[EM] Kristofer: Yes, maybe Condorcet could overcome its disadvantages.

Kristofer Munsterhjelm km_elmet at lavabit.com
Sat Apr 28 10:13:37 PDT 2012


On 04/27/2012 07:32 PM, Michael Ossipoff wrote:
> ...but wouldn't it be better not to have the disadvantags?

Sure. It would be better to make Arrow and G&S go away, but it's not 
going to happen. If you want to have Condorcet, you have to give up some 
desiderata. If you want FBC, you have to give up quite a bit, even for 
"weak" FBC which is fairly strong by itself.

> As you suggested, if Condorcet were proposed municipally, maybe (as
> was the case with IRV in some municipalities) big outside money won't
> come in to emphasize rank-balloting methods complete novelty,
> unknown-ness, unfamiliarity and unpredictability.

They didn't emphasize any of the weird properties of IRV. Warren et al. 
did that. What they *have* emphasized is the unnatural nature of any 
method that isn't Plurality, and that is a hurdle to Approval, too. Yes, 
you may counter by saying you can explain to people that Approval is 
"setwise voting" and thus is one man one vote (or that in Approval, any 
vote can be countered by an opposite inverted vote), but that still 
means you have to show people how to look beyond their immediate intuition.

> But maybe the reason why that wasn't always done to IRV proposals
> was because the big opponents knew that IRV wasn't going to really
> improve on Pluralilty anyway, given current electorates.

I doubt it. Of course it could be the case that high-profile people who 
supported IRV, like Howard Dean, were complicit. But if so, they're 
doing a *really* good job at faking their enthusiasm.

> And you further suggested, when people become familiar with
> Condorcet's properties, by experience in municipal use, maybe they'll
> realize that FBC violations are rare--usually no one will regret
> ranking their favorite in 1st place. Right, and a good enough boxer
> might be able to win while holding one hand behind his back. Heroic,
> but maybe not the best idea.

Arrow and G&S implies every "boxer" has to hold something behind his back.

> Why would we want to have the disadvantages of Condorcet, if they can be
> avoided?

Because we want the *advantages* of Condorcet. Advantages like rank 
expressivity when at least some significant fraction of the voters are 
honest. Advantages like the ease that significant fraction can vote 
without having to use dice, know their preferences to the level of 
olympic judges, or keep updated with iterations of polling; and 
advantages like defensibility of the victor when there's a CW, both 
directly and indirectly. The CW would beat any other candidate 
one-on-one in a runoff; on the other hand, Approval's "the candidate 
most people approve of" begs the question because Approval "approvals" 
are so often strategic. If enough people are honest, Condorcet's 
defensibility is real, too. It means no angry majority can be unified 
behind a repeal simply by the cause "you didn't like the winner".

> Not only would approval have an incomparably better chance for
> relatively quick national adoption, but Approval (nothing other than
> an elimination of Plurality's nonsensical falsification-requirement)
> rightfully qualifies as a voting-rights issue, a rights remedy that
> could and should be required by the courts.

Alright, if you can push directly for a national change, do so. I'm not 
going to stop you :-) However, I think (and this may be wrong, of 
course, but it's the impression I get) that national change is pretty 
locked tight, at least for something as unfamiliar as voting reform. 
Voting affects everybody, and it's brittle - use a nonfunctioning method 
and democracy is compromised. It's not like the tax code or an 
environmental tweak where small changes to the regulation leads to small 
changes in the outcome.

> And, if anyone is going to propose something more complicated than
> Approval, it should be something that better gets rid of the
> co-operation/defection problem (C/D). ICT greately mitigates that
> problem, is defection-resistant. That can't be said for Condorcet.
> Compare Condorcet and ICT in the usual Approval bad-example, the
> 27,24,49 example.

CT is Condorcet (Smith, even), and you said CT handles the defection 
problem. So Condorcet in itself doesn't preclude defection-resistance. 
So if you can find a Condorcet method that passes defection resistance 
as well as the criteria that the advanced Condorcet methods do (clone 
independence, ISDA), that would be something. But we might have to pick 
our criteria compliances even then. Schulze argues that his method, by 
finding strongest beatpaths, makes it possible to rebut any arguments of 
the form "X won yet Y beats X" by a strong claim of the sort "but X 
beats Z beats W beats Y and each of those victories are stronger than 
Y's victory over X". It might very well be true that beatpath methods 
can't solve the defection problem; if so, one would have to choose 
between strategic resistance within the method ( defection problem 
resistance) and resistance outside of the method (from arguments that 
the winner did not deserve to win).




More information about the Election-Methods mailing list