[EM] Kristofer: MMPO objections

[Kristofer Munsterhjelm]

On 12/30/2011 10:59 PM, MIKE OSSIPOFF wrote:
>
> Kristofer:
>
> First, let me agree that "not-valid" is only a subjective opinion. I was using it as shorthand to mean
> that I don't consider the objection to be important.
>
> So I don't deny the subjective-only value of "not-valid" when I said it.

Thank you.

> I'm not saying that the A and B voters, as a class aren't wronged.
>
> I'm saying, "How serious is it really,considering the above hypothetical conversation?"
>
> Let me put it differently:
>
> I'm just not seeing the problem that you're seeing.
>
> When I say "problem", I mean a strategy dilemma like the need for favorite-burial, or the
> co-operation/defection problem. When I speak of a "problem", I'm referring to a genuine,
> big, problem to voters. A practical problem. A problem that will keep the electorate
> from achieving the change that they want.
>
> You haven't shown that Kevin's MMPO bad-example is a problem in that sense.

I think my primary objection is that it doesn't make sense. Perhaps that 
is aesthetic, perhaps it is based on logic. But, as I'll say below, 
Plurality (the criterion) is much more than Plurality (the method); and 
I think, though I am of course not certain of this, that the vast 
majority of people would find something wrong with how MMPO resolves the 
situation.

Or to try to put what I think (and how I think people would react) into 
more direct terms: this would be like "everybody but Bush" voters and 
"everybody but Gore" voters (A and B plumpers respectively) wake up and 
find that, because a single voter voted some obscure candidate (say 
Hagelin) equal to Gore, and another voter voted the same obscure 
candidate equal to Bush, that obscure candidate won.

Sure, you could say that the everybody-but-Bush and everybody-but-Gore 
voters should have clarified, in their ballots, that they do prefer Gore 
and Bush (respectively) to Hagelin - but I think it would seem wrong to 
the voters that they would have to do so, and that the method would not 
just degrade gracefully if they did vote "everybody but Bush/Gore" right 
out.

If we take that more generally, the contention might be about what not 
ranking someone actually means. A Plurality (criterion) failure is worse 
the greater the chance is that when someone votes for A alone, he means 
"the rest are so bad I don't want to rank them". In this respect, the 
pairwise opposition mode would be the most "forgiving" (so to speak) of 
equally ranking someone last, margins would be somewhere in the middle, 
and wv would be the most strict.

> What I see in Kevin's MMPO bad-example is an un-plurality-like
> outcome. We want results better than those of Plurality.
 >
> The more improvements we want over Plurality, the more our results might sometimes depart
> from what we're used to in Plurality. Especially if we're greedy for the super-brief
> definition of MMPO, or its great flexibility as a full-rankings method, or the
> simplicity of only requiring unqualified unilateral support, and its better job of
> electing unfavorite CWs.

The Plurality criterion isn't just failed by methods that return an 
un-Plurality-method-like result. It is also failed by methods that 
return an un-Approval-like result. Recall that the Plurality method says 
"if A is ranked first on more ballots than B is ranked at all, B 
shouldn't win". Failures of this is a subset of failures of the form "if 
A is ranked on more ballots than B, B shouldn't win". Now, the latter 
criterion may be desirable or not, but it would show that the Plurality 
criterion is about more than just getting the result that you would 
under Plurality, the method.

(Approval itself would give a tie between A and B.)

> So, when asking for
> so much, yes I admit that the method's results could depart from those of Plurality so as
> to bother people who are accustomed to Plurality.
>
> MMPO and MDDTR get their advantages from their big departures from Plurality.
>
> I don't deny that those departures from Plurality could cause a problem for a public enactment
> proposal. That's why I consider my conditional-middle-ratings proposals to be better public
> proposals.

Alright.

> Someone on this list was terribly bothered by the mutuality-requirement, referring to it as "sordid".
>
> He'll think this is terribly sordid, but if a faction of voters want coalition support for their
> candidate, why would it be important to them that they not support that coalition?
>
> They want to not help the people whose help they need?
>
> Oh what a cruel strategy-need to burden someone with! :-)

When I read this, I recognized something that may explain others' 
disagreement about the relative value of these criteria, and 
mono-add-plump in particular.

In my mind, a preference ordering is (or should be) disconnected from 
the method to which it is fed. In other words, a honest ranked ballot is 
an expression of the wishes or preferences of the voter independent of 
the voting system. In essence, he is saying "if it was up to me, I'd 
have A, but if I couldn't have A, I would have B" and so on.

In that light, a criterion-based approach becomes much more sensible. If 
an election method is a black box that takes preferences as inputs and 
tries to "mechanically mediate" between the preferences to find the best 
social outcome, then it doesn't matter what happens inside the black 
box. What matters is the relation between the outputs and the inputs. If 
a method fails mono-add-plump, that means that when someone says "As for 
me, I prefer A to everybody else", then the method uses that preference 
to decide that "oh, A wasn't that good a choice after all". That may 
happen because of mutuality requirements, or because of some unintuitive 
internal logic, but the outcome changes in a way that is detrimental to 
the voter who made the outcome change in the first place, and this voter 
doesn't know anything about mutuality or coalition preferences or the 
internals of the method when voting honestly.

If others have this view of preference orderings as well, that could 
explain why they don't like a method failing mono-add-plump, and why an 
explanation of how the method works doesn't change their minds.

Obviously, no method can pass every criterion, but mono-add-plump is 
relatively "cheap" in that most methods out there pass it.