[EM] Re: Blake's margins arguments

[MIKE OSSIPOFF]

I replied to this message some days ago, and sent my
reply in, but maybe I sent it to the old list address,
because it isn't in the archives:

I'd said:

>Majority matters because it's a group of people whose need for defensive 
>strategy can be minimized to a degree qualitatively better than
>can be said for a submajority group of people. As described by the
>definitions of the majority defensive strategy criteria.

Blake replied:

That's certainly a unique way of defending majority rule.

It shouldn't be surprising if the wishes of most of
the voters can be important in more than 1 way. I
doubt that my majority arguments here are "unique", since
there are others here who value the voted preferences
of most of the voters, and who agree with the goal
of minimizing need for defensive strategy.

But, though majority rule matters with regard to
defensive strategy need & SU, maybe majority rule is
mostly a fundamental standard, one that many people value for its own sake.

As such, it really doesn't make sense to argue with it.
Either you consider important a voted preference of
most of the voters, or you don't. If Blake doesn't, that's ok. Largely, 
Blake is fighting about issues on
which I don't really dispute him.

Blake continued:

> > That's a strategic, not a majoritarian argument.
>
>You wanted a majoritarian argument to justify majority?

Here's my point. Sometimes one of your arguments for wv is that it
protects the only true majorities

I reply:

Blake is the only one who keeps referring to something
that he calls "true majorities". He's still playing
"dueling majorities". Blake's "majority" is a pairwise
defeat. No one disputes the imporance of pairwise
defeats. But the problem is that sometimes it's impossible to honor all of 
them, when there's no unbeatn
candidate. Then, unavoidably, we must elect a candidate
who has a pairwise defeat against him, who has Blake's
"majority" against him. Then, it's reasonable to compare these "majorities". 
What if one candidate has
a defeat voted for by most of the voters, and the
others don't?

But I re-emphasize that it isn't a question of "should".
Some consider it important when most of the voters
vote a certain same preference. Some want to minimize
defensive strategy need. No one's saying that Blake
should agree that those are important, and so it isn't
quite clear what Blake's arguing against.

Blake continued:

, and that since everybody (or at least
a lot of people) believes in majority rule, we should use wv.

I reply:

No, I don't say what Blake should use, or what he
should propose to committees of which he's a member.

I do suggest that if someone cares about a voted
preference of most of the voters, or about the goal
of minimizing the need for defensive strategy, then
that person would prefer wv to margins.

But I say nothing about what Blake should believe or
use.

Blake continued:

This is
what I call the majoritarian argument

I reply:

Well, the argument wasn't quite as you quoted it.
Yes, lots of people value a preference voted by most
of the voters. You'll have a difficult time selling
margins to them, because margins so often violates,
ignores, erases such majority information. But that's
your business, your problem, and not something for me
to argue against.

Blake continued:

Arguments about how wv is good at
preventing strategic voting are a different kind of argument.

I reply:

You keep saying that, but it isn't about preventing
strategic voting. It's about minimizing the need for
defensive strategic voting. Defensive strategy means
voting in a way that doesn't express all of your sincere
pairwise preferences, and in a way intended to protect
the win of a CW or to protect majority rule (as I defined
majority rule on EM).

Sure, wv deters offensive strategy, but avoiding or
minimizing the need for defensive strategy is what
I & some others have been emphasizing.

Blake continued:

I feel I
dealt with this adequately in the page I originally linked to.

I reply:

Blake has a right to feel like that. I've replied to
his website arguments about strategic voting. Basically,
Blake said that wv too can have strategy incentive.
Of course. Gibbard & Satterthwaite showed that all
nonprobabilisitc methods must have situations with
strategy incentive. But, if we consider the need for
defensive strategy, then we find a big difference between
wv & margins. Margins is the method that does poorly
in that regard.

Blake continued:

I agree that the pairwise preferences of a majority group matter. I
even agree that the pairwise preferences of a majority group, as you
define it, matter. I just don't agree that the majorities as you define
them should always take precedence over other majorities, as I define
them.

I reply:

It depends on what criteria and standards one considers
important. You value, more than I do, the number of
losing voters, voters who lost the public vote on
whether candidate X is better than Y. You believe that
they're important enough to overrule the fact that
most of the voters indicated that they'd rather elect
X than Y. Again, I don't argue with you on that. You
should be able to believe or value what you want to.

Blake continued:

Furthermore, "can't be good for social utility" isn't much of an
argument. Personally, I don't think it can be good for social utility
to favour a lower margin majority over a higher margin of majority. So
there!

I reply:

Your margins are meaningless for SU unless voting is
sincere. And if voting is sincere and we're judging by SU, then Borda is the
best rank-count, and CR is the best method. So there!
Maybe insincerity won't be as rampant in margins as in
Borda, but margins has more need for it than wv does.

You're claiming SU advantages, manifested only under
sincere voting, for a method that avoidably creates
a relatively great need for insincere voting.

I'd said:

MMC doesn't apply
>generally to the classic lesser-of-2-evils problem example.

Blake replied:

You have a point, but as I understand it, the lesser-of-2-evils problem
isn't really solvable in a general case. So, its really an argument
about what special cases you want to solve it for.

I reply:

No, it's also a question of how general you want the
LO2E guarantees to be, how drastic you want the strategies to be, and how 
necessary you want the strategy to be to the voter. Margins has no LO2E 
guarantees
that compare in their generality or power to the
defensive strategy criteria. MMC is met by Margins, wv
, & IRV. We can do a lot better than that.

As I said, your strategy criticisms of wv aren't
about defensive strategy need. They're about strategy
incentive, something that's unavoidable for nonprobabilisitic methods. 
Defensive strategy need is
where margins does so poorly in comparison to wv.

I'd said:

It's often said that Plurality fails Condorcet
>Loser. No it doesn't, when that criterion is defined in terms of
>actual votes.

We've had this discussion before. I could define all the criteria in
terms of sincere votes, but this would rely on a precise definition of
what constitutes a sincere vote

I reply:

I've posted such a definition here.

Blake continued:

, possibly on a method by method basis.

I reply:

No, my definition is universally applicable. I don't
know if you could come up with an unproposable method
that it wouldn't apply to intuitively, but it applies
as expected to all proposed methods.

A voter votes sincerely if s/he doesn't falsify a preference, or fail to 
vote a sincere preference that
the balloting system in use would allow him/her to vote
in addition to the preferences that s/he actually did
vote.

(A preference is a pairwise preference)

[end of definition]

Blake continued:

Also, it would make my definitions much different from those found in
academic literature.

I reply:

And we don't want that! :-)

Mike Ossipoff




_________________________________________________________________
MSN 8 with e-mail virus protection service: 2 months FREE*  
http://join.msn.com/?page=features/virus