We've discussed this before, but I'd like to comment on it again: ------------------------------------------------------------------------------------------ Ideal Society with completely honest voters:
Try to satisfy Rawls' standard. Score balloting. Instruct voters to rate the candidates proportional to utility, to the extent that they can put a number to it. Elect the candidate or alternative whose lowest rating on any ballot is the highest. Remember that all voters are completely honest. Of course there's a possibility of ties, in which case another method would have to be used. Maybe, among the tied candidates, just do an ordinary Score count, using the same ballots. Maybe hold a 2nd election, among the tied candidates, by Approval or MMPO. (Approval would elect the candidate acceptable to the most people; MMPO would elect the candidate with fewest people preferring another to it--If an alternative is really bad for you, you'll rank it at bottom with lots of alternatives over it.) ------------------------------------------------------------------------------------------- Voters vote for self-interest, but agree about Rawls' desideratum: Shouldn't be Condorcet (including ICT)--too majoritarian. Approval, or MMPO (for the reasons given above) would be good. Or, because of MMPO's greater vulnerability to burial, maybe MDDTR (burial wasn't a problem with honest voters, in the previous section). Reducing burial-vulnerability would come at the price of MDDTR's top-count, which isn't in keeping with the desideratum. I suggest that MMPO's burial vulnerability needn't be such a great problem. Every strategy has a counter-strategy (pre-emptive defensive/retaliatory burial?). It would be like a (secondary) chicken dilemma, but not as bad, because it's secondary. MMPO, like ICT and MDDTR, has no primary chicken dilemma. Other than its burial vulnerability, MMPO has good properties. Unlike any Condorcet version (including ICT), MMPO and MDDTR meet Later-No-Harm (LNHa). Unlike any Condorcet version, MMPO (but not MDDTR) meets not only Participation, but also the stronger Mono-Add-Top (as do Approval and Score). Yes, MMPO fails the Plurality Criterion, and has Kevin's MMPO bad-example (which, it seems to me, is a Plurality Criterion failure). But I remind you that the assumption was that the people agree on Rawls' desideratum. That means that MMPO's intrinsic advantage, a count rule that is a standard in and of itself, would be more important to people than an embarrassment criterion. And I remind you that Kevin's MMPO bad-example, though it would be loudly trumpeted by opponents, consists of the election of a winner such that _no one_ thinks that that winner is worse than the other candidates. (I criticized Margins Condorcet for its Plurality failure the other day. I did so because that embarrassment criterion would likely be fatal to an enactment proposal under presently-existing conditions. It's a reason why I abandoned MMPO for proposal under existing conditions). Condorcetists tell us that Plurality Criterion failure is worse than Participation and Mono-Add-Top failure. But failure of Participation and Mono-Add-Top (but especially Plurality) are monotonicity failures, instances of something worse than unresponsiveness, when a method acts contrary to what you express on your ballot. You turn your steering wheel right, and the car goes left.. Could it be that those Condorcetists have convinced themselves of that because they've already decided and agreed that they want to propose Condorcet? You know, "group-think". Those monotonicity violations are instances of your ballot acting oppositely to your intent, and are, as such, worse than Plurality failure or the MMPO bad-example. What might be an example of group-think? Well, just hypothetically of course :-) ,suppose there were a group of people who had decided that they were going to advocate unimproved Condorcet. Well, I'd bet that maybe they'd debate endlessly about which kind of unimproved Condorcet is the best, under the dogmatic, group-think, not-subject-to-question, assumption that their initial belief and choice is right. Remember that this is just hypothetical. (I didn't mean to make this section so long, but I digressed) -------------------------------------------------------------------------------------------- Currently-existing conditions (Count fraud problem, voters pursuing self-interest. No moral consensus): No rank method, due to count-fraud problem, and difficulty of agreement on a rank-count. Approval or Score, for the reasons that I've been giving here at EM. Approval is probably a better proposal, due to its simpler (and therefore less fraud-prone) count, and because it's the natural, obvious, minimal improvement on Plurality. ------------------------------------------------------------------------------------------------ Informational polling, to inform and guide voting in an upcoming Plurality election: Condorcet's Critrerion is very desirable, to find the candidate on whom votes should be combined (something that is necessary in Plurality, with its split-vote problem). Condorcet's Criterion is useful only if the worst preference-distorting strategy needs are avoided. For that reason, I recommend ICT for informational polling. If a rank-count can't be agreed on, or if a handcountable method is needed, then I'd suggest Score, as a 2nd choice. Approval would do, if a still easier handcount is needed. ---------------------------------------------------------------------------------------- Voting on what movie to go to or rent: Favoriteness is all important for that application. Just use Plurality. --------------------------------------------------------------------------------------------- Voting on an ice-cream flavor, or a kind of pizza: Dislike becomes more important (as compared to voting on a movie choice). I'd suggest Condorcet (Of course I'd suggest ICT), Score, or Approval. In Score, voters should be encouraged to vote strategically, unless you believe in the discredited moral philosophy of Utilitarianism. Comparing Score to Approval, I prefer Score's easier fractional rating, useful for SFR, or when it's uncertain whether an alternative qualifies for top-rating. ------------------------------------------------------------------------------------------------------ Approval's simpler (and therefore more secure) count, and Approval's minimal change from Plurality probably make it a better proposal for official public elections. -------------------------------------------------------------------------------------------------------- Mike Ossipoff Mike Ossipoff ---- Election-Methods mailing list - see http://electorama.com/em for list info
