5 ms·

Isn't the problem that although rank-based systems have many desirable properties, they are particularly prone to strategic voting and manipulation? I personall

by 13415 6y ago

Isn't the problem that although rank-based systems have many desirable properties, they are particularly prone to strategic voting and manipulation? I personally think they are good in cooperative settings where voters have no strong incentive to vote strategically, but are they really so desirable for adversarial situations like national elections?

tunesmith 6y ago
I suppose you could extend that all the way back past the practice of voting... ultimately democratic voting isn't as good as some kind of scalable deliberative consensus mechanism that educates people, but we don't have one of those. But anyway, there's also a lot of misinformation out there about voting methods. For instance, most sources say that all Condorcet methods are flawed, but that's because they conflate Condorcet counting with the various tie-breaking methods when there are cycles. The latter can be flawed, but the former is not. If there's a Condorcet winner, it doesn't run afoul of Arrow's theorem or anything. The main "flaw" with a Condorcet winner is that you can have a Condorcet winner without that candidate having "majority approval", depending on how you measure "approval".
scythe 6y ago
>they are particularly prone to strategic voting and manipulation? IRV and AV are particularly prone to strategic voting. Amusingly, the best catalog of IRV's failures is generally provided by AV proponents, and the best catalog of AV problems can be found on FairVote.org. The ranked-choice systems which satisfy a criterion called ISDA -- independence of Smith-dominated alternatives, or more poignantly, independence of losers -- are harder to manipulate. These are all "Condorcet" methods, because ISDA is a generalization of Condorcet. ISDA means roughly that if a candidate is not in contention to win (the "Smith set"), it doesn't matter where they appear on your ballot. I could rank Vermin Supreme and Ron Paul #1 and #2 and the algorithm basically cancels it out and starts me at #3. The reason that ISDA is so effective at preventing strategic voting is that in order to vote strategically, I have to misrank a candidate who might actually win; in other words, I must accept a realistic risk of encouraging an outcome I don't really want. That's a pretty good disincentive, in my opinion. However, the worry about strategic voting seems to be at least a little overblown. Australia does deal with "how-to-vote" cards[1], but this is a way to prevent "allied" parties from accidentally competing with each other. In 2009, there was concern over a monotonicity violation in Australian elections where supporters of one of the parties were encouraged to misrank another party. However, the voters didn't go along with the plan, and instead voted honestly: https://www.theage.com.au/national/victoria/howtovote-cards-awaken-the-rebel-in-innercity-voter-20101115-17uga.html https://www.theage.com.au/national/victoria/howtovote-cards-... https://www.abc.net.au/news/2011-05-04/an-example-of-non-monotonicity-and-opportunites-for-tactical-vot/9389768 https://www.abc.net.au/news/2011-05-04/an-example-of-non-mon...
jlokier 6y ago
I believe it's the other way around. Good rank-based systems are less prone to strategic voting and manipulation than plurality voting (= an ordinary vote), because each voter can safely put their preferred candidate first, even if that candidate is unlikely to win, without harming the chances of their next-preferred candidate. They don't need to vote strategically to try to keep someone they particularly dislike out. The no-so-good rank-based systems are less effective at this separation (though no system can be perfect at it), so people still vote strategically of course. But the pressure to do so is still reduced compared with plurality. Plurality is probably the worst of all options for pressuring people to vote strategically.

by 13415 6y ago

tunesmith 6y ago

I suppose you could extend that all the way back past the practice of voting... ultimately democratic voting isn't as good as some kind of scalable deliberative consensus mechanism that educates people, but we don't have one of those. But anyway, there's also a lot of misinformation out there about voting methods. For instance, most sources say that all Condorcet methods are flawed, but that's because they conflate Condorcet counting with the various tie-breaking methods when there are cycles. The latter can be flawed, but the former is not. If there's a Condorcet winner, it doesn't run afoul of Arrow's theorem or anything. The main "flaw" with a Condorcet winner is that you can have a Condorcet winner without that candidate having "majority approval", depending on how you measure "approval".

scythe 6y ago

>they are particularly prone to strategic voting and manipulation? IRV and AV are particularly prone to strategic voting. Amusingly, the best catalog of IRV's failures is generally provided by AV proponents, and the best catalog of AV problems can be found on FairVote.org. The ranked-choice systems which satisfy a criterion called ISDA -- independence of Smith-dominated alternatives, or more poignantly, independence of losers -- are harder to manipulate. These are all "Condorcet" methods, because ISDA is a generalization of Condorcet. ISDA means roughly that if a candidate is not in contention to win (the "Smith set"), it doesn't matter where they appear on your ballot. I could rank Vermin Supreme and Ron Paul #1 and #2 and the algorithm basically cancels it out and starts me at #3. The reason that ISDA is so effective at preventing strategic voting is that in order to vote strategically, I have to misrank a candidate who might actually win; in other words, I must accept a realistic risk of encouraging an outcome I don't really want. That's a pretty good disincentive, in my opinion. However, the worry about strategic voting seems to be at least a little overblown. Australia does deal with "how-to-vote" cards[1], but this is a way to prevent "allied" parties from accidentally competing with each other. In 2009, there was concern over a monotonicity violation in Australian elections where supporters of one of the parties were encouraged to misrank another party. However, the voters didn't go along with the plan, and instead voted honestly: https://www.theage.com.au/national/victoria/howtovote-cards-awaken-the-rebel-in-innercity-voter-20101115-17uga.html https://www.theage.com.au/national/victoria/howtovote-cards-... https://www.abc.net.au/news/2011-05-04/an-example-of-non-monotonicity-and-opportunites-for-tactical-vot/9389768 https://www.abc.net.au/news/2011-05-04/an-example-of-non-mon...

jlokier 6y ago

I believe it's the other way around. Good rank-based systems are less prone to strategic voting and manipulation than plurality voting (= an ordinary vote), because each voter can safely put their preferred candidate first, even if that candidate is unlikely to win, without harming the chances of their next-preferred candidate. They don't need to vote strategically to try to keep someone they particularly dislike out. The no-so-good rank-based systems are less effective at this separation (though no system can be perfect at it), so people still vote strategically of course. But the pressure to do so is still reduced compared with plurality. Plurality is probably the worst of all options for pressuring people to vote strategically.