[EM] Juho: Different answers to your questions. You're right...

[Juho Laatu]

On 27.9.2012, at 9.21, Michael Ossipoff wrote:

> ...about some things.
> 
> But first, regarding some of the other things:
> 
> 1. You seem to imply that you think that there is a single, objective,
> right ideal sincere winner. Of course you'll deny that, but you've
> repeatedly fallaciously based on argument on that assumption.

I deny that I'd think that there is one definition of right ideal sincere winner (since I think different elections, societies and individuals are free to define themselves who is the best, and different elections may well have different requirements). But in each election, with a set of votes (that do not contain exactly similarly or exactly equally well ranked candidates), one (each opinion holder separately) should at least in theory be able to tell which one of the candidates is the best and should be elected if the votes are sincere. There is thus always some understanding of what kind of candidate should be elected, and this philosophy could in principle be defined and presented as a "best sincere winner definition / criterion".

> 
> Sometimes something is a matter of subjective, individual choice,
> without a single objective right answer.

Yes.

> 
> How do you think that you might go about proving that one particular
> winner is the ideal right sincere winner? Proofs and conclusions have
> to be based on some objective premise.

Free choice.

> 
> You're giving us an unsupported assumption.
> 
> 2. You seem to assume that, if there is a single, objective, right
> ideal sincere winner, then it must be found from rank-balloting.

I assumed that the best sincere winner definition will be given together with some assumption of the type of ballots. It would be ok to define the best sincere winner in terms of ratings, although one would choose a ranking based method for use in the elections. There might be various reasons for not using ratings directly in the actual method. Obviously one would pick such a rankings based method that elects as good winners as possible (according to the best sincere winner criterion), and has other good properties like sufficient strategy resistance. The resulting method would not always elect the best winner even if the votes were fully sincere, but it is now a good compromise between different needs (election of good winners, startegy resistance, simplicity etc.).

> 
> Again, that's just your unsupported assumption.
> 
> 3. You seem to think that, for any two desirable method-attributes,
> one of them can only be achieved by something that spoils the other.
> In other words, you think that any pair of desirable method-attributes
> must be mutually incompatible.

In some sense yes. I have mentioned the assumption that methods that have modified to be strategy resistant, usually do not implement any sincere best winner definition that has been determined for sincere votes.

> 
> ...a pair of desirable attributes such as freedom from the worst
> strategy needs, and choice of the unique, objective, right ideal
> sincere winner...or maybe any desirable sincere winner.

Yes.

> 
> In fact, it doesn't occur to you that free-ness from the worst
> strategy needs can result from a result that would be desirable as the
> sincere winner...or the best ideal sincere winner, if there is such a
> thing.

Different needs could sometimes lead to the same answer. It is possible that the sincere winner definition naturally points in a direction that also is strategy resistant (maybe to some extent even typical). I only believe that if the method is tweaked or modified just to be strategy resistant, then one would with high probability deviate from the path of electing the best winner with sincere votes. Tweaking means here making decisions that aim at defending against some identified weaknesses of the method (strategic vulnerabilities or other other problems that are not related to pickng the best sincere winner).

My point thus is that if you first think about sincere votes and who should be elected, and then find a method that implements exactly that definition, and then you abserve some stratgy problems and modify the method so that it can better answer those chanllenges, then the method (by definition) is not any more the method that elected the best winners with sincere votes.

> 
> I've told you why that, in fact is so (but without the assumption
> about there being a single best ideal sincere winner).
> 
> Now: Something that you're right about:
> 
> It made sense to ask me if I prefer any sincere winners or ways of
> choosing them. Yes I do.
> 
> For Official Public Elections:
> 
> Approval. (maybe Score too).

I guess Approval means the candidate that is most approved among the voters (not necessarily the one that gets the highest number of approvals from voters that give approvals to candidates using their best strategy to optimize the result from their point of view). That is, sincre Approval. Same with points in Score.

