Re: Constitutional amendment: Condorcet/Clone Proof SSD vote tallying
Raul Miller wrote:
> On Fri, May 23, 2003 at 04:40:49PM -0700, John H. Robinson, IV wrote:
> > correct me if i am wrong, but, isn't quorum suppoed to _prevent_
> > minority rule? now you are saying that minority rule is good, and
> > desired?
>
> What do you mean?
i mean to point out a hypocrisy. on the one hand, say that minority rule
is bad, but implement in such a way as to allow minority rule.
> There are forms of minority rule which quorum prevents, and there are
> forms it allows.
and the proposition uses a form that allows it.
> Certainly, quorum isn't about having a small group of people dictating
> to a larger group of people without the larger group having any say.
true. quorum is intended to ensure that enough people participate to get
good sampling. ie: no secret, behind the doors meetings to decide what
really happens.
> > if that is the case, i recommend scrapping the entire idea of quorum, as
> > it breaks condorcet in strange and interesting ways.
>
> Condorcet itself allows certain forms of minority role and prevents
> others.
example? other than the case of extreme voter apathy, of course.
> Sometimes I get the idea that you're saying things, not because you're
> trying to achieve some worthwhile end, but because you like the way
> they sound. Is that the case here?
no, i am afraid not.
> > in this case, the options were not fairly close at all. 10 people
> > prefered A over B. only five people prefered B over A. that is a 2:1
> > margin. that is a 2/3'ds majority in favour of A, and it still lost.
>
> So what?
>
> In the example I presented earlier (ballot: A, B, default D, one
> vote: ABD), option A was infinitely preferred over all other options.
> You're arguing that we should accept such a vote, even though (other
> than one person) no one wanted to (or was able to) participate in it.
in a classic Condorcet, yes. but as i explained in a previous email
(http://lists.debian.org/debian-vote/2003/debian-vote-200305/msg00133.html
or Message-id: <20030523205148.GG14022@ucsd.edu>) i outlined five major
choices we could make, along with their relative merits.
Arrow's Theorem shows that no voting system is perfect. so we _must_
accept some flaws. the nice thing is, we get to choose our own poison.
we use a basis of Condorcet, because that is the one that is the best
amongst a collection of flawed systems.
two of the flaws, as far as Debian is concerned, is that Condorcet does
not support a tyranny of the status-quo, nor does it prevent minority
rule.
we have attempted to solve the first flaw by the use of supermajorities.
in so doing, we have created a new, always-present option called the
Default Option which is basically a null-option. if a supermajority
requiring option fails to beat the default option in a one-to-one race
with respect to its supermajority requirement, that option is discarded.
we have attempted to solve the second flaw by the use of quorum. in so
doing, we have created a new, always-present option called the Default
Option which is basically a null-option. if an option fails to beat the
default option in a one-to-one race by the quorum requirement, that
option is discarded.
immediately, we see that the Default Option plays double duty. in the
one case of supermajority, and the other case in quorum.
i agree that Condorcet does make the best basis for a voting system. i
do not agree, however, that our application of quorum, to prevent
minority rule, is the correct or even the simplest approach.
i believe that the per-option quorum opens us up to strategic-voting
(rank all other options below default, so they don't make quorum, but
otherwise rank them in the order that you prefer to allow the proper
Condorcet ranking in the case that the other options do make quorum).
my proposed solution opens up another problem, that of a single vote
causes an otherwise nullified vote to declare the opposition a winner.
the latter case will only occur in cases of extreme voter apathy, or
when the quorum requirement is very close to the number of voters (such
as in the case of small voting body). example: the technical committee
has a membership from four to eight members, and quorum is always two.
thus quorum can be from 25% to 50%.
Manoj pointed out that his proposal would cause more votes to end in a
default state than my amendment would. he is correct. under my
amendment, if a vote fails to make quorum, it is as if the vote never
occurred. i see this as drastically different from the case where a
binding default option wins. in practice, it might not be so different
(exception: DPL elections. a binding None of the Above would be a lot
different from a nullification).
> And yet, you're arguing that we should accept this kind of election as
> valid, and apparently the basis you're using for your argument is that
> minority rule is bad.
>
> What's the point?
under my amendment, the only way a single vote of ABD would be binding
is when the quorum requirement is one. this is the exact same case for
the proposal.
the more i think about this, the less and less i like all the special
weight applied to a single option, even if that options is effectively a
null-option. unfortunately, i do not have good answers as to how to
allow a tyranny of the status-quo and prevent minority rule, which i
accept as being desirable.
anyway, i have a long weekend ahead of me. i think i will enjoy it.
-john
Reply to: