Re: Call for votes for SPI board of directors membership election

At its October 14, 2003 meeting the board of directors accepted
resolution 2003-10-14.iwj.6, which details the procedure for extending
the board with three new members.

During the period from November 7 to (but not including) November 21
(UTC) contributing members can cast their vote online. People who
become contributing members during this period are also eligible for
voting. The voting system can be found at the following address:

https://members.spi-inc.org/vote/

Please take a look at the statements submitted by the candiates at
http://www.spi-inc.org/secretary/votes/vote2/cfv.xhtml before casting
your vote.

Votes will be counted using the "Concordet" election method system which
will be used to select the most preferred candidate. Conceptually, the
election will be broken into a a series of pairwise races between each
possible paring of the candidates. If one candidate beats each of the
others in pairwise races, that candidate wins. Otherwise, the
"Cloneproof/Schwartz Sequential Dropping" method is employed to choose
the most preferred candidate from those remaining.

As such, vote preferences should be made by ranking the candidates. A
vote that simply specifies "X" is does not imply preference. Your
preference will be shown best when you specify all candidates in order
of preference (i.e. "XYZWTUV"). Voting "XY" implies that you prefer
candidate X to candidate Y but will not affect the races between X and
the other candidates or Y and the other candidates. Voting for a single
candidate has no meaning with this method and will not be possible.

Some background reading on preferential voting and Concordet is
available online at http://www.electionmethods.org/ .

Wichert.

--
Wichert Akkerman <wichert(at)wiggy(dot)net> It is simple to make things.
http://www.wiggy.net/ It is hard to make things simple.

Wichert Akkerman writes ("Call for votes for SPI board of directors membership election"):
> As such, vote preferences should be made by ranking the candidates. A
> vote that simply specifies "X" is does not imply preference. [...]

I am very disappointed, Wichert, that despite detailed explanations,
and clear consensus to use a more sensible system, you haven't heeded
everyone's wishes.

I'm now wondering whether the Board should