> 
> I've discussed Approval's unique optimizations. Approving someone is
> actually quite relevant to choosing. I'm referring to procedural
> approval, as in "Yes, I will approve that proposal." Choosing the
> candidate for whom the most voters have chosen to say, "I approve this
> candidate" is an important and valuable optimization.

Yes, this sounds like sincere approval. I agree that in some elections the best sincere winner can be defined as the candidate that is approved by most voters.

> 
> It retains what's good in Plurality, without any of what's undesirable
> about Plurality.
> 
> You can approve those whom you like or trust, or the candidates who
> are acceptable to you.
> 
> Approval, then, is choosing the candidate who is either liked or
> trusted by, or acceptable to, the most people.
> 
> Approval's optimal strategy, in all of its forms, amount to approving
> the above-expectation candidates.
> 
> For that reason, Approval gives the result that will pleasantly
> surprise the most voters (give them a result better than what they
> expected of the election).

For the very same reason these strategic voters may also deviate from electing the most (sincerely) approved candidate. In order to elect always the most approved candidate, the voters would need to approve those candidates that they sincerely approve. If voters use the "above-expectation startegy", then the method (with the given votes) only approximates the ideal result with sincere votes.

> 
> Approval's result confirms and rewards optimism.
> 
> Approving the above-expectation candidates can be regarded as voting
> optimistically, and that optimism is reflected in Approval's result.
> 
> Informational Polling, To Inform Plurality Voting in a Subsequent
> Official Public Election:
> 
> Symmetrical ICT, and its usual sincere choice, the CW (by which I mean
> the legitimately-defined CW).
> 
> As you may know, I advocate only Approval, and maybe Score, for
> official public elections.
> 
> But the CW has special relevance and important for informational
> polling to inform Plurality voting. When there's a CW, that's the best
> candidate that you can get a majority for.
> 
> And, because of its strategic properties, SICT will encourage the
> sincere voting that is needed to ensure the usefulness of that
> informational poll.
> 
> But I'll make this admission: If there's any likelihood of chicken
> dilemma (in Approval), then (if all of the rank-method disadvantages
> and problems were to somehow go away) I might prefer SICT for official
> public elections too, because SICT avoids Approval's chicken dilemma
> nuisance. (_possible_ nuisance, if and when it happens. A nuisance,
> not a problem)
> .
> But there's no reason to believe that rank-balloting will lose its
> problems, at least not anytime soon.

I guess the mathematical properties of the rank-balloting methods will stay there for practically ever :-). It's just a question if they are small enough or too bad (and a question on if the majority orientation of the ranked methods is the correct choice).

> 
> I refer to the count-fraud-vulnerability resulting from
> labor-intensive count, and especially resulting from the need for
> machine balloting and computerized count;

labor-intensive count: Yes, more labor-intensive, but probably not overwhelmingly so. If the ballot is not too labout-intensive for the voters to fill, then the counters can also find the time to digitize the votes. Full counting of the final result is not required.

need for machine balloting: No need. If the ballot is too complex to be filled manually, it may be also too complex to be filled using a machine. One can often also simplify the ballots (without too much damage) if they are about to grow too complex.

computerized count: No problem if the votes are first digitized in a reliable way and there are alternative independet counting processes. (A fully automated process with no paper ballot trail and no visibility to anyone to the given votes would be a problem in any method.)

> and to the apparent
> impossibility of agreement on any one particular of the infinitely
> many rank-counts, for proposal, adoption and enactment; and to the
> fact that it could be difficult (impossible) to demonstrate a rank
> method's FBC compliance to people. Approval's simplicity makes it easy
> to show that there's no need to not approve Favorite.
> 
> So there are sincere winners that are favorite to me.
> 
> By the way, are you trying to say that the chicken dilemma won't be a
> problem? Good, because I say that too. I've been saying it for quite
> some time.

Depends on the method :-) (and on the society too).

> 
> 
> The chicken dilemma isn't a problem. It's a nuisance.
> 
> I've made it clear that the chicken dilemma is the thing that comes
> nearest to being a problem in Approval. Therefore, you don't
> significantly improve on Approval unless you get rid of the chicken
> dilemma. Therefore, don't bother with any rank method that doesn't get
> rid of the chicken dilemma.
> 
> Regarding my 75,51,100 chicken-dilemma example for Dodgson and MinMax(margins):
> 
> You say that both sides should co-operate.

