[EM] Re: S/WPO

[James Green-Armytage]

James G-A replying to Forest, on the subject of cardinal-weighted pairwise
(CWP)...
>
>I've delayed bringing this up because I didn't want to dampen your 
>spirits; I think that Cardinal Pairwise suffers from a "bunching up" near 
>the extremes problem similar to straight CR, and that S/WPO and Approval 
>Weighted Pairwise avoid this problem.

	This doesn't dampen my spirits; you're not the first person to make this
point. I've answered it before. First of all, I'm not fully convinced that
this incentive for extreme ratings will be as strong as you say. (So far,
no one has presented me with a proof that the incentive will even exist,
given moderately incomplete information) Secondly (and much more
importantly) even if there is a strong incentive for extreme ratings
assignments, then the worst case scenario is that it will reduce to AWP.
Which is not bad at all. If CWP offers even a little bit more resolution
than AWP, it can only be an improvement. Plus, a 0-100 scale is more
intuitive to voters than an "approval cutoff", even though an approval
cutoff is technically simpler.
>
>To get the additional resolution that you seek from Cardinal Pairwise you 
>should interpret your Cardinal Ratings ballots dyadically.

	How do you propose to do that?
>
>Dyadic ballots or dyadic interpretation of ratings avoids this bunching
>up 
>problem the same way S/WPO does, but with a greater variety of strengths 
>for the preference relations.
>
	How do you propose to count these varying strengths?

my best,
James