About the hypothetical polling with verifiable results: For one voter to deceive another voter about things like who will have a majority or who will outpoll whom, it would be necessary to for him to know something that the other voter doesn�t know. I can�t deceive you about something that you know about. And if I myself don�t know about it, then there�s no way for me to know that I�m deceiving you about it. And it seems a necessary assumption that all voters have access to the same predictive information.
To the claim that a deceiver could harm his cause by deceiving his co-partisans, someone could answer: What if it�s known that Best won�t have a 1st choice majority, but might have a 1st choice plurality. And Worst might or might not have a 1st choice majority? Isn�t it then safe to bullet-vote for Worst in the poll, since there�s no danger of giving away a Best majority win? Sure, but Best could still have a 1st choice plurality, and if enough Worst voters randomly guess otherwise, Best could win--but not of Best voters have been deceived into being convinced that Worst will outpoll best. Obviously, whatever undeceived strategy the Best voters are going to use, whether 0-info or probability-info, is better for their expectation-maximization than a strategy based on being deceived. And what�s better for them is better for you too, if you�re a Best-preferrer with preferences like theirs.
And, as I said, you don�t even know if your false voting for Worst is convincing the Worst voters of something untrue about the outcome.
So giving false poll answers doesn�t sound very profitable.
But if there were a poll-lying problem, it would just mean that knowledgable voters would disregard the polls.
One could suggest a scenario where each faction has its own pollsters whom it trusts. Voters tell the truth to their own pollsters, who never lie in a way that would hurt their faction. But voters may lie to other factions�pollsters, and pollsters may report false results that will help their own faction. That would become a mess in which, as I said, no one would take the polls seriously--just as no one should take polls seriously now.
Anyway, it�s really meaningless to discuss lying by poll-answerers without taking into account lying by those who count and report the polls. For the purpose of these comments I�ve falsely assumed that the polling uses verifiable balloting.
Using Condorcet for the election, or at least using pairwise-count for the (verifiable-balloting) poll would be much more convenient than having a whole series of Approval polls.
And the concern about poll-lying was explicitly based on the mistaken notion that Approval strategy needs polling. Approval voting is reliably informed by previous election results. And in the 1st Approval election, when no previous results exist, then Approval strategy can be o-info strategy (voting for the above-mean candidates) or probability-info strategy based on conversations, and assessments from writers or commentators whom one trusts, etc. A number of probability-info strategies have been described on EM. All the strategies that aren�t o-info and aren�t based on perceived certainties are probability-info strategies.
Mike Ossipoff
_________________________________________________________________
Express yourself instantly with MSN Messenger! Download today - it's FREE! http://messenger.msn.click-url.com/go/onm00200471ave/direct/01/
---- Election-methods mailing list - see http://electorama.com/em for list info
