[EM] Scoring (was Re: OpenSTV 2.1.0 released)

[Michael Ossipoff]

On Mon, Sep 24, 2012 at 4:13 AM, Juho Laatu <juho4880 at yahoo.co.uk> wrote:
> On 24.9.2012, at 8.43, Michael Ossipoff wrote:
>
>> You said that you don't support unimproved
>> Condorcet. But then, later, maybe in the same post, you cited Dodgson
>> and Beatpath as methods that you seem to be endorsing or advocating as
>> methods that do better than SITC, with regard to the ideal sincere
>> winner.
>
> Unimproved Condorcet, as you defined it, may contain both good and bad methods. I don't >support the bad ones.

It comes with all of its elements. if you don't accept some of its
elements, then you don't want it.
If you don't "support" some of its elements, then you don't "support"
a method that has those elements.

You said:

I have not taken position on Dodgson. I included Beatpath in what I
referred to as "basic Condorcet methods".

[endquote]

When I defined "unimproved Condorcet", to include Beatpath, etc., you
said that you don't support unimproved Condorcet.

You said:

The only method that can be linked to my example sincere winner
definition is Minmax(margins).

[endquote]

No. You spoke of electing the candidate who can be made into CW by
changing as few pairwise votes as possible. That's Dodgson, not
MinMax(margins). As I said, MinMax(margins) only looks at a
candidate's largest defeat (largest margin of defeat). Dodgson looks,
as you described, at all of the margins against a candidate, to find
the candidate who could be changed into CW by changing as few pairwise
votes as possible.

But don't worry about that. Neither method can be shown to be better
than SITC, with sincere voting.


>
>> But, as I explained in my previous post, you haven't told us what
>> exactly you want in the way of an ideal sincere winner, or what method
>> you think would choose it better than would SITC. (If you're sure you
>> want to commit yourself to Dodgson or Beatpath, then say so
>> unambiguously.)
>

You said:

> I think I already said this a few times

Then you think wrong.

You said:


, but the idea is that the definition of which candidate should be
elected with sincere votes can be election specific. I presented one
possible definition.

[endquote]

Dodgson, or maybe you'd really prefer MinMax(margins). I commented on that.

You said:

That definition happened to be the same as Minmax(margins).

[endquote]

No. Dodgson. But that's ok. Don't worry about it. If you think that
MinMax(margins) or Dodgson is better than Symmetrical ICT, under
sincere voting, you have yet to tell why.

You said that it helps an officeholder's status, against opposition in
office, if s/he could be changed into CW by changing as few pairwise
preferences as possible. Do you really think that would help hir
status against opposition in office better than being the most
favorite candidate in the top cycle?

You said:

I don't claim that it is the best criterion for all elections, but it
could be chosen for some.

[endquote]

So you're saying that different voting systems should be used for
different elections. But, as each new election comes near, who decides
which method will be used for that particular election? Or does
someone choose the method for the election after the ballots have been
looked at?

If you mean "for different offices", then I'll clarify that I'm
referring to presidential and Congressional elections in the U.S.
Should we use different voting systems for presidential and
Congressional elections? If so, then which one would be better (by
ideal sincere winner) for the presidency,and which would be better for
Congress?

Of course, judging by how well they choose the ideal sincere winner
assumes that you still think that there won't be a chicken dilemma,
and can tell why.


>
>> Nor did you answer my question about the chicken dilemma.
>
> I answered something on the chicken dilemma. What more do you want to know?

You incorrectly claimed that unimproved Condorcet has less chicken
dilemma than Approval does.

Apparently much of what you've been saying depends on that incorrect
assumption. If there will be defection in situations like the chicken
dilemma examples, then can you still advocate Beatpath,
MinMax(margins) or Dodgson over SITC, by saying they will get sincere
rankings? (and that, with sincere rankings, they do better at choosing
the ideal sincere winner?)

>> Basically, you haven't answered any of my questions.
>
> I tried to cover all the questions in your mail. You may point out the unanswered ones, so I can check what I can do with them.

1. What makes you think that MinMax(margins), Dodgson, or Beatpath
won't have a chicken dilemma? In other words, what makes you think
that there won't be defection when those methods are applied to the
chicken dilemma examples that have been posted here, if such a
configuration of faction-sizes occurred in an actual election?

If you ask me why they would, then I'll suggest that you try them in
the usual chicken dilemma examples. Must I do that, to show you their
chicken dilemma? Request it and I will.

2. What makes you so sure that the United States won't have a
significant amount of favorite-burial, when unimproved Condorcet, such
as Dodgson, MinMax(margins) or Beatpath, is used?
...given the fact that our media are strongly committed to promoting
favorite-burial, and the fact that the public are easily deceived by
the media regarding such things as candidate winnabilities, and
apparently even about the matter of where to draw the
"acceptable/unacceptable" line.

Sometimes you seem to say that you're just speaking in general, about
most societies, or many societies. Sometimes, though, you make
assertions about what won't happen here.

3. What is your best argument to support your belief that Dodgson,
MinMax(margins) or Beatpath would do better at choosing the ideal
sincere winner, if voting were sincere, than SITC would do?

