[EM] Rank methods, contd.

[MIKE OSSIPOFF]

People suggest rough ways to estimate which acceptable candidate, in IRV, is most able to be helped to 
take victory from an unacceptable. But understand that those rough estimates can't be more than a
best guess.

I don't think that _any_ method has as simple and clear strategy as Approval. In fact, is the strategy
of other methods even _known_? Sure, you can say "In IRV, rank the acceptable candidates in order of their ability
to thereby be helped to take victory from an unacceptable", but no one can say exactly how to do that. 

I've spoken of Approval's unique result-optimizations. I'm referring to the fact that, when people are voting in
their best interest (expectation-maximization), Approval will elect the candidate who is above-expectation for the
most voters. The result that will pleasantly surprise the most voters. That's true whether or not the election is u/a,
and it's true whether or not the election is 0-info.

Someone will say that this is a "truism", or a circular statement, but obviously Approval elects the candidate who is
approved by the most people. Truism or not, that is something of value. The candidate for whom the most voters have said,
"I give that candidate an approval".

Of course, if it's a u/a election, that's the candidate who is acceptable to the most people.

I claim that our public political elections are u/a for pretty much all voters. No matter what your political
persuasion, there would be (in a non-Plurality election at least), for every voter, unacceptable candidates
who could win. But of course we won't have proof of what the winnable candidate lineup will look like with Approval, or other
good methods, until we have them in use.

In fact, if people dislike corruption and bought-ness as much as they say they do (and I think that they do), then
they've been voting for unacceptable candidates all this time, and our elections, even now, are u/a for pretty
much everyone. 

Ask someone if they think that bought-ness and corruption are really acceptable. Does anyone really feel that it is?

As I said in my reply today, Approval's unique optimizations and its uniquely simple, clear strategy make it
the best method in a strong meaningful sense. 

And, due to Approval's easy hand-countability, maybe its kind of best-ness is the kind that would be actually feasible
in practice. It's certainly the kind that's attainable, due to Approval's unequaled elegance and simplicity, as the
obvious and natural improvement on Plurality--in contrast to the many elaborate contraptions known as rank-methods, each
seemingly an arbitrary machine whose possible problems and unintended consequences will seem suspect, to enactment-campaign
voters. ...And each of which leaves people asking, "But why this elaborate contraption instead of a different one?"

Still, some (including even me, I admit) would like more preference-level expressiveness. I claim that that
is highly over-valued, of questionable value. But I myself would use Approval options that give that expressiveness. 
Call that a brave, Quixotic ambition.

For that goal, I suggest that ICT is the best. As I said yesterday, maybe there could be a much more complicated method
that can improve on it.

For the purpose of trying to find the sincere CW, to inform Plurality strategy, I suggest that ICT is the only good choice.
Condorcet? No, because any burial or defection strategy would spoil its Condorcet efficiency. And I wouldn't use MMPO or MDDTR
for that purpose either, due to their successful burial strategy.

But, in public elections, as the main method, I have nothing against MDDTR and MMPO. The burial problem is still there,
but I don't think it would be a problem, as people discover its downside. It isn't very well deterred, but I still don't
think it would happen on a problematic scale. 

The great beauty of MDDTR and MMPO is their _great simplicity of definition_. MMPO's simplicity of definition is unmatched
by anything other than Approval. So I'd like to propose MMPO or maybe MDDTR for public elections. But others here have
pointed out some criticisms that they'd be vulnerable to. Though I've told why I consider those criticized "faults" entirely
unimportant, it's also true that opponents and media would make much of them, and would have the advantage when it comes
to having the public's ear. So reluctantly, I'd decline to publicly propose MMPO and MDDTR.

Someone argued for Condorcet vs ABucklin. As I've said, if Condorcet's result is distorted by favorite burial, then its
Condorcet Criterion compliance is meaningless. ABucklin meets FBC, as do Approval, MCA, MTA, MTA2, MTA/MCA, etc.

I recently defined ABucklin2. I'd like to define ABuckline3: Same as ABucklin2, except that, when 1 or more candidates
attain a majority, then, instead of deleting all of the ballots that have already given a vote to a majority-possessing
candidate, delete only the ones that give 1st rank position to a majority-possessing candidate. I think that ABucklin3
is better than ABucklin2. Either is an interpretation for how MTA2 could be generalized to unlimited rankings, but I now
feel that ABucklin3 is more in the spirit of MTA2.

It goes without saying that MCA is simpler than MTA, MTA2 or MTA/MCA; and that ABucklin is simpler than ABucklin2
or ABucklin3. I'd probably therefore mention MCA first, then, as a 3-slot method; and ABucklin first, before
ABucklin2 or ABucklin3.

When ABucklin2 and ABucklin3 use the rankings' approval-cutoff, someone could argue that a strategically-placed approval
cutoff among the reduced set of majority-possessing candidates might be different from where it would be among the entire
set of candidates. Yes, but not if it's a u/a election. And I feel that it's always really a u/a election.

Mike Ossipoff





wwwwwwwwwwwwwwwwwwwwwwwwwww
've said seemingly contradictory things about IRV. It's particularly flagrant FBC failure makes it entirely inadequate for
public political elections, more so than Condorcet, which, too, is inadequate due to FBC failure.

And I also said that IRV would be a fine method, were it not for the public's inclination towards resigned, cowed overcompromise,
and their very sad and disastrous lowering of standards for acceptability. "Vote for the least bad of the corrupt candidates, because they're the
winnable ones".

But how could IRV be alright, even with better voting? Wouldn't people wanting to maximize their expectation still favorite-bury sometimes?
Sure. But if their judgement was at all reasonable, they wouldn't do so in a way that would bother me. It comes down to what one calls
"acceptable". 

IRV's strategy in non-u/a elections would be especially complicated to describe, but it's a sure thing that it would often involve favorite-burial,
to maximize a voter's expectation. Without knowing the details on how to vote (no method matches Approval's clearly-defined simple strategy,
for any kind of election), voters would still do favorite-burial.

As with Approval, and probably all methods, IRV's expectation-maximizing strategy is simpler in a u/a election. But that doesn't mean that
it's simple enough for anyone to know what it is, exactly.

IRV has two great advantages: It is, as I understand it, entirely defection-proof. And it meets the Mutual Majority Criterion.

If we had the kind of electorate who could handle FBC failure, then we could discuss the relative desirability of Condorcet and IRV. But we
needn't bother, because they both fail FBC, disqualifying both of them under existing conditions.

Anyway, ICT has all of what I consider to be the most important properties. Regrettably, of course, it's almost surely too complicated to be considered as
a first proposal. But, when the public are more informed and used to voting system reform, ICT would become a good proposal. ICT is probably the
best, unless something much more complicated could improve on it.




 		 	   		  
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.electorama.com/pipermail/election-methods-electorama.com/attachments/20120410/9ef3c0c1/attachment-0003.htm>