( 75: A>B>>C, 51: B>A>>C, 100: C>>(A=B) )

I say that in Minmax(margins) the B voters have no critical need to rank A second, i.e. no need to co-operate at this intensity level. The B supporters may truncate quite freely, but that may not make any good to them (with close to 100% probability), and that can harm them (with higher probability, e.g. if also some of the A supporters truncate).

> That doesn't mean they
> will. The chicken dilemma is quite common in the animal kingdom. You
> flatter voters too much  you believe that they won't have it too.
> 
> You say that both sides would have the same amount of defection. Why?

I don't say that there is the _same_ amount of defection. I say that there are many kind of voters, with many different kind of opinions, and there are many polls, and many interpretations on what the true final opinion of the voters will be. In the example above A and B are not in quite symmetrical position. If they were, the voting behaviour of their supporters could be quite similar. In this example B voters coud be more eager to truncate (for the possible strategy related thoughts that you mentioned, although the stratgy is quite irrational), but one can not exclude the possibility that some A voters would truncate. Actually that is quite certain. Check for example the number of voters that truncted in Burlington. One can truncate also just because one is lazy or does not want to take position on any others that one's favourite. In the example above, A supporters could truncate more than B supporters because they may think that A is clearly ahead of B, and therefore there is no need to support B (who is not likely to win anyway).

> 
> For one thing, the point of the example was to show that defectors can
> easily take advantage of co-operators in Dodgson and MinMax(margins).
> The example shows that.

I hope you read my comments :-).

> 
> You say that real national elections are more complicated. No sh*t !  :-)
> 
> Perhaps you hadn't heard that examples are necessarily
> simplifications, for the purpose of showing that something can happen.

That's fine. But we must remember also that in real elections the opinions and votes will be more complex (and will change in time, and they are not easy to control).

> 
> No one claims that an example is really the exact party, candidate,
> and faction configuration of an actual presidential election :-)
> 
> You think that both factions would have the same number of defectors.
> But voters who prefer policies that are more violent, and less humane,
> ethical--voters who are more likely to tolerate, or even applaud,
> criminality in their political leaders--Do you really think that those
> voters will be just as conscientious, honest, co-operative and
> responsible as voters who are not like them?

There are almost as many kind of voters as there are humans.

> 
> ...and likewise their party organizations and campaign organizations?
> 
> Will the chicken dilemma situation always occur? Will it always occur
> on a scale sufficient to change the outcome?  Of course not. You're
> right.
> 
> Remember that I've long been saying that the chicken dilemma is a
> nuisance, not a problem. It's a nuisance that often won't happen (I've
> several times posted a list of reasons why it usually won't be a
> problem). And can be easily dealt with if it does happen, via SFR.
> 
> But that doesn't excuse Dodgson and MinMax(margins) from failing to
> get rid of that nuisance that can easily be avoided with a good rank
> method.
> 
> You're still saying that choosing the candidate who can be made into
> CW adding (or disregarding, or reversing) the fewest pairwise votes is
> MinMax(margins), and not Dodgson.  Fine.

Yes, adding new votes where the candidate in question is ranked first (disregard and reversal are not exact matches).

> 
> I've explained that to you already. If you still remain confused about
> the difference between Dodgson and MinMax(margins), then I',m not
> going to repeat the explanation for you.
> 
> Some people would look those two method names up, at electowiki or a
> search-engine, instead of just continuing to repeat their belief--but
> not you.

Or Wikipedia. We'll both keep our own beliefs. I'm almost fine with that deal. :-)

Juho


> 
> Oh yes, the definition of "pairwise vote":
> 
> A pairwise vote for X over Y is an instance of someone ranking X over Y.
> 
> Mike Ossipoff
> ----
> Election-Methods mailing list - see http://electorama.com/em for list info