[EM] Score DSV

[Jameson Quinn]

I've just put up a page on the electorama wiki for Score
DSV<http://wiki.electorama.com/wiki/Score_DSV>(aka Range DSV), a
condorcet/score voting single-winner method in which a
voter's score ballot is counted as if they'd used the "*D*eclared *S*trategy"
of renormalizing so that the Smith set spanned from 0 (minimum score) to 100
(maximum score).

I'm going to say some provocative things about this method, to ensure that
Marcus and Warren will respond and so to start the debate.

Score DSV is superior, theoretically, to Schulze. Looking at compliance with
results-based criteria, it either complies or does better than compliance
with all the same criteria, and comes closer to complying with the
participation criterion. The only way that Schulze is clearly better is the
summability criterion (which is not a results-based criterion); Score DSV
requires (at most) two rounds of counting, though it does have a variant
which is order-3 summable (that is, using a cube-like matrix instead of most
Condorcet methods' square one). Note that "better than compliance" includes
either compliance with a stricter criterion (in the case of the Strong
Defense Criterion) or non-compliance which actually results in a
higher-utility winner (in the case of: Smith criterion, Mutual majority
criterion <http://wiki.electorama.com/wiki/Mutual_majority_criterion>, local
independence from irrelevant
alternatives<http://wiki.electorama.com/wiki/Independence_of_irrelevant_alternatives>,
Schwartz criterion <http://wiki.electorama.com/wiki/Schwartz_set>, and
the Generalized
Strategy-Free criterion<http://wiki.electorama.com/wiki/Generalized_Strategy-Free_criterion>
.)

I expect that Score DSV will have lower Bayesian Regret than any other
system, even score (aka range) voting, given voters with any reasonable
amount of non-false information making a rational choice of strategies, with
any given (ideologically-biased or -unbiased) mix of minimum
expected-benefit thresholds for voting strategically rather than honestly,
as long as at least half of voters have an "attainable" strategic threshold.
Obviously, this is an empirical question, and "put up or shut up" is a
reasonable response (Warren, what's the link for your source code and for a
clear explanation of your monte carlo models for voter utility?)

(Note: I actually do believe that this method is objectively better, but I
can easily imagine intelligent good-faith disagreements, so I'm
intentionally being a wee bit trollish by stating it so unequivocally. I do
not intend to give offense, though, and I would be happy to be corrected if
I'm wrong.)

Jameson
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.electorama.com/pipermail/election-methods-electorama.com/attachments/20090829/1fe663f3/attachment.htm>