MIKE OSSIPOFF nkklrp-at-hotmail.com |EMlist| wrote:
Markus doesn�t like the mention of preference in a criterion, and implies that it�s somehow necessarily imprecise. I don�t agree. Here are a few approaches to the meaning of preference:
I can't read Markus's mind, but I suspect that his objection is not so much to the definition of "prefer" but rather to the fact that it refers to a state of mind of the voters separate from the *actual* votes cast. As I wrote earlier today, "normal" criteria (i.e., that do not originate from Mike) make no reference whatsoever to voter preferences except in terms of the actual votes cast. In other words, "normal" criteria apply to the tally rules only and not the voting strategy. They may have *implications* for voting strategy, of course, but they do not address it directly.
Having said that, I think I agree with Mike on this one. I think it is reasonable to model true voter preferences as a "sincere" rank list for each voter. I also think it is reasonable to model the true preferences of each voter as a cardinal rating for each candidate, though that would be a more specific model. Either case is reasonable so long as the model is clearly stated. Unless I am missing something here, the definition of a preference is then trivial: if the voter ranks or rates one candidate over the other, the voter "prefers" the higher ranked or rated candidate.
Unlike others here, therefore, I think Mike-style criteria can be, and usually are, well defined. I just think that some people are confused because, unlike other criteria, they directly address both voting strategy *and* tally rules. The question remains, of course, of how *useful* Mike-style criteria are even if they *are* well defined. On that matter I differ with Mike.
----
Election-methods mailing list - see http://electorama.com/em for list info
