Kristofer: > A few methods that pass CD: > > ICT > Symmetrical ICT > MMPO > MDDTR > > A few methods that fail CD: > > Beatpath, RP, Kemeny, VoteFair, MinMax(wv) and apparently all > traditional unimproved Condorcet versions. > > Approval and Score don't pass CD either. But, as I said before, it's > so easy to automatically avoid the chicken dilemma nuisance with a > rank method that there's no excuse for a rank method to fail CD, and > so no rank method that fails CD should be considered.
You replied: All of the methods above fail independence of clones. [endquote] So you're saying that Beatpath, RP, Kemeny, and VoteFair all fail Independence of Clones? (You said "all of the methods above fail Independence of Clones.") You continued: Therefore, I disagree that "there is no excuse for a method to fail CD", unless you can show that it's possible to construct a method that passes both clone independence and CD. [endquote] Ok, you've answered a question that I asked last week: You've told of a property that you regard as important enough to justify losing FBC and CD. That's a start. The next step, of course, must be to tell _why_ Clone Independence is that important. Tell us what majorly adverse societal consequences would result from failure of Clone Independence. And of course it would be necessary that the benefits of Clone Independence remain available under the extreme and drastic strategy-needs resulting from failure of FBC and CD. I've said a lot here about the adverse societal results of failing FBC and CD. But one obvious bad result is that, with the extreme and drastic strategy-needs that I referred to above, that result from failure of FBC and CD, then compliance with the Condorcet Criterion doesn't mean anything. One more thing about that: Though I'm not sure exactly what Clone-Independence is, it seems to me that someone said that Approval and Score pass Clone Independence. Is a clone-set a set of candidates such that everyone who votes one of them over someone votes all of them over hir, and everyone who votes someone over one of them votes hir over all of them, and everyone who votes someone equal to one of them votes hir equal to all of them? And does Clone independence require that removing a non-winning member of a clone-set from the ballots and the election, and then re-counting, shouldn't change the matter of whether or not the winner comes from that clone set? What I've been saying is that the reason why there's no excuse for a method to fail CD is because meeting CD is the only way to significantly improve on Approval and Score. If Approval and Score meet Clone Independence, then TUC's compliance with Clone-Independence doesn't improve on Approval and Score at all. So, if Approval and Score meet Clone-Independence, then Clone-Independence can't be considered an advantage of traditional unimproved Condorcet that improves on Approval and Score. But you could still use Clone Independence to justify traditional unimproved Condorcet's loss of FBC and CD, in a comparison with ICT. (For a comparison with Symmetrical ICT, you'd have to add Zero-Info Probabilistic LNHe (ZLNHe) to the list of things you'd give up for Clone Independence.) Do the FBC/CD methods lose something that Approval has, when they lose Clone-Independence? All of the Condorcet methods lose something that Approval has, when they lose Participation and Consistency. You said: As it stands, the CD methods and the wv/margins methods are mutually Pareto-nondominated with respect to each other*. [endquote] Ok, you're saying that the traditional unimproved Condorcet methods using wv or margins, and the FBC/CD methods don't dominate eachother by desirable properties Of course you're right, because so many criteria have been defined. It's always possible to find a criterion that a method fails. Typically, to gain something desirable, you trade something else that sounds desirable. (I'm going to abbreviate "traditional unimproved Condorcet as "TUC".) But lets not imply that all criteria are equal, and that it's only necessary to name a criterion that TUC(wv or margins) passes and the FBC/CD methods fail, and then call it a stalemate, a draw, an undecideable choice, a matter of subjective opinion. That won't do. It depends on how important, how necessary, Clone-Independence can be shown to be. You've named a criterion that you consider more necessary than FBC and CD. The next step is to tell in what way Clone-Independence is so important and necessary. You said: (That is, assuming that ICT reduces to Condorcet//FPP if there are no truncated votes nor equal-rank ones. If that is wrong, please tell me what method ICT reduces to when all voters rank every candidate, ranking nobody equal.) [endquote] If, in ICT, everyone ranks all the candidates, and no one equal-ranks, then, unless I'm mistaken, the result would be the same as that of TUC//FPP, as you said. By the way, on a different subject, what if Symmetrical ICT's bottom-end improvement were applied to MMPO? I'm not suggesting it, at least not now. But might it not get rid of MMPO's "random-fill incentive", and tame MMPO's burial problem? I realize that MMPO fails the Plurality Criterion, and would probably still have Kevin's MMPO bad-example (I haven't checked). But if some hypothetical future electorate really wanted MMPO's rule as a standard in and of itself, they wouldn't care about those embarrassments. Mike Ossipoff ---- Election-Methods mailing list - see http://electorama.com/em for list info
