[EM] Sincere methods

[Kevin Venzke]

Juho,

 --- Juho Laatu <juho4880 at yahoo.co.uk> a écrit : 
> On Apr 4, 2005, at 23:38, Kevin Venzke wrote:
> 
> > Maybe you have in an earlier post argued that majority rule (in this 
> > sense)
> > is not as necessary in a "sincere method." But I doubt I can be 
> > convinced
> > of that.
> 
> I'm not sure if I understood all you wrote, but anyway, if one 
> sincerely thinks that majority rules should be respected, then 
> corresponding sincere methods shall reflect that wish. Majority 
> defending modifications should be excluded but sincere majority needs 
> kept.

I don't see a difference. What you call "majority defending modifications"
in e.g. winning votes is nothing more, nothing less than the use of a
defeat strength measure that inherently views majority-strength defeats
as being stronger than sub-majority-strength defeats.

Suppose I think it's important to have peanut butter rather than jelly
on my toast, so I decide that instead of using jelly, I will use peanut
butter. Based on this discussion, I might expect you to claim that peanut
butter toast is a form of jelly toast, but with strategic modifications
designed to defend people who prefer peanut butter.

> > WV is more likely
> > to give an approval-ish result than margins is. I'm pretty sure 
> > simulations
> > could prove this easily.
> 
> May be, but the "ish" at approval-ish probably already means that we 
> are not talking about our sincere target method but just about an 
> approximation of it.

The example used below:
49 A
24 B
27 C>B

WV and Approval elect B, and this isn't a coincidence: both methods 
consider the actual number of voters voting a certain way.

> > Does it matter which measure is "more natural"? Does that affect 
> > whether a
> > method can be a "sincere" one?
> 
> Yes. I believe people think that that their sincere methods are 
> natural, and on the other hand natural looking methods could be 
> someone's sincere methods. Easy to understand strategy defence 
> mechanisms can also be said to be natural (although they are not 
> sincere). So, for all x, SincereMethod(x) => Natural(x)

Well, I'm afraid I don't understand either how margins is "natural" or
how being "natural" suggests that something could be used as an SVM.

> > Do you have a favorite example in which, when all abstention in 
> > pairwise
> > contests is informed and deliberate, margins produces a "more ideal" 
> > result
> > than WV?
> 
> I don't have any such favourite at hand. I guess I would favour margins 
> in any example where the results differ. Maybe you have some 
> interesting example in mind.

The 49 24 27 scenario is good enough.

> >> Yes, most notably term "sincere criteria" excludes all criteria whose
> >> target is to fight against strategies. Sincere criteria aim at 
> >> electing
> >> the best candidate and nothing more.
> >
> > I'm not sure this is well-defined. I can imagine interpreting 
> > Condorcet,
> > or the Majority criterion for solid coalitions, as anti-strategy 
> > criteria.
> 
> If some criterion has both sincere and strategy defence based 
> reasoning, then I'm happy to call it a sincere criterion.

What I mean is that sincere criteria almost always have a strategic
justification.

> P.S. I'm not familiar with "Majority criterion for solid coalitions"

That is the one that says if more than half of the voters prefer every 
candidate in some set of candidates to every other candidate, then the winner 
must come from that set.

> >>> SVM: Schulze (wv), PVM: MinMax (pairwise opposition) and CDTT methods
> >>
> >> Schulze (wv) is to me a good PVM but I haven't considered it to be a
> >> SVM (since I believe many of its features are related to fighting
> >> against strategies, not to electing the ideal winner (with sincere
> >> votes)).
> >
> > I still haven't understood what you mean by "ideal winner."
> 
> I haven't defined that term, sorry. I note that I have used "ideal 
> winner" pretty much as a synonym for "winner of a SVM".

If that's what "ideal winner" means, then your statement above that you
"believe many of [Schulze(wv)'s] features are related to fighting against
strategies, not to electing the ideal winner" doesn't make any sense,
since my claim is that Schulze(wv) is a good SVM.

Maybe you think "SVM" is already well-defined, but you've suggested that
approval and margins could be used, and those aren't similar.

> > I consider Schulze(wv) to be a good SVM.
> 
> But you include also some strategical concerns in the justifying text. 
> => In my mind when defining a SVM you should try to describe the ideal 
> results without considering strategies. (I hope my definition of ideal 
> winner clarified my use of term "ideal".)

The more I talk with you, the more I am convinced that there is no way to
justify a method except in strategic terms. The only justifications you have
used so far are "naturalness" and a kind of "immunity to mutiny" (which I
don't think counts).

> > 49 A
> > 24 B
> > 27 C>B
> >
> > I say the most intuitive winner is B. It's true that I have some 
> > strategy
> > concerns in mind.
> 
> A seems to be the minmax (margins) winner (= my default favourite :-). 
> Do you see B as a sincere winner too? For what reasons? Not being 
> ranked at last place in majority of votes?

I see B as a sincere winner. A can't win because this would be contrary
to majority rule (in any way I've defined it). C can't win because by any
measure, C seems to have done worse than A.

You haven't offered an SVM choice that would preserve majority rule, in my
sense.

> > I want to minimize voter regret, and get candidates an accurate number 
> > of votes.
> 
> These sound more like sincere wishes about how a SVM should work.

No... "Minimize voter regret" is a strategy concern, and "get candidates an
accurate number of votes" is already done if voters are not using strategy.

If you think "Minimize voter regret" is not a strategy concern, then I note
that in the above scenario, if A is elected, then the C>B voters will regret
that they ranked C, since surely if they had just voted "B," B would have
been elected.

And of course if they don't vote for C, then C is not getting an "accurate
number of votes."

> > Unfortunately these MinMax methods aren't clone-independent. Surely the
> > "ideal method for sincere voters" would be clone-independent?
> 
> I'm not quite sure if clone independence is a strategical defence 
> criterion or a sincere criterion. Clones may also sometimes be real 
> clones and sometimes not. Any good example in your mind for me to 
> comment?

Obviously it is a strategy criterion; it's just about nomination strategy
rather than voting strategy.

If you really are not convinced that a sincere method should be independent
of clones, I won't argue with you, but that makes me doubt how "good" the
method is, even given sincere voters.

> You caught me a bit unprepared in this. I'll try to study more to be 
> able to give better answers later. I have some favourite techniques for 
> handling clones but I'll come back when my thoughts are stable.

Ok.

> >> (Note that reason why I fear that sometimes strategy defence examples
> >> could be misused is that one can claim that some method gives correct
> >> result despite of certain strategic votes, but in this case the same
> >> votes could be as well a result of sincere opinions, in which case 
> >> they
> >> should of course not be corrected.)
> >
> > If they're the result of sincere opinions, then there's no problem in
> > "correcting" them.
> 
> I didn't get this. Did you say that sincere winner could be replaced 
> with someone else? (="corrected")

I'm saying that strategy defense mechanisms don't operate by detecting 
insincere votes.

Kevin Venzke



	

	
		
__________________________________________________________________
Découvrez le nouveau Yahoo! Mail : 250 Mo d'espace de stockage pour vos mails ! 
Créez votre Yahoo! Mail sur http://fr.mail.yahoo.com/