No, it’s team bullet voting.
“Bullet voting” means voting only for your favorite candidate, so you’re inventing a new term here. Okay, fine, call it “team bullet voting” if you like. Point is, it works well. We have the Bayesian regret numbers to show it. And we have a mathematical theorem that in the “worst case” strategic scenario, it elects Condorcet winners, which is a pretty good outcome in most people’s eyes.
And if the green gives the democrat less than full points, they’re basically handing the election over to the republicans.
Not necessarily. Doing so might actually switch the winner from Democrat to Green, which would be good for that Green voter. It just depends on the circumstances. Again, the strategy is well studied.
You cannot afford to vote honestly with score voting.
I repeat: tactical voting is an issue with all deterministic voting methods. Including Condorcet. Score Voting with about 50/50 honest/strategic voting is as good as Condorcet with 100% honesty. And radically simpler and more transparent (you can present every candidate in a bar chart, allowing society to see support for minor party and independent candidates).
You can only vote offensively or defensively, not honestly. And in that case you’ve basically reduced the system to approval voting, except for the few idiots that try to vote honestly (or who pick the wrong strategy) and end up throwing the election one way or the other.
For every person who decides to vote honestly, the quality of election outcomes increases. It is nonsensical to talk about honest voters “throwing the election one way or the other”, given that the tactical voters are the ones having a greater impact on the outcome.
That’s not how elections should be decided.
You’re again making the fallacy I illustrate in this image.
It’s not even bayesian.
Whatever that means. I cited Bayesian regret calculations which show Score Voting leads to better outcomes.
It’s just forcing every voter into the prisoner’s dilemma.
Just like every deterministic voting method.
Basically, unless you’re defining “bayesian regret” as “used it in an AI and got it to beat montezuma’s revenge” then it’s arbitrary.
No, it is not arbitrary. I cited numerous links to proof that the social welfare function is just the sum of cardinal utilities. I doubt you even read/understood them.
If it doesn’t demonstrate actual bayesian behavior then it isn’t bayesian.
Making nonsense statements really isn’t helping your case. If you think some aspect of the calculations wasn’t modeled realistically, you could argue as such. But Smith varied all of the parameters massively (e.g. changing strategic voting from 0% to 100% in small increments), and Score Voting won in all of those parameter settings.
The alternative, which is accepted universally by all social choice theorists, is to define optimality according to the condorcet and dodgson criteria.
Ludicrous. I have cited a multitude of proofs that the social welfare function is cardinal, not ordinal. Incredibly you have never heard of Arrow’s Theorem.
That hardly counts as debunking. It doesn’t apply to real world situations at all
You’re simply wrong. People use cardinal utilities in investment under risk (probability) all the time. PRICES are cardinal.
AND IN VOTING. If my expected utility is 3, and I think X=0, Y=4, Z=5, then it makes sense to vote for Y and Z. If I think Y=2, then it only makes sense to vote for Z. This is cardinal utility, even with the same ranking.
Never in any election are you asked to weight a choice between a pair of candidates vs a single candidate. Never. That doesn’t even make sense.
I have no idea that you’re talking about, and I’m pretty sure you don’t either.
In economics you might make that choice, but a choice like that can also be described accurately by minimax+tideman-score, at least when averaged over many voters.
Okay, run some Bayesian regret calculations on that system compared to Score Voting, and see what results you get. If it does well, present it to the community, here. There are people on there with Harvard stats PhD’s and Princeton math PhD’s. You’ll get some good ego-damaging feedback.
The only voting systems which “properly” pass independence of irrelevant alternatives are stochastic methods like random ballot and sortition. Score voting, at best, only weakly deals with that issue via tactical voting.
Score Voting passes independence of irrelevant alternatives. Instead of “via tactical voting”, you meant to say, “ignoring tactical voting”. Sigh.
The sort of tactical voting that has to be used in special situations like that both requires that the voter has perfect or near-perfect knowledge of the outcome prior to the election, and that they know the optimal strategy, which may be significantly different from normally used strategies.
You’re literally getting this entirely backwards. Score Voting satisfies IoIA given a set of cast ballots without any strategic readjustment when removing an irrelevant alternative.
However in the paper I linked it was argued that irrelevant alternatives aren’t actually irrelevant anyway, so it’s a useless feature.
Of course they are irrelevant. If you’re trying to decide between chocolate and vanilla ice cream, it doesn’t matter what you think of strawberry.
There are countless people who have argued against irrelevant alternatives, including the Marquis de Condorcet himself.
Any organism which acts in such a way is exploitable and will be weeded out by natural selection.
As for the mathematical issues, it has the same problems as borda counts.
Utterly false. Borda is, like most ranked methods, extremely vulnerable to tactics, because e.g. a Green voter wants to rank the Democrat #1, which forces her to rank the Green #2. Thus even if the Green is the most popular candidate, he will lose if people don’t think he can win.
Whereas with Score Voting, giving the Democrat the max score doesn’t prevent you from also giving the Green the max score, so the Green can still win even if people don’t expect it.