4. Tell the requirements that describe the ideal sincere winner.


> Btw, is this a promise that also you will answer all my questions?

I always answer all questions.

>>> But if you allow me to define the sincere winner for you, and for your >election, you could take my example definition and compare it to SITC.
>
>>> There is a difference between those two definitions.
>>
>> What two definitions would those be?

You said:

> SITC and the quoted lines below. I believe it is obvious that they are different.
>
>>>>> (In my text above I asked you to provide a definition of the best candidate. A simple Condorcet oriented definition could be e.g. "the candidate that requires least additional support/votes to beat any of the other candidates in a pairwise comparison/battle should be elected". This target could be selected because it gives one rational argument why the winner could be able to rule well (= only little bit of additional support needed (if any) to gain majority support for his proposals while in office).)

[endquote]

Certainly SITC is a different method from Dodgson or MinMax(margins).
It wasn't clear that that's what you meant. But, if that's what you
meant, then you're right about that.

So what?


>> MinMax(margins) looks at the biggest defeat. Dodgson looks at the
>> number of pairwise preferences that would have to be changed in order
>> to make your candidate the CW. Those are two different considerations.
>> You explicitly referred to the one that is Dodgson.

Say your candidate, X, is pairbeaten by Y with a margin of 100, and by
Z with a margin of 50.

Say my candidate, A, is beaten by B with a margin of 101, and by C
with a margin of 1.

It's a circular tie. Our candidates are in that circular tie.

Dodgson elects A. Minmax(margins) elects A. Minmax(margins) elects X.

Whether you referred to Dodgson or MinMax(margins) depends on what you
meant by "any".

"Any" is an ambiguous word that I avoid using in definitions.

"I can beat _anyone_ at chess. Can you beat anyone at chess?"

I assumed that, by any, you meant "every". But if you, instead, meant
"any one candidate", then yes, you referred to MinMax(margins).

That ambiguity between "every" and "any one" is the same one in the
chess example.

But, as I said, I don't care which one you meant. You're willing to
answer my questions, listed and numbered above, for MinMax(margins).


> Check again. I didn't refer to changed votes but to least number of additional votes, which >happens to be the same thing as biggest pairwise defeat in margins.

As I said, it depends on what you meant by "any". But it doesn't
matter. Let's say you meant MinMax(margins).

>> Then you need next to show that Dodgson (or is it MinMax(margins)?) is
>> indeed better than SITC, with regard to ideal sincere winner, "for
>> some needs". Then you need to show that those needs are socially
>> important and frequently needed.
>
> No reference to SITC nor to Dodgson. But my quoted example best winner definition above includes also a short explanation on why this criterion might be considered useful for the society ("This target could be selected because..."). You can take that as a starter.

You mean the status against opposition in office of the candidate
whose largest margin against him, in favor of another candidate, is
the least. And being the most favorite, having the largest faction,
doesn't confer any status against opposition in office? :-)


>> But yes, if there's a top-cycle, the choice, under sincere voting, is
>> less important, as compared to when there's a candidate who pairbeats
>> each of the others.
>
> I take that to mean that a complete definition of the best sincere winner should define also which >candidate wins if there is a to loop.

The definition of a Condorcet method must, of course include such a provision.

But you're the one who wants to use the best sincere winner. No,
saying that the choice, under sincere voting, becomes much less
important when there's a top-cycle instead of a CW certainly doesn't
imply that top-cycle solution is an important part of a method, as
regards choosing the best sincere winner. In fact, it implies the
opposite. It means that the top-cycle solution is much less relevant.


>> So you're now saying that Dodgson (the method you specified there)
>> always chooses the ideal sincere winner.
>
> The example definition (not Dodgson) could be used for some elections.

...whether it's a good idea or not. So could Plurality.

>
>> For one thing, I claim that Dodgson is based on an illegitimate
>> definition of "CW".

You said:

> Ok, I have by now learned that you want to change the regular Condorcet criterion and interpretation of the rankings on the ballot to something else, and you consider that definition better, I guess both for defining the sincere winner and for practical elections.

[endquote]

Respecting voters' preferences, intentions and wishes is desirable for
the purpose of choosing a legitimate winner under sincere voting. But
that respect for voters' preferences, intentions and wishes also
doesn't give them incentive to favorite-bury.


>> But you'd claimed that favorite-burial won't be a problem (but, when
>> pressed, you become entirely vague about where it won't be a problem).
>
> I can't exclude the possibility that some society would turn into widespread burial, but by default I >assume that it is not a problem.

So the default hypothetical country or society consists of people who
wouldn't have favorite-burial incentive. I'm not proposing FBC
complying methods for your default country. I'm proposing them for the
country in which I reside.

In any case, it's often a mistake when your default assumption is the
optimistic assumption. I don't criticize optimism. I think voting
should be based on optimism. But obviously it can be a mistake to make
the optimistic assumption your default assumption:

"I don't know if a car is coming toward me, on the other side of the
road, around that curve that is ahead. My default assumption is the
optimistic assumption, and so I'm going to assume that no car is
coming on the other side of the road, and so I'm going to pass that
large truck in front of me, on this curve, along this cliff."

Sometimes, often, the best default assumption isn't the optimistic assumption.

Mike Ossipoff