[EM] paradigms...

[Jobst Heitzig]

Dear Dave!

I truly apologize for giving you a headache, which is surely the last I
want to cause... The thing is, although it seems to me that I try to
make the most obvious things clear, it seems to need explanations over
and over again. So, excuse me for repeating some arguments in reply to
you...

you wrote:
> We are electing ONE PERSON, 

Of course we do. But we do it as a group of people, by aggregating
individual preferences which are interpreted as containing valuable
information about which candidate is both socially optimal and a just
and stable choice. That does *not* require each voter to have so much
information that s/he can tell a unique first choice.

>       Given an IRV ballot I proceed as you describe above (except, if I am 
> a serious voter, I will likely do less coin tossing).

So, may I ask what you would do instead when pressed to give a strict
ranking? Because this was the imagined situation in my example... when
ties had been allowed, the ranking would of course have been A and B
first, C and D second. But note that that would still distort my true
preferences in erroneously indicating a preference of B over C and of A
over D...

>       Given a Condorcet ballot I proceed in EXACTLY the same way, 
> expecting IDENTICAL results, even though the debating may use different words.

Sorry, I can't follow you here. What is a "Condorcet ballot"? I know
ranked ballots with or without ties and with or without truncation. And
I know pairwise ballots with the possibility to vote either A>B or B>A
for each pair A,B, hopefully with the additional possibility of voting
A=B (equivalence) or A?B (abstention). On the other hand, the term
"Condorcet" refers to a method not a ballot, isn't it?

>       Correction - liking X1 better than X2, I do not need a coin toss to 
> prefer B over C!

Are you sure? Well, I'm not: X1 may only be slightly better than X2, but
Y2 may be FAR better than Y1 without me knowing that because I'm not an
expert on issue Y. In that case I would surely prefer C to B, the only
problem is that at the moment of voting I don't have that information.
So I would be stupid to express a preference B>C when it can easily be
that my true preference would be C>B as soon as I get enough information.

>       Also, not caring as to C>D vs C<D, I should not vote a nonsense 
> implied preference between them.

Your absolutely right!!! That's what the whole example is about!! I
don't want to be forced to express nonsense preferences which I don't
have only to be able to express some other preferences I do have. BUT
THE RESTRICTION TO USE RANKED BALLOTS FORCES ME TO DO EXACTLY THIS, and
that's what the example shows.

> But, in your demonstration voting, you indicated a preference about Y that 
> you admit here was nonsense.

See above.

>>> However, let's assume that the ranking system in question allows you
>>> to, rather than flipping a coin, simply rank A and B equally.  In
>>> other words, declare them a tie.
>> 
>> 
>> That would be fine as long as I could really do so! But as long as I can
>> only express rankings I cannot do as you suggest! In a ranking, I cannot
>> tie A=C, B=C, A=D, and B=D and simultaneously express A>B and C>D.
>> 
> Agreed that the target you offer is impossible, BUT, it has nothing to do 
> with whether the sentence you are responding to is valid - and I see 
> validity there.

Yes, of course, the sentence is right: if I were allowed to tie all
pairs of candidates which I'm actually undecided about, then fine. But
again ranked ballots do *not* allow me to do this as I explained in the
above sentence. I did not suggest to falsly express undecidedness as
equivalence, I only responded to that suggestion and showed that it
doesn't lead us nowhere...

All the best,
Jobst