[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: supermajority options

On Wed, Nov 20, 2002 at 11:04:10PM -0800, John H. Robinson, IV wrote:
> > > i tend to agree with the philosophy that you need to convince at least
> > > half of the voting populous.
> > 
> > [1] Who is the voting populous?
On Fri, Nov 22, 2002 at 01:11:12PM -0800, John H. Robinson, IV wrote:
> depends upon where you count, you have all elgible voters, then you have
> those voters that actually vote, then you have the minimum number of
> voters to vote for a vote to be binding.
> the last one refers to quorum.
> i specifically meant half of those that are interested enough to
> actually vote. if that number fails to meet quorom, then there is not
> enough interest in the issue to perform any action on it.


Given that criteria, what would the reason for saying that all decisions
should be made at the 50% mark?  In particular, if at time (A) we can
make a unanimous decision, should we agree that at some later time (B)
we should be willing to overturn that decision based on 51% approval?

> > [2] Why are they the voting populous?
> depends upon context. in the US, it is citizens over 18 that have not
> had their right to vote (ie: convicted of a felony) revoked.
> in this context, it is each Debian Developer as listed in the
> Constitution.

Yes.  And we're talking about revising the constitution.

The point I was making is that debian developers are a set
of people selected based on interest (self-selection) and
on competence (which helps maintain interest).

If nearly half of the interested developers think something is a bad
idea, what would be a reason for saying we should adopt that idea for
all future developers?

> in other contexts, by virute of being in the set defined as The Set of
> Eligble Voters,or however you want to phrase it
> > [3] Is competence an issue?  Why or why not?
> yes it is. in the US system, it is assumed that you are competent at
> age > 18, until proven otherwise (ie: convicted of a felony)
> in this context, each DD has been vetted by the NM procress, in whatever
> form that was at the time of induction.
> in this context, we also demand competence by forcing voters to follow
> directions on filling out and returning the ballot.

I think that's an incomplete of the process.  Our mailing lists are
important for of opinions and discussing them.  You've left out that
our primary focus is on technical issues, and that our administrative
stuff is only useful to the degree that it helps us solve the technical
issues [like "this package broke on 68k in the last stable release,
but it worked before that".]

Most useful decisions are made by individuals.  In group decision making,
we use "degree of popular agreement" to measure competence.  Which leads
right back into the question of majority requirements.

> > [4] Is involvement an issue?  Why or why not?
> yes. you have to have at least ONE person vote. if NO ONE participates,
> then you have no results. you can increase the demands of participation
> via quorum requirements.
> i can see requiring a higher quorom for certain actions (Constitutional
> Ammendment) than for other things (Official Debian Mascot) to ensure
> that there is enough interest in the issue. i beleive that the more
> interest you have, the more likely you are to see a greater reflection
> of the groups desire. you get a perfect picture at 100% participation.
> if every eligble DD voted, and the Smith set had only 415:414 would that
> indicate that the Debian Project should move in the direction indicated?

That's one of the things we're trying to decide when we revise the
voting draft.

> > [Hint: for most things in Debian, you need to convince at least one
> > person who happens to be the package maintainer.]
> yep. in the case of GR's, you need to convince six (sponsor and five
> seconds)
> > > Condorecet seems pretty resilient to insincere voting. for each method
> > > of counting Supermajorities, it has been shown to where it possible, in
> > > some cases almost trivial, for an insincere vote to change the result of
> > > an election. that appears to defeat the whole purpose of using Condorcet
> > > to begin with.
> > 
> > For some methods, this is true.
> > 
> > You seem to be assuming this is true for all methods, but you offer
> > no proof.
> that is correct. i do not have the math to do that. however, those with
> the math (here i refer to electionmethods.org) do not address the issue
> of Supermahorities at all.

Debian's concept of supermajority is tied to debian's concept of quorum.

Our constitution has tried to define a decision making where a small
group of self-selected activists can make decisions for the whole.
That's a very different sort of situation than the sort of thing which
electionmethods people usually try to address (where if significantly less
than half the people are involved the entire process becomes suspect).

I think we can get away with it because we're not making life and death
decisions about people.  Instead, we're making technical decisions
about computers.

Electionmethods normally are based on some sort of "everybody must be
represented" or "most everybody must participate" -- we've tossed that
idea out on its ear, and we're using supermajority to give us an
approximation of the "unanimous consent" that we used to enjoy.

> > > just out of idle curiosity, has anyone asked the electionmethods people
> > > about Condorcet+Supermajority?
> > 
> > Yes.
> > 
> > Unfortunately, most of them seemed to lose interest in the discussion
> > before we had much discussed the underlying issues.
> :(  so we are floundering on our own on this one?

Pretty much.

Then again, the issues are subtle but not impossible to understand.
And flawed understanding is relatively easy to point out (using examples).

> > That said: Debian 3:1 supermajority is LESS OF A CONSTRAINT than a
> > requirement that a majority of the voting population agree.
> depending upon quorum requirements, and how you defice ``voting
> population.'' most methods define it as those that are eliglble to vote
> _and_ actually vote. this is where quorum comes in: that there is
> sufficient interest to use as a statistical model that the subset
> reflects the entire set.

I'm assuming current debian quorum requirements when I say "Debian 3:1

> > Are you suggesting that we prefer majority rule because it's more of a
> > constraint ["more tyranical"] than supermajority?
> i will agree with branden that the use of the word tyrannical in this
> case is bad. however, i will still answer the question.
> i am suggesting that we prefer majority rule because with our election
> method (Condorcet) it has been shown that at least some methods of
> counting Supermajorities can lead to insincere voting for strategic
> purposes being effective. this is one of the things we wish to _avoid_.

Ok.  That is what I believed your reason was, but I wanted to have a
clear statement of it.

I'm assuming that we can describe an implementation of supermajority
with CpSSD where supermajority does not encourage insincere voting.

Right now, I'm trying out examples against Anthony's most recent
implied draft.  I feel that even if it's not perfect, I'll have a better
understanding of the properties of our voting system if I can find a flaw.
However, I've not yet found a flaw.

If we can't find a decent implementation of supermajority, I'll be
forced to agree with you that we should drop the idea of supermajority.
However, I'm not ready to give up yet.



Reply to: