Winning votes intuitive?

[Rob LeGrand]

Mike wrote:
> Ok, winning in violation of majority rule, by truncation, in the way that
> can happen in margins methods--how will that happen in wv methods
> by flipping a coin?

Say a voter's sincere vote is A>B>C and he could gain some advantage under
margins by insincerely voting A>B=C.  Under winning-votes, all he has to do is
flip a coin.  If it's heads, he votes A>B>C; if it's tails, he votes A>C>B. 
Doing this gives him exactly the same expected outcome as would voting A>B=C
under margins.  Besides, generally, when voting A>B=C would help, voting A>C>B
would help even more, which would work just fine under winning-votes.  As Blake
has put it before, winning-votes is "easy to get around".  Now one might say
that voters will be too dumb to flip a coin or too principled to engage in
strategic reversal.  Fair enough.  If I were sure of that, I'd probably support
winning-votes too, despite its lower social utility.  But like I've said, I
prefer to be pessimistic and realize that voters will likely catch on.

> When you said that wv doesn't guarantee anything that margins
> doesn't guarantee, I told you of some guarantees, which I call
> SFC & GSFC.

Yes, and Blake and I have explained patiently why, in our opinions, those
guarantees are ultimately useless, even if they sound attractive.

> In CSSD(wv), BeatpathWinner(wv), or Ranked-Pairs(wv),
> if X is in the sincere Smith set, and Y is not, and if a majority
> prefer X to Y, and vote sincerely, and if no one falsifies a preference,
> then Y can't win. Not even if someone flips a coin :-)

Sure, if no one "falsifies a preference".  Which means it's okay for them to
vote insincere ties (like A>B=C) but not to reverse (like A>C>B).  But I'd say
voting A>B=C is insincere just as A>C>B is, if a little less extreme. 
According to your definition, voting A>B=C isn't falsifying a preference, but
voting 50% A>B>C and 50% A>C>B is.  And what if every voter is willing to vote
strategically (which is obviously the kind of case I'm most interested in)? 
Your guarantees don't mention that case at all.

> Yes, I know that you think that, with wv, voters who'd otherwise
> have only truncated would order-reverse. Even if the necessary number
> of people were inclined to do that, which is really doubtful, it
> would carry a great danger of backfiring.

Perhaps, but strategic reversal is no more dangerous in winning-votes than in
margins, so that's also an argument for not worrying about strategic reversal
under margins.  I personally would vote sincerely in a Condorcet election
unless I could somehow be *really* sure that a reversal would be to my
advantage (except maybe for voting high ties if a winning-votes method were
used).

> If the insincere voting that you're referring to is ranking in
> different rank positions candidates whom you sincerely rate equally,
> do you really think that's important?

That's not what I actually had in mind, but sure, I think it's important.  I
see voting preferences you don't have as insincere.  Having said that, unless
there are two candidates who are literally clones of each other and exact in
every way, I think the only possible reason for not preferring one to the other
is laziness.

> Rob LG, if you'll check the introductory web-page for EM, it points
> out that it isn't productive to repeat refuted arguments. We've been
> over all this before.

Obviously I don't believe that my arguments have been refuted.  I think Blake
and I have done a good job refuting your arguments.  But in the end it comes
down to assumptions we prefer to make.  I don't see truncation as a case of
insincere voting to be treated specially, and I don't see winning-votes's
handling of tied preferences as doing any good if the voters are informed, as
I've explained.

> Yes, it isn't an offensive strategy that works. It won't steal
> the election from a sincere CW, or violate majority rule in the
> way that I've pointed out that margins will.

Again, winning-votes's guarantees only hold when no voter "falsifies a
preference".  Pretty fragile.

> As for flipping coins, that of course would most likely have no
> effect in a public election, as the coin-flips and the resulting
> AB votes would tend to cancel out.

Exactly.  Under winning-votes, a bunch of voters voting half A>B>C and half
A>C>B will get the same result as if they had all voted A>B=C under margins. 
If they benefit from the insincerity, then winning-votes's guarantees are
rendered ineffective.  Winning-votes works only when voters are stupid enough
not to realize that they have just as many strategic options as under margins;
in fact, they have more, since margins allows no equivalent to winning-votes's
A=B>C.

> "Every method has problems, and so we shouldn't try to minimize
> the magnitude of the problems"--That sounds like what we hear
> from our IRV friends.

That isn't at all what I said.  Of course I think we should minimize problems
created by strategy.  I pointed out how winning-votes shares every single
strategic problem that margins has when voters aren't ignorant, and noted that
*any* ranked-ballot method will have serious strategy problems when the
electorate is well-informed and there is no sincere Condorcet winner.  All we
can do is to try to minimize them, which is difficult.

Mike is comforted by the fact that winning-votes methods make it easier to
protect a sincere Condorcet winner when no voter reverses strategically.  I'm
not.  He wants to prevent offensive "truncation".  But as I've explained, a
voter could make his offensive "truncation" just as effective
(game-theoretically equivalent) under winning-votes as under margins, whether
by randomization or coordination.  If you don't agree, if winning-votes makes
guarantees that are still important to you, then you're right to prefer
winning-votes.  I don't say it's wrong.  But if you agree with me that
winning-votes is actually "easy to get around", then margins makes more sense,
being better at social utility and arguably more intuitive for the voter.

I don't expect to convince Mike.  If I'd done all the work he has on Condorcet
methods, I'd be proud too, and I have the highest respect for his objective
approach to evaluating methods.  I've just tried to explain how *my* reasoning
leads me to reject winning-votes as (as we say in Texas) all hat and no cattle.

--
Rob LeGrand
honky98 at aggies.org
http://www.onr.com/user/honky98/rbvote/calc.html

__________________________________________________
Do You Yahoo!?
Yahoo! Tax Center - online filing with TurboTax
http://http://taxes.yahoo.com/