[EM] Methods

[matt welland]

On Wed, 2011-10-19 at 00:51 +0000, MIKE OSSIPOFF wrote:
>  

> 
> Matt-- 
> 
> You're a bit unfair to rank methods. You said that it's difficult to
> figure out the right strategy. 

Hmmm... I'm not sure I said that it is difficult to figure out the right
strategy. One thing that I did say was that "rank methods hide or lose
information". I still think that is true but of course rank methods are
all over the map and are a mathematicians wet dream. so if you say they
can convey the same info then I'll have to  agree to disagree. 

> Approval voting is easiest when some candidates are acceptable and
> some are entirely unacceptable: Just vote for the acceptables.

I think approval voting is always very easy, but yes, it is trivial to
decide in the case of highly polarised candidates.

> If you have no information about winnability, then the strategy is
> simply to vote for all the above-average candidates.
> 
> But if there aren't "unacceptables", and if there is winnability
> information of some kind, then Approval is inherently a strategic
> method. 
> 
> Approval voting is strategic then. As I said, a good strategy is to
> just vote for all candidates who are better than what you expect
> from the election.

Voters may be ill-informed but they are not stupid. I'm pretty sure
99.999% of the voters will have no trouble mastering the trivial
"strategy" needed by approval in some races.

> Bucklin has Approval-like strategy, with, as I said, 3
> protection-levels instead of two. If it's clear who belongs in each
> protection category,
> 
> then the strategy isn't difficult. Other than that, I don't think
> Bucklin's strategy is known, to the extent that Approval's simpler
> strategy
> 
> is known. But, knowing who you'd vote for in Approval can inform your
> Bucklin voting, because Bucklin lets you rank people you wouldn't vote
> for in Approval, safe in the knowledge that you've equally top-ranked
> the best set of candidates.
> 
> Condorcet(wv), MDDA & MAMPO can be more free of strategy if no one
> falsifies preferences, due to their SFC compliance.
> 
> But, if some voters are likely to use burying strategy, then it's
> desirable to thwart them, and enforce the methods' SFC benefit,
> 
> by refusing to rank the candidates of the likely reversers. If that
> sounds complicated, it isn't really. Just use some judgement about
> how far down you rank. Don't rank the really odious candidates, or the
> ones who (or their supporters) are antagonistic to your
> candidate. 
> 
> Additionally, of course, the buriers intended victims have the same
> polling information as do the buriers.  And it isn't possible to
> organize a large scale burial strategy without it leaking to the
> intended victims. Burial only works against people who are trying to
> help you. 
> 
> And, finally, what if the burial succeeded, this time. What about
> subsequent elections. Do you think that party will get ranked in the
> victims' rankings again?
> 
> Oh, one more thing, in the above-listed methods, as I said, to steal
> the election for a candidate by burial requires that a large fraction
> of his favorite-supporters do burial. And thwarting and penalizing the
> burial requires only a small fraction of the intended victims to
> truncate the buriers' candidate.

> I don't agree that rankings are awkward or painful to vote. I know
> whom I like better than whom. But I'll agree with you on this:
> 
> I consider our elections to consist of acceptables and unacceptables.
> That kind of election is _made for_ Approval. Bucklin, however, 
> can let you rank among the acceptables, while still giving them full
> SDSC protection from the unacceptables, if they have majority
> support. And if they don't have majority support, not method can save
> them.
> 
> Likewise, MDDA and MAMPO make the same thing possible.
> 
> And no one needs to vote 50 times in succession.

Of course not, that misses the point. A proponent of any of these
systems is going to be blind to the costs of the system. I've done the
exercise of casting a ballot using different systems many times in
succession. My original hypothesis was that I just wasn't used to it.
After a dozen times I knew that the mental energy in ranking was *much*
higher than for approval or plurality. By casting a ballot in each
system fifty times in a row I think people will get a more balanced
sense of just what a pain ranking is. For the exercise to be meaningful
the candidates need to be presented randomly for each iteration. I'd add
ranking to my (bitrot aflicted) approvalvote site for folks to do
side-by-side comparison but unfortunately easy-to-do ranking is too hard
to implement :) . Can anyone recommend a side-by-side approval and
ranking site? I guess I could implement it similar to how St. Paul has
implemented paper ballots:

http://www.minnpost.com/twocities/2011/10/17/32440/st_paul_ready_to_give_ranked_voting_its_first_try

> Someone quoted James Green-Armytage as saying that Approval is
> "vulnerable to strategic manipulation"  (!)
>  
> What does he mean? (Rhetorical question)
> 
> Approval, like all methods to some degree, of course has strategy. But
> calling strategy "manipulation" implies that the strategy
> is tampering with and changing a rightful result. What rightful
> result, in Approval, is being tampered with by strategy.
> 
> Strategy is merely the determination of how to vote, to maximize your
> expectation. Some kinds of elections require strategy
> in Approval, but to me, our current public political elections don't
> require any strategy decisions, other than "vote for acceptable
> candidates and don't vote for the entirely unacceptable ones."
> 
>  "Vulnerable to" implies the same thing.
> 
> Those words imply an unsupported assumption.
> 
>  If James is here, maybe he'd like to explain what he means.
> 
> ----
> Election-Methods mailing list - see http://electorama.com/em for list info