[EM] MMPO has incentive for reversa l, but not compromise-reversal.

[MIKE OSSIPOFF]

James--

You said:

    * pro MMPO: 1. LNHarm, 2. zero compromising-reversal incentive

Mike:

I replied (in part):

    his renaming is incorrect.


You say:

I wasn't trying to rename your criterion.

I reply:

No, you're just calling it by a different name :-)

You say:

You're welcome to talk about
FBC. But when I compare methods, I prefer not to talk about FBC, but
rather about compromising-reversal incentive.

I reply:

Then you just go ahead and do that, but understand that, when you talk about 
compromise-reversal incentive,  you're talking about favorite-burial, and 
you're not talking about something that has anything to do with FBC. 
"Compromise reversal incentive" is not another way of saying 
"favorite-burial", and doesn't refer to what FBC is about.

As I pointed out before (apparently with no effect), "compromiise-reversal" 
can be rewarded in MMPO. What isn't rewarded in MMPO is voting someone over 
your favorite. So I'll tell you again what I told you yesterday: 
compromise-reversal does not mean the same as "favorite-burial". But of 
course you can talk about what you want. But when you talk about something 
that I didn't talk about, then that can't be called a reply.

But of course no one says that you have to consider FBC important. Maybe it 
doesn't matter to you if voters have strategic need to bury their favorite. 
If not then of course no one has any reason to expect you to talk about FBC.

By the way, if you know of a method in which no one can ever benefit from 
"compromise-reversal", one would hope that you would tell us about it.

In Approval it's possible to benefit from order-reversal that doesn't 
involve one's favorite. It isn't likely, and it's been argued that its 
likelilhood is low enough that it should be ignored. But in principle the 
order-reversal benefit is there in Approval (and therefore in CR). And, as I 
said, in MMPO.

You continued:

Although the two concepts
are similar, they are not intended to be identical. I made this clear when
I first proposed the terminology.

I reply:

Well, there are methods in which no one can benefit from favorite-burial, 
but there aren't methods in which, even in principle, no one can benefit 
from order-reversal.

I'd said (in two different parts of my posting):

    What could be called "compromising reversal" can profit you in MMPO, if 
the candidate ovewr whom you raise
your compromise isn't your favorite.

...
    Compromising reversal can profit a voter in MMPO, when the candidate 
over whom the compromise is raised isn't one's favorite.



You reply:

That's interesting. Are you sure that there is compromising-reversal
incentive in MMPO? I can't see how. Can you give an example where
compromising-reversal is more effective than compromising-compression?

I reply:

Sure: Say the method is MMPO, and you reverse your 2nd and 3rd choices, 
voting your 3rd choice in 2nd place, and your 2nd choice in 3rd place.

Doing that can't benefit your 3rd choice, because ranking him/her equal to 
your 2nd choice would have been enough to remove from him/her one 
vote-against. But doing that _can_ keep your 2nd choice from beating your 
1st choice.

Say that your 3rd choice is the candidate ranked over your 2nd choice by the 
most voters. By voting your 3rd choice over your 2nd choice, you're adding 
one vote-against to your 2nd choice's MMPO score. If you and one other 
person do that, you could change the winner from your 2nd choice to your 1st 
choice, if your 1st choice would have been the close runner-up, only one 
point behind, had you not done that order-reversal.

You continued:

Definitions:

Compromising strategy:  Insincerely ranking an option higher in order to
decrease the probability that a less-preferred option will win.

I reply:

No objection to that definition, which is clear enough. I just refer instead 
to order-reversal, truncation, and equal ranking, where the order-reversal 
and equal-ranking could, but needn't, be compromise as you define it. I say 
it that way because, if it's compromise,  it makes all the difference which 
kind of compromise it is. I talk about different distinctions than you do. 
And if it's done to protect the win of a CW, or to protect majority rule, 
then I call it defensive. It's just a question of what someone wants to 
express or emphasize, what distinctions they want to express.  As I said, 
they used to say that the Eskimos had a thousand words to distinguish 
between different kinds of snow. Though it wasn't true, it demonstrates the 
point.

You said:

If I am raising D on the ballot to decrease the chance that F will win,
in MMPO, how could order-reversal help this cause in a way that equal
ranking could not?

I reply:

If you're raising D above F, where D would sincerely be below F, that adds 
to the number of people ranking D over F. If D is the candidate ranked over 
F by the most people, then, by so doing, you've increased F's 
unpreferredness score. If someone you like better than F was only one point 
worse than F, and if you and one other person both add to F's 
unpreferredness score, then you change the winner to someone you like 
better, and are rewarded by the order-reversal.

But that isn't compromse, as you define it, because you aren't doing it to 
make D win. You like D less than F, so you wouldn't want to change the 
winner from F to D.

It's reversal, but it isn't compromise-reversal.

You continue:

Example 1: sincere preferences R>S>T>U. Why could R>T>U>S be better than
R>S=T>U in terms of decreasing the chances that U will win?

I reply:

It wouldn't be better for that purpose.

If you want to keep U from winning, why would you raise him/her in your 
ranking? What you're doing in your example is lowering, or "burying" S. That 
can stop S from winning in MMPO.

You continue:

In either
case, I am removing an opposition vote against T

I reply:

Correct.

You continue:

, while maintaining my
opposition vote against U.

I reply:

Not so. When you change to RTUS, you're no longer ranking S over U, and so 
you could be making U win, if you're lowering his worst votes-against. If 
only you do that, you could make a tie that U wins. If you and another voter 
do that same reversal, you could change U from loser to winner.

You continue:

I don't see how adding an opposition vote
against S will help my cause.

I reply:

It will help you if, by adding a vote against S, you (and one other voter 
doing the same thing) can make R win instead of  S, where previously S was 
the winner and R was the close runner-up. Changing from the sincere RSTU to 
the insincere RTUS can accomplish that for you.

You continue:

Example 2: sincere preferences A>B>C>D>E>F>G. Why could A=D>B>C>E>F>G be
better than A=B=C=D>E>F>G in terms of decreasing the chances that F will
win?

I reply:

It wouldn't. You're not changing your pairwise votes against F, in either 
instance.

You continue:

In either case, I am removing my opposition votes against D, while
maintaining my opposition votes against F.

I reply:

Quite so.

You continue:

It's true that the second strategy allows me to keep some opposition
votes against B and C.

I reply:

Yes.

You would do ranking #2 if you wanted to help D. You would do ranking #3 if 
you wanted to help D, B, & C.

Obviously helping D, B, or C could make A lose. So MMPO doesn't completely 
protect your favorite from your strategy, but at least you don't vote 
someone over your favorite.

Likewise in Approval or CR, helping lower choices could make your favorite 
lose, though you don't vote them over your favorite.

You continued:

But if I'm willing to compromise by removing
opposition to D, with the intent of electing D, why would I be unwilling
to remove opposition to candidate whom I like better than D?

I reply:

No _likely_ reason that I know of. If you want to help D win, you should 
also want to help A and B. So your 3rd ranking is better than your 2nd one. 
But that's just my first impression. If, improbably, you knew that if D 
could win, it would be with an electorate that wouldn't make A winnable; but 
that if B or C could win, it would be with an electorate in which A would be 
winnable, then your 2nd ranking could be better than your 3rd one. That 
answers the question quoted above. But that kind of specific detailed 
knowledge about combinations of winnability doesn't sound at all likely, and 
so I'd say that, in practice, there would be no reason to vote your 2nd 
ranking instead of your 3rd one.

You continued:

It seems to me that for any possible compromising-reversal strategy in
MMPO, there is a compromising-compression strategy that is just as
effective. Hence, it seems that MMPO has no compromising-reversal
incentive.

I reply:

Yes, that sounds right. If I earlier said that MMPO can have 
compromising-reversal incentive, then I should have just said that MMPO can 
have reversal incentive. MMPO can have burying reversal incentive, as you 
define "burying": lowering someone in order to keep him/her from winning. 
("him/her" may seem awkward, but the  unacceptable candidates aren't all 
men).

Of course, as I use the term "burying", it just means voting someone lower. 
So I don't add anything to the ordinary meaning of "bury".

Mike Ossipoff

_________________________________________________________________
FREE pop-up blocking with the new MSN Toolbar – get it now! 
http://toolbar.msn.click-url.com/go/onm00200415ave/direct/01/