[conlang_learners] voting system

[Jim Henry]

On Sun, Aug 16, 2009 at 2:26 PM, Philip Newton<philip.newton at gmail.com> wrote:
> 2009/8/16 Jim Henry <jimhenry1973 at gmail.com>:
>> 2. Voters should email their ballots offlist to Philip Newton and one
>> or two other volunteers (we need at least one more volunteer who
>> doesn't have a conlang of their own nominated)

> BTW, may the ballot counters also vote? Or may they not only have no
> conlang of their own under consideration, but may not vote, either?

Hmm.   I suspect that since we plan to  have two or three vote
counters, and they're unlikely to have identical preferences, they can
keep each other honest.  I think we should allow them to vote.  What
do the rest of y'all think?


2009/8/16 Olivier Simon <cafaristeir at yahoo.com>:

> Points 1° and 2° are independent from each other, as the paradox of
> Condorcet can already occur with only 3 candidates (while we have a
> selection of several tens of conlangs...)

> "10" was a pure example; I meant that, if voters are entitled to propose
> more than one conlang in their ballot, there must be a prealable agreement
> about how many points gets each conlang ranked in order of preference (think
> at the "European Song Contest").

Sure, if we're using range voting.  With instant-runoff voting that
doesn't matter.


> It is the assignment of points which prevents the paradox of Condorcet from
> happening. And the assignment of points can be done only by starting from
> the highest value.

I've been reading the Wikipedia article on "Condorcet method":

http://en.wikipedia.org/wiki/Condorcet_method

Assignment of points to candidates by itself doesn't prevent the
Condorcet paradox.  The Condorcet paradox can't be entirely avoided,
only somewhat ameliorated; and not all range or ranked voting methods
are Condorcet-criterion methods.

Can you please be more specific about what you're proposing, either
pointing to an online source such as a Wikipedia article on a very
specific Condorcet-criterion voting method or explaining the method
you have in mind at greater length?


> We can also consider that all of the 25 proposed conlangs on the Frath Wiki
> List (+ 2 contested not taken into account) can be proposed in each vote,
> which will entail that Chosen Conlang n°1 will get 25 points, n°2 gets 24
> points, n°3 23 points, etc...

Wouldn't it be better, in terms of reflecting fine-grained differences
in people's preferences, to allow people to assign any arbitrary
number of points from 0-25 to all the conlangs they list on their
ballot?   That was the reason some people gave in earlier threads for
preferring range voting to instant-runoff voting: IRV requires you to
give every candidate you list a different rank and doesn't let you
specify how strong your relative preferences are or say (for instance)
that after your favorite conlang Foo, you next prefer Bar and Baz in
no particular order.  But maybe allowing people to give two conlangs
equal rank on their ballot makes Condorcet paradoxes more likely...?
Again, I'm not mathematician enough to know.  But I consulted a
mathematician friend by email and he recommended IRV.


> Considering the number of auxlangs on the list, it is unlikely that we'll
> get two (or more ?) conlangs with an equal highest number of votes.

There are only 4 auxlangs on the shortlist, out of 25, by my count.
Are you using "auxlang" as a synonym for "conlang"?


> In any case, I support the idea of a multi-choice vote, for it is the most
> descriptive of the aspirations of all voters. (For example, we may have
> completely different opinions on "our" preferred conlang n°1, but we may all
> be quite near about n°2 or n°3; a multi-vote system will then favour
> conlangs that interest the biggest number of voters, and not

That is exactly the merit of IRV.   Range voting does pretty well at
that too, but IRV is particularly good at finding a compromise
candidate that a majority people like second or third-best when
they're no majority among their first choices.  It doesn't satisfy
Condorcet's criterion, but neither do any of the range voting variants
that have been proposed, including yours, as far as I understand it.

Also, it sounds like the vote-counting methods that satisfy
Condorcet's criterion are significantly more complicated to administer
than the ones that don't.  And even Condorcet methods don't always
find an unambiguous winner on the first pass, and need various
somewhat arbitrary ways to pick a winner when voters' collective
preferences are cyclic.  The voting paradox is kind of like the Second
Law of Thermodynamics; an unpleasant fact about reality, not a flaw in
particular voting systems or machines.  No possible voting system will
produce an inarguably fair result under all possible circumstances,
just as no possible machine can work with zero waste and inefficiency.

-- 
Jim Henry
http://www.pobox.com/~jimhenry/