[EM] strong defensive strategy criterion

[James Green-Armytage]

>
>To conclude that James' interpretation is most reasonable,
>I think one must take Mike's words out of context, since 
>elsewhere Mike wrote that truncating a preference shall 
>_not_ be considered falsely voting two candidates equal.

	Yes, I assumed that Mike didn't intend it my way, but I thought it was an
interesting definition all the same.
>
>But I'm just nitpicking, since there's a legitimate
>question as to which version is the better criterion.

	I agree, in that I think that what Mike intended in his definition is
hardly worth worrying about. Also, I don't know if one definition
necessarily has to be better; it is perfectly possible to use both, as you
seem to have done, with "sincere defense" and "minimal defense".
>
>James' version appears to me to be equivalent to my
>"sincere defense" criterion, 

	Yes, it looks that way to me, too. I can refer to it as the "sincere
defense" criterion from now on, to avoid confusion.

>which can only be satisfied 
>if the voting method is more complex for the voter than 
>merely expressing an order of preference.  The added 
>complexity means it may not be a better criterion.
>It depends on what the voters can handle and will 
>tolerate...

	Right, again, it's not a matter of "better", necessarily.
>
>The cardinal pairwise strategy James suggested for the 
>majority trying to ensure x cannot win, which involves 
>giving a rating of 100 to some compromise y they all 
>prefer over x, raises the question whether voters can 
>be relied on to carry it out.  Giving a maximal rating to 
>a compromise candidate is similar to ranking a compromise 
>equal to one's favorite

	Not really, because in my cardinal pairwise method, you can give equal
ratings while maintaining unequal rankings. The rankings information is
what determines the winner of the pairwise comparisons, so giving extra
candidates a maximal rating will not alter the direction of pairwise
defeats, will not change the members of the Schwartz set, etc.
	Also, actually ranking later choices in the compromise group at 100 is
rarely necessary to prevent members of the greater evil group from
winning. It depends on the size of the majority, the size of the cycle,
etc. I just wrote "100" because that is the simple, mathematically
tautological way to do it.

>Long ago when I wrote about the sincere defense criterion 
>and several families of voting methods that satisfy it,
>all of which allow the voter to optionally insert a 
>dividing line somewhere in her order of preference,
>I stressed that the dividing line should NOT be interpreted
>as an approval/disapproval line; it should be publicized
>as a strategic device that does not correspond to any 
>sincere preferences.  

	This is an interesting concept. Jobst has been pressing me to define
sincere cardinal ratings, but I have tended to take your point of view,
i.e. that the ratings are largely a counterstrategic device, and hence
should not be expected to conform to a definition of sincerity.
>
>The second wording of minimal defense (which I described 
>as "nearly equivalent") is not strictly weaker than sincere 
>defense; it's different.  It specifies a particular simple 
>voting strategy that must be effective and does not mention 
>any input beyond orders of preference.  I don't recall the 
>definition of James' cardinal pairwise method, so I cannot 
>say whether it satisfies the second wording.

	No, I don't think that it does.
	But, in general, if you don't remember what cardinal-weighted pairwise
is, I heartily invite you to read a current version of my proposal. I've
updated it quite a lot recently. Currently, there are two versions
available in pdf format, a longer version and a shorter version.
the long version (21 pages):
http://fc.antioch.edu/~jarmyta@antioch-college.edu/cwp21.pdf
the short version (13 pages):
http://fc.antioch.edu/~jarmyta@antioch-college.edu/cwp13.pdf

my best,
James