I like Approval, and MJ is Approval. 

True, it's an unnecessarily complicated, elaborate and expensive Approval, but it's still Approval.

In Ruderman's poll, I voted:

1. Approval
2. RV
3: MJ

I would have put the Approval cutoff just below MJ, but I didn't know that it was an Approval cutoff.

I don't oppose ICT. I consider it one of the better full-ranking methods for public elections.  ...along with
AOCBucklin and ACBucklin. And maybe IC(wv-conditional) or IC-ACBucklin. I don't claim to know which of all those
is better.

There are certain strategy problems shared by the pairwise-count methods. Condorcet(wv) avoids &/or minimizes some of
them, but ICT avoids Condorcet(wv)'s worst problems: FBC-failure and co-operation/defection dilemma.

The pairwise-count problems that ICW doesn't get rid of don't really seem problematic to me. Problems sure, but
secondary problems. Compare them to the kind of problems Plurality and IRV and most rank methods have.

It's my opinion that the best full-ranking method is AOCBucklin. But that's really a matter of opiniion.

The optional conditionality gives the voter more to do other than just sincerely rank. But that's the trade-off, for what
I consider a more deluxe method. ICT is effectively automatically conditional, like ACBucklin. For the voter, more to do means
more choices to make, more work in voting; but it also means more freedom, expressiveness and control of what one's ballot
will do.

It's difficult to compare a pairwise-count method to an Approval-based method, because they're so different. I feel that the
Approval-based methods are closer to the rightness of Approval. Approval's simplicity and rightness are there in stepwise
Approval too (that's ABucklin). In comparison, the pairwise-count methods are more like a machine that does certain things
convincingly, but also has unintended consequences, as do many machines.  ...not that all methods don't have unintended
consequences too, but that seems less for the more natural Approval-based methods.

The pairwise-count methods are the ones for reliably choosing the CW. But we all know that Condorcet's Criterion is
incompatible with FBC anyway.

ICT does something that's more similar to looking for a CW, but when we abandon Condorcet's Criterion, it isn't obvious
what else is most important.

One difference between ICT and ACBucklin is that Bucklin starts at the top and looks for a majority that can be assembled
from the top of people's rankings. In a way, doesn't that make more sense? Say one ballot has y in 100th place, and x
in 99th place. Another ballot has y in 1st place and x in 2nd place. Another has x in 1st place, and y in 4th place.

Pairwise-count counts those ballots' treatment of x & y equal. But are they? Might it not make more sense, the way ABucklin
starts from the top down, in its search for a majority?

ACBucklin, like ABucklin and AOCBucklin meets the Mutual-Majority-Criterion. 

(Though IRV meets MMC, and doesn't have the co-operation/defection problem, it has (as you know) a particularly bad and
frequent FBC failure).

In optionally conditional Approval (AOC), you should give a conditional vote to a candidate if you want to help hir, for protection
against someone worse, but you don't want your help to be usable against candidates whom you like more. S/he's better than
the worse ones, but s/he's still a rival to the better candidates. Likewise in AOCBucklin.

That's the obvious first improvement on Approval. Sure, MTA, MCA & ABucklin improve on Approval too, slightly, by adding
more levels of majority-rule protection. But that's insignificant in comparison to getting rid of the co-operation/defection
problem. 

So: Obvious first Approval improvement: AOC. After that, maybe MTAOC, MCAOC or AOCBucklin.

And, even if ICT is better than ACBucklin, a problem of ICT is: How do you get there from here? If you propose ICT, people
can ask, "But, with so many full-ranking methods, and with so many methods, why choose that one particular one?"

That question isn't a problem for Approval and its natural and obvious improvements.

One advantage of the optional conditionality of AOC is that an option is much more difficult to criticize than a non-optional
change or feature.

Returning to the comparison of ICT and ACBucklin, if you truncate in ICT, you still can benefit from help given by the
supporters of the candidate you truncated from your ranking.  That isn't so in ACBucklin.

I wanted to post these comparison of these methods.

Mike Ossipoff

















A fair comparison would be one between ACBucklin and ICT, because they're both automatically conditional. In the Approval
bad-example, it isn't enough for the A voters to co-operate. C will win unless the B voters co-operate too. The efficacy of the A voters'
support for B is conditional upon co-operation by the B voters.

With the deluxe AOCBucklin, the A voters have a choice between giving to B a conditional vote or an unconditional vote.














 		 	   		  
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.electorama.com/pipermail/election-methods-electorama.com/attachments/20120221/028bc8ce/attachment-0003.htm>