The relative difference between 0 score and 1 score is infinite.
No, it is 1. It is not meaningful to use division.
What that means in practice is that score based methods can tell the difference between least-liked candidates and all others, and can discriminate between the least-liked and second least-liked very well, but most of the time it cannot tell the difference between the most-liked and second most-liked candidates, which becomes worse as you add more candidates to the ballot.
This entire paragraph is nonsense. We have Bayesian regret figures. We know Score Voting works.
Assuming people even bother to vote honestly rather than just going all or nothing, which they won’t because honesty a losing strategy.
On the contrary, we have strong evidence that upwards of 50% of people will vote honestly. It is astonishing how little you know about this subject given you could have easily read about over an afternoon before spending all this time typing.
Unless you have a machine to read a quantified “utility sum” directly out of a person’s brain, then this is meaningless.
I don’t need to read anything. I made these people up. I can give them whatever utilities I want to.
> More specifically, if all voters use rational optimal strategy (attempt to game the system to the fullest extent), then they get the Condorcet winner.
You cannot pick a condorcet winner unless people actually express an ordinal preference between candidates.
Wow, you aren’t reading anything I’m linking you to. Again, there is a MATHEMATICAL PROOF. ==> http://scorevoting.net/AppCW
ProTip: Cardinal ratings give you everything rankings give you, PLUS more.
When gaming the system maximally nobody ever votes any way besides all or nothing. You can’t afford to do anything else, because again, anything else is a losing strategy. If I vote for candidates A and B, then I have to give both a full score even if I have a real preference between them, which means that the information needed to determine the true condorcet winner never even makes it onto the ballot.
If you had read that proof I keep linking you to, you wouldn’t have spent that whole paragraph embarrassing yourself by being blatantly wrong. Sigh.
No. Let’s say you support the green party. Too bad, you have to give full score to both the green candidate and the democrat, otherwise the republican wins.
And if the Republican votes for just the Republican, then your vote and his vote perfectly cancel out. You have equal “weight” even though you vote for different numbers of candidates. Again, this was related to the point about how Score Voting makes voters equal unlike most ranked methods. You seem to be confusing two different issues.
If everyone does that, then either the green and the democrat will tie
No! Because some people will only vote Democrat. I cited tie probability calculations.
You’re assuming, incorrectly, that anyone will vote honestly under score voting, when there is a strong incentive against it.
Shows how little you know. We have massive evidence that MANY people will vote honestly.
And it turns out honesty is a pretty good strategy.
Bucklin voting was either abandoned or outright declared unconstitutional because so many people were bullet voting and the authorities kept having to throw cheating ballots out.
It’s perfectly fine that many people will bullet vote. If you held a Score Voting election right now, most Democrats and Republicans would be perfectly reasonable to bullet vote. We see no evidence that anything “bad” happened with Bucklin.
I mean, you’re arguing that too many people were bullet voting, and that was bad, so they forced everyone to bullet vote. WHAT?
The repeal of various superior voting systems seems to have happened because these systems were deemed to complicated, or perhaps because they worked, and special interests didn’t like that. For instance, Marquette, Michigan once used Condorcet voting (Nanson), and that got repealed. Nearly two dozen US cities once used STV/IRV, and that got repealed. There is nothing special about Bucklin here. This country has just not been very favorable to progressive voting reforms.
Of course reversal symmetry is a desirable property. If all individuals have the exact opposite preferences, it is logically necessary that the group also must.
Oh I see you’ve read that paper I linked. Or any of the other articles I wrote. Oh wait..
Nothing in your paper is going to refute something that can be so trivially mathematically proven. Period.
That’s hardly an authoritative source on the resistance to control and manipulation, either. There are dozens of authors who have covered the subject.
I don’t care about authority. If you want me to take your voting method seriously, you need to measure its performance. There is exactly one correct way to do this. Measure its social utility efficiency. You can express that as Bayesian regret, or its opposite.
As for the complexity, it’s exactly the same complexity as any other version of minimax, which is significantly simpler than ranked pairs or schulze. It’s as simple as any condorcet method can be, which is still hardly an argument one way or the other.
Yeah but that’s insanely stupidly complex compared to Score Voting, which also happens to be better.
In comparison, the non-linearities suffered by score voting are a far subtler form of complexity that your average voter who thinks they understand the system will utterly fail to comprehend.
There is no non-linearity. You don’t know what you’re talking about. I assume this is based on your bizarre confused division issue from above.
A bunch of complicated-looking math is not a proof, and this is just another crappy argument for irrelevant alternatives, probably the most useless property for any voting system to have.
The correct social welfare function must obey independence of irrelevant alternatives. This is logically necessary, full stop. If we are asking whether you prefer X or Y, what you think about Z is BY DEFINITION irrelevant.
I already linked you to a paper that thoroughly covers those issues, but apparently you’re either illiterate or being purposefully belligerent.
Your paper cannot possibly refute IoIA. Its necessity is definitional to choice.