[EM] James, SFC

[James Green-Armytage]

James replying to Mike...

>There's a big difference between a problem that can happen if
>signifilcant 
>numbers of voters attempt offensive order-reversal, and one that happens 
>automatically, or by mere truncation. Offensive order-reversal on a scale 
>sufficient to affect the outcome isn't likely. 

	I'm not sure if I agree with this. In a ranked ballot system, might not
we expect voters to bury the main competitor to their favorite candidate
in last place, even if the competitor isn't actually their sincere last
choice? For example, wouldn't we expect a Bush backer to rank Kerry in
last place even if there were other candidates on the ballot whom that
voter would like even less than Kerry if that was the only choice?
	I suppose that this kind of non-strategic order-reversal will probably
tend to exist on a smaller scale than non-strategic truncation, but it
should still be taken into account somewhat. 

>When SFC says "if noone 
>falsifies a preference", that's a convenient way to say it,but obviously
>the 
>same thing is true if preference-falsification isn't done by a group
>large 
>enough to change the outcome.

	Right.
>
>
>I always say that under certain plausible conditions, it won't be
>necessary 
>to do other than rank sincerely.

	You think that voters should give full sincere rankings in WV? I thought
we agreed that voters should only rank one of the two major party
candidates (or truncate after reaching their approval cutoff, or truncate
after reaching their compromise candidate, or something along those
lines...), in order to assure that attempted burying strategies (OOR) by
the other "side", as it were, could only backfire. 
	By the way, we should try to solidify this a bit. Do you think that
"always truncate after your approval cutoff when a burying strategy
(offensive strategy) is a possibility" is good universal advice for voters
in WV elections.

>You object that that isn't true under all 
>conditions. But such a nonprobabilistic method is impossible. You're
>asking 
>too much. But, under the plausible conditions under which such a
>guarantee 
>can be made, wv makes it, because wv meets SFC.

	Methods that use information other than rankings (like CWP, AERLO/ATLO)
can make it so that voters have more freedom to vote their full sincere
rankings while still employing effective counterstrategy.

>One other thing that you're forgetting: While offensive order-reversal is
>an 
>extreme and drastic offensive strategy, truncation needn't even be an 
>offensive strategy at all. 

	I agree with this; I'm not forgetting it. I also remind you above that
order-reversal is not always done with a specific strategy in mind.

>Truncation can be done just to deter offensive 
>order-reverssal. 

	Yes.

>In fact truncation can be done for no strategic reason at 
>all. Usually truncation won't have a strategic motive. Some people will 
>truncate because there are just people for whom they don't want to vote, 
>from principle or due to distaste, etc. And people will truncate because 
>they don't know much about all the candidates, or because they just don't 
>want to take the time, or don't want to take the trouble, to rank all of 
>them. 

	Yes.

>Maybe the merit differences among their lower choices aren't so 
>important to them. Studies have shown that to often be the case, that 
>perceived utility differences between lower choices tend to be less. So
>it's 
>not surprising if some don't rank all the candidates.

	Yes. And I take this into account when I reason that CWP won't be
vulnerable to truncation in practice, despite the SFC failure. Recall that
the ratings are the determinant of defeat strength in CWP. If I truncate
when I get to candidates at the bottom of my utility range, it usually
won't make much difference, because even if I did rank them out, I would
rate them close to zero anyway. Either way, my ballot doesn't lend much
strength to defeats between them, and my ballot lends close to maximum
strength to defeats where they are beaten by a candidate whom I rate
highly. Hence, I think that only a remarkably low-utility CW could be
deprived of victory in CWP as a result of this kind of non-strategic
truncation. 
	I also argue that CWP won't be vulnerable to strategically motivated
truncation. If supporters of a candidate A are truncating against B (the
sincere CW) in order to cause an A>C>B>A cycle that resolves in favor of
A, I reason that A and B are likely to be very different candidates, in
that the average rating differentials on both sides of the A-B preference
will be high. (If they were similar candidates, then the reward/risk ratio
of the strategy would be fairly low, and hence it would probably not
occur.) If the rating differential on both sides of the A-C preference is
high, then the B>A defeat will be strong, and an attempt to overrule it
via a burying strategy will be unlikely to succeed.
>
>So your effort to equate truncation with offensive order-reversal won't 
>work.

	I'm making no such effort. They are obviously separate and distinct
concepts. However, I do say that order-reversal may be more likely than
you expect, and that CWP can provide similarly high resistance to both
truncation and order-reversal.
	I'm not arguing with you on the WV versus margins question. I believe
that WV is substantially better (as you say, more stable). I also think
that GSFC is a useful criterion. However, I do say that CWP's GSFC failure
will not have the practical significance that you might expect, and that
CWP (without the majority-beats patch) will still be more
strategy-resistant than WV. As I wrote earlier, CWP has anti-strategic
properties that make it highly resistant to both truncation and
order-reversal. I suggest that these properties are too subtle to be
measured by existing ranked ballot criteria, because they work based on a
(hopefully correct) understanding of the interaction between strategy and
preference strength, and preference strength information is not included
in ordinal data.

Sincerely,
James
>
>