[EM] Re: Kiss your paradigm goodbye

[MIKE OSSIPOFF]

James--

You posted an example in which, in wv, offensive order-reversal took victory 
from the CW, and elected the offensive strategists' candidate contrary to 
majority wishes, as I define that term.

Then you said:

IRV will pick A either way, and in general IRV will not be vulnerable to
manipulation in this example.

I reply:

It's common knowledge that IRV isn't vulnerable to manipulation. As I've 
been trying to get through to you, you're barking up the wrong tree when you 
talk about vulnerability to mainiulation.
I've told you why that is, in recent postings.

You continue:

	So, if the strategy is successful, it causes a problem that is not found
in IRV.

I reply:

No, not really. That problem is found in IRV. IRV will often fail to elect a 
CW, and/or violate majorilty wishes. So then, how are IRV and wv different? 
Two ways:

1. In IRV that happens automatically, easily, without anyone using offensive 
strategy. In wv, it can only happen when someone uses risky offensive 
order-reversal.

2. In IRV, preventing that result requires defensivse order-reversal, often 
favorite-burial. In wv, even when someone is trying offensive 
order-reversal, it can be thwarted, and the problem thereby avoided, by mere 
equal-ranking. In fact, tuncation can be used as a deterrent that will 
prevent the offensive strategy from even being tried.

You continue:

Which is to say that the burying-vulnerable methods contain at
least the _possibility_ of a type of problem occurring that cannot occur
in IRV.

I reply:

Wrong. The type of problem consisting of failing to elect a CW, or violation 
of majorilty wishes is a type of problem that will often happen in IRV. In 
fact, with IRV that problem will happen without any offensive strategy being 
used by anyone, and, and it will require favorite-burial to protect against 
that result.

Now, does it always happen with IRV in the same examples as in wv? Of course 
not. They're different methods. Different methods give different results 
with the same ballots. That's why they're called different methods. If 
you're saying that you expected different methods to have their problems in 
the same examples, then let me be the first to tell you that different 
methods often have their problems in different examples.

How often will it happen in IRV? For simplicity, let's consider 3-candidate 
examples. It will happen when the middle CW's favoriteness support is less 
than the difference between the favoriteness support of the two non-middle 
candidates.

A sufficient, but not necessary, condition for that, when there's a middle 
CW, is for the favoriteness support of all the candidates to be within a 
factor of 2 of eachother.

How likely is the possibility? That is another, much more
complicated discussion.

What does it take for that to happen in wv?

You already know part of the answer to that:

1. The A voters must be willing to attempt risky offensive strategy.

2. No one must use any counterstrategy. That would be odd, for the A voters 
to organize their offensive order-reversal, without tipping the B voters off 
about it. Having heard about the A voters' plans, the B voters wouln't rank 
A. When refusing to rank A, the B voters would publicize that refusal in 
advance.

And that would make it distinctly inadvisable for the A voters to attempt 
their offensive strategy.

3. A would have to be certain to pair-beat C. Otherwise the offensive 
order-reversal would make C into the BeatsAll winner, cause C to win as 
would a CW.

4. The B voters must prefer A to C, and rank A, in keeping with 2 above. But 
it's worth adding that this means that, in wv, you can only steal the 
election from voters who are trying to help you. Congratulations, you must 
be proud of yourself.

And don't expect another opportunity to do that. Don't expect anyone to 
trust you again. That's it.

5. Of course A's defeat in the resulting strategic circular tie would have 
to be the smallest of the three. And, at the very least, to avoid giving the 
election to C, it's necessary that C not have the smallest defeat. Could you 
predict those things? But you think the A voters could?

For the offensive strategy to succeed, in addition to no counterstrategy 
being used, and in addition to A pairwise beating C,  in addition to the B 
voters preferring A to C, it's also necessary that A be the sincere 
Plurality winner. That isn't necessary for IRV to have its problem.

Mike Ossipoff

_________________________________________________________________
Express yourself instantly with MSN Messenger! Download today - it's FREE! 
http://messenger.msn.click-url.com/go/onm00200471ave/direct/01/