[EM] James--CP, AERLO, ATLO, defensive strategy definiltion

[MIKE OSSIPOFF]

I'd said:

>Fair enough, stating a first impression, but it´s also fair to ask you
>why you doubt that AERLO & ATLO will be as effective and easily
strategy-resistant as Cardinal Pairwise.


You repied:

	I did explain this a bit already, as you might have seen later in the
message, in this paragraph:
>Also, I believe
>that cardinal pairwise achieves many of its counterstrategic benefits
>without any need for conscious counterstrategizing by the voters.
>Lots of voters giving Bush a very low rating and Kerry a very high rating
>makes it
>very hard to overrule a Kerry>Bush defeat, but the voters don't all need
>to understand majority rule cycles, or anticipate a particular strategic
>incursion, in order to guard against an incursion that attempts to
>overrule the defeat. The neat thing about cardinal pairwise is that rating
>the candidates intuitively is usually a good protection against strategy.
>AERLO and ATLO don't have this benefit, because the concept of truncation
>lines and equal ranking lines is never going to be as intuitive to voters
>as ranking the candidates on a scale from 0 to 100.

I reply:

Sure, and I agreed with that. But I then ponted out that to achieve certain
guarantees, one must vote strategically, with either system. AERLO & ATLO 
merely
ask for a line to demark the candidates one wants to protect. Strategy, but
obvious and natural strategy. With CP you can get the same result by
strategicallly using your point assignments. Explicit stategy, however, is
easier strategy. How to use that line in your ranking to protect someone is
obvious, though of course it would
be easy to tell someone how to accomplish that with CP.
.
But I re-emphasize that I have no quarrel with CP. Either CP or AERLO & ATLO
would be a powerful strategy-need-reducing enhancement for wv, and I'd just
about just as soon have one as the other.

Just as with the RP vs SSD or BeatpathWinner/CSSD dispute, I claim that 
there's
no significant merit difference, and therefore no genuine dispute or
disagreement.

You wrote:

I think that both AERLO/ATLO line placement and cardinal pairwise rating
assignment should be understood to incorporate strategic thinking, rather
than being "purely" derived from voters' sincere attitudes toward the
candidates outside of the context of specific poll results.

I reply:

True. But voters who know what they're doing won't base their voting on poll
results, unless the polls are conducted, counted, and reported by an
organization that they have good reason to trust. That can't be said for the
pollsters now.

You wrote:

	My definition of "burying" is a slightly modified version of Blake
Cretney's definition. Quoting from my paper:
	"Insincerely ranking an option lower in order to increase the probability
that a more-preferred option will win.

I reply:

Ok, just wondered. That could be done offensively or as a deterrent to 
offensive
strategy, and I make that distinction. But I don't criticize those who 
don't.

You continued:

	"Compromising", on the other hand, I define as follows:
	"Insincerely ranking an option higher in order to decrease the
probability that a less preferred option will win. For example, if my
sincere preferences are R>S>T, a compromising strategy would be to vote
S>R>T or R=S>T, raising S’s ranking in order to decrease T’s chances of
winning. (The drawback is that this often decreases R’s chances of winning
as well.)"

I reply:

So compromising could be order-reversal or equal ranking. I make that 
distinction, but I
don't say that everyone should.

You continue:


	I've never fully understood the difference between your terms "offensive"
and "defensive".

Someone once told me that a strategy is offensive if it
seeks to elect someone other than the CW, but I feel that this is a
somewhat awkward way to look at things, because the CW/Smith set is not
always known before the election.

I reply:

Of course, but how is that an awkward way to look at things? Who says that
strategy must be based on perfect, complete, and reliable informaiton?

You strategize because of what you believe to be so, and your information 
may
well be mistaken.

I define defensive strategy as strategy intended to protect the win of a CW, 
or
to protect majority wishes.

(Yes, I've been saying "majority rule", but some time ago I replaced that 
with
majority wishes, for the purpose of defensive strategy, and stated that 
change on EM).

If you vote for Kerry in Plurality, though you prefer Nader, believing
that the Nader voters and the Kerry voters constitute a majority who prefer 
Kerry to Bush, or you believe tha Kerry is CW, then you're voting defensive 
strategy.

If you object that there's too much subjectiveness in that definition, I 
point out that your definitions of burying and compromise both describe 
wishes and intents on the part of the voter, and his belief that what he 
wants might not happen if he doesn't strategize, but would be more likely if 
he did strategize.

If you vote for Kerry instead of Nader, without expecting Kerry + Nader to 
be a majority, or expecting Kerry to be CW, but just in the hopes that 
you'll help Kerry somehow get lucky and win, that isn't defensive strategy, 
as I define it, though, by your and Blake's definition of compromise, you're 
doing exactly the same compromise strategy in both of those situations.

I told Markus that I was going to define majority rule soon. My definition 
of
majority wishes is similar, and I guess that I'd better state that 
definition
now, instead of being vague about what I mean by majority wishes and 
majority
rule.

If a majority prefer X to Y, that's a majority pairwise preference (MPP).
The strength of that MPP is measured by the number of voters who prefer X to 
Y.

An MPP for X over Y is outdone if there is a sequence of MPPs from Y to X,
consisting of MPPs that are all at least as strong as the MPP of X over Y.

To violate majority wishes means to elect someone who has an MPP against him
that isn't outdone.

Protecting majority wishes means avoiding a violation of majority wishes.

[end of definition of protecting majority wishes]

To repeat my definition of defensive strategy:

Defensive strategy is strategy intended to protect the win of a CW, or to
protect majority wishes.

[end of defensive strategy definition]

Majority rule:

X has a majority pairwise vote against Y if a majority vote X over Y.

Substituting majority pairwise vote for majority pairwise preference in the
definitions above leads to a definiltion of majority rule instead of 
majority
wishes.

Well, it's better to say it explicitly:

X has a majority pairwise vote (MPV) against Y if a majority vote X over Y.

An MPV's strength is measured by the number of people who vote X over Y.

An MPV for X over Y is outdone if there's a sequence of MPVs from Y to X
consisting of MPVs that are all at least as strong as the one for X over Y.

Violating majority rule means electing someone who has an MPV against him 
that
isn't outdone.

[end of definition of violating majority rule]

I define offensive strategy as strategy intended to take victory from a CW, 
or
to elect someone in violation of majority rule.

[end of definition of offensive strategy]

I admit that that definition is about votes (the majority rule part) and 
about preference (the CW part), and that, because the offensive strategy 
definition refers to votes and the defensive strategy definition doesn't, my 
offensive strategy definition isn't the exact opposite of my defensive 
strategy definition. But they're what I mean by defensive and offensive 
strategy.

Now, if someone finds fault with my definition of offensive strategy, that 
will
confirm my feeling that I wasn't entirely satisfied with it when I wrote it, 
a
long time ago.

Of course strategy to improve your utility expectation might be neither
offensive nor defensive.

Anyway, I told Markus that I'd post a definition of majority rule, one that 
is
similar to something that he has posted. The difference between majority 
rule
and Markus's Beatpath GMC is that my majority rule definition speaks of the
strength of the majority pairwise votes in a sequence from Y to X, while
Beatpath GMC only requires that there be a sequence of majority pairwise 
votes
from Y to X.

I also told Markus that I'd post a new definition of majority-rejected:

A candidate is majority-rejected if someone has a majority pairwise vote 
against
him that is not outdone.

[end of majority-rejected definition]

That term, majority-rejected, isn't used by me now, but I wanted to define 
it
anyway, because Markus was so interested in that term.


I'd said:

>
>Well, several of us told you why Condorcet wv won´t have a strategy
>problem. Every suggestion that you made about a Condorcet strategy problem 
>was
answered by me and others.

You replied:

	I don't think that this is true.

I reply:

Well, we spoke of why, with wv,  order-reversal isn't going to be a problem 
in practice.

Obviouslsy you and I can disagree on whether that's so.

You said that if offensive order-reversal ever successfully stole an 
election, then people would throw-out Condorcet's method. But, before they 
do, I (or whoever is alive in that century) should point out to them that 
Plurality and IRV, without any offensive strategy being attempted, have a 
much worse defensive strategy problem than offensive order-reversal can 
cause in wv*, and that Plurality and IRV will routinely commit violations as 
bad as the one that offensive strategizers could achieve by their 
well-deterred strategy.

* because prevention in wv doesn't require the favorite-burial that it 
requires in Plurality and IRV, while the creation of the problem requires 
risky, well-deterred offensive strategy in wv.
I'd said:

>Of course. But when we make statements, we can fairly be asked to justify 
>them.
And when our claims are answered, then the answers should be
>answered before the claims are re-stated.
>

You reply:

	We disagree also on which of us has best supported our claims more
effectively. I believe that I effectively supported my claims on this
issue. You believe that I failed to support them adequately. Again, I'm
not sure that this is a fruitful line of discussion at this particular
time.

I reply:

Of course there's no point pursuing the issue of who has better made their 
case. I merely was asking if there were any additional arguments.

I'd said:

>Condorcet has been used on a small scale in many polls, here and
>elsewhere on the Internet. Condorcet has been in use in organizations, and,
from
>what I´ve heard, the results have been very satisfactory.

You asked:

	That's good. Which organizations use Condorcet?

I reply:

The Debian organization uses Cloneproof Schwartz Sequential Droppinlg 
(CSSD).

The Free State Project, a Libertarian organization, uses PC.

And of course a number of polling websites use various versions of 
Condorcet, usually wv.

These include the website at http://www.masquilie.org  ,which has facilities 
for members to set up polls, using BeatpathWinner or Approval, which then 
counts and displays the results; and the international U.S. presidential 
poll Leader of the Free World, which also used BeatpathWinner.

Nader won the international Leader of the Free World poll, and some of the 
presidential polls at masquilier.

Rob Lanphier has a wv polling website, and so do Eric Gore and someone else 
on EM.

>
>1. plurality or runoffs
>2. equal-rankings IRV
>3. minimal dominant set completed by IRV.
>4. cardinal pairwise??
>
>Well, we´ve already got Plurality and Runoffs, so hopefully we could soon
>be out of stage 1.
>For #3, why not use ERIRV instead of IRV to choose from the minimal
>dominant set (Smith set?)?

	I meant that to be implied.
	By the way, I'm not really devoted to that progression or anything...
there are several other possible progressions that I might find attractive.

I comment:

I'd like to add that last spring I found a demonstration that convinced me 
that ERIRV(fractional) meets WDSC. Steve and I had previously looked for a 
WDSC failuire example of ERIRV(fractional), and failed to find one. Then, 
last spring, I thought that I'd found a demonstration that there can be  no 
such failure example, that ERIRV meets WDSC.

Maybe it does, but now I can't find a demonstration of that--or a failure 
example. Now IL don't know if my demonstration that ERIRV(fractional) meets 
WDSC contained an error.

Steve, if you read the list e-mail, can you post a demonstration that 
ERIRV(fractional) meets WDSC?
ERIRV is IRV with equal ranking allowed (at least in 1st place). When you 
have several candidates at a rank position in ERIRV, and your vote is at 
that rank position, ERIRV divides your vote between those candidates at that 
rank position.

As for progressions to a better voting system, by far the best chance is to 
start out with CR. Then, from there Approval could be a natural move, when 
people understand that Approval is just a CR version, rather than illegal 
Plurality voting. But maybe, when we have CR, and people know that there can 
be a genuinely better voting system, and find out about the guarantees that 
good voting systems can offer, the public might want more guarantees, the 
ones offered only by Condorcet.

If it just stays CR, that's fine too, because CR is an excellent method.

So, 3 possibilities:

1. CR
2. CR and then Approval
3. CR and then Condorcet

Speaking for myself, if CR is going to lead to another method, I'd prefer 
that it be Condorcet rather than Approval, though Approval would be good 
too.

I'd said:

>It doesn´t seem so to me, because it would be enough for voters to 
>understand
that AERLO & ATLO can be used to give fullest protection to a favored set of
candidates.


You reply:

	Okay, fair enough. Let's say that the voting method is AERLO/ATLO
Condorcet, and I'm a voter who doesn't know much about voting methods, and
my choices in order of preference are C1, C2, C3, C4, C5, C6, and my
"favored set of candidates" is {C1, C2}. How do I give full protection to
these candidates using the AERLO/ATLO lines? Can I do it just with the
AERLO line, or do I need to use the ATLO line as well?

I reply:

Of course you know that, strictly speaking _full_protection is possible only 
by applying both AERLO and ATLO, since both protect the favored set under 
more than one set of conditions.

True, that's asking something of the voter, judging which to use,or both. 
Well, giving those candidates both is an obvious way to give them the best 
protection.

I defined AERLO and ATLO as separate options for flexibility, but you have a 
point: Flexibility isn't necessarily something that the public want to deal 
with. Maybe AERLO and ATLO should be combined as one option, effectively 
AERLO + ATLO, for simplicity.

Mike Ossipoff

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