[Election-Methods] USING Condorcet

[Kevin Venzke]

Hi Rob,

--- En date de : Jeu 3.7.08, rob brown <rob at karmatics.com> a écrit :
> > Hi Rob,
> >
> > By the way, I'm not complaining about Condorcet
> here.
> >
> > I guess you're talking about the camp that
> believes that voters will use
> > burial strategy for no reason at all. I can't say
> that they wouldn't,
> > but that isn't my position.
> 
> Ok, good to know.
> 
> > > Maybe I misunderstood your statement, but it
> seems to say
> > > "even if it is an
> > > unwise strategy, the fact that people may do it
> anyway is a
> > > matter of
> > > concern".  Given that:
> >
> > No, that's not what I meant to imply. There's
> a difference between a
> > strategy that is unwise and a strategy that may be
> wise but happens to
> > backfire.
> >
> 
> Well that I'm confused then.  If there is a significant
> chance of it
> backfiring, doesn't that make it less wise?

Yes. My main point is really that we should make it as likely to backfire
as possible so that nobody thinks it's a good idea.

What I was trying to say to Juho is that if people think the strategy is
wise, it isn't just a problem if the strategy works. It's a problem also
when it fails badly. And it's at least somewhat undesirable for it to be 
a decision for the voter.

> For instance, if someone offers you $1000 to play one round
> of russian
> roulette, this can be distilled down further than "5
> out of 6 times it is a
> wise move, the other sixth of the time it is unwise." 

With 5/6 odds it is too unwise for anyone to try. No problem there. The
problem is when you improve those odds to the point where some people
start taking the gamble and a few end up in the morgue.

My point was that if you don't approve of this game, it isn't just a
problem when people win at it. There's no consolation in people losing
their lives for playing, if they're using sound logic to try. The mere
fact that you can lose at a game doesn't mean the game is stupid.

> > > a)  a prisoner's dilemma situation:
> >
> > I don't understand in what context you want to be
> concerned about a
> > prisoner's dilemma? Is this for election methods
> in general?
> 
> Yes.  I am using it to represent something that is in an
> (independently-operating) individual's self interest,
> but when everyone does
> it, everyone is hurt.  Really anything that is
> "gameable" or "strategically
> manipulable" falls under this category.  What does
> *not* fall under this is
> behavior which is non-rational, or altruistic.

Does this mean that if all voters are rational (and not altruistic or
motivated by some other principle), you can label all strategy incentive
problems as a prisoner's dilemma? I'm having trouble imagining that.

> > > b) a situation where people's self interest
> is in
> > > conflict with some other
> > > goal, in particular, their sense of ethics.
> > >
> > > I don't see your scenario -- voting
> > > strategically/selfishly but in an unwise
> > > way -- to fit into either a or b, so I see it as
> issue of
> > > minor concern, at
> > > least in the long term.
> >
> > If people rank the worse of two frontrunners strictly
> last, just because
> > that "feels right," I would categorize that
> under B. But this isn't my
> > scenario.
> 
> Hmm, I don't think I would, I would classify that as
> non-rational,
> especially if people are educated that it isn't usually
> in your benefit to
> do rank them insincerely.  But ok.

I'm not talking about uninformed voters though. I think it's perfectly
reasonable to bury the worse frontrunner as a matter of principle. I
doubt I would ever care enough to do that, but I think it's possible that
someone would. It's similar to a protest vote, in expressing contempt.

> > The scenario is more like "chicken." If I
> think you will be sincere, then
> > I should bury your candidate. If I think you're
> going to bury my candidate,
> > then (if I only care about who wins) I should vote
> sincerely. Or else I
> > can be stubborn and bury your candidate, thereby
> refusing to let you bury
> > my candidate and get away with it. When we both bury,
> then we crash and
> > elect the worst candidate.
> 
> Again, I think with chicken you can distill it down to
> either unwise or wise
> based on the your guess at the probability of the other
> party's move.

Sure.

> Chicken is a game specifically contrived to be unstable, of
> course (that is
> what makes it interesting), the fact that there are only
> two people playing,
> and the all-or-nothing nature of the outcome, creates and
> amplifies that
> instability.  The more people playing, the more stable and
> predictable it
> becomes -- i.e. it becomes more of an equilibrium rather
> than a feedback
> loop.

Great. Where will the equilibrium be?

> For instance, imagine playing chicken against 50 other
> people (maybe using
> some machine that can give an electric shock?), where if
> some percentage of
> them chicken out and some percentage don't, the level
> of injury and rewards
> to those who don't chicken out will be an equivalent
> proportion of maximum.
> What might be expected is that an equilibrium is found,
> where a few (those
> who are the most competitive and/or tolerant of pain) will
> choose not to
> chicken out, and the level that they are injured and
> rewarded approximately
> balances out for them, so that they don't regret their
> decision. As opposed
> to everyone not chickening out and all ending up in the
> hospital, or
> everyone chickening out and feeling like they missed an
> opportunity to win.
> 
> I suppose if you have an election with a tiny number of
> voters, it can
> become less stable and more "chicken like", but
> even with two voters, I am
> not convinced that burial would often be wise unless there
> is an assumption
> the other side is behaving non-rationally.

I'm not sure what you mean by "non-rationally" in this context. Whether
you should swerve depends on what the opponent will choose to do, not on 
whether he's rational.

Rationally if they would hate to have a car accident worse than anything
else, and they are rational, then burial (by you) could be wise, 
particularly if you don't mind the car accident as much (perhaps you are driving a large truck).

> The point is that what we should be looking at with
> condorcet is whether,
> when all the probabilities and utilities are calculated
> out, is insincere
> burial actually a truly wise strategy for a significant
> number of voters?

My point is that it depends, but can cheaply be made unwise virtually
all the time. The only thing you give up is the novelty of being able
to say that voters can and should rank everybody they can. Personally I'd
rather have the method just work well, than have maximum expressiveness.

Kevin Venzke


      _____________________________________________________________________________ 
Envoyez avec Yahoo! Mail. Une boite mail plus intelligente http://mail.yahoo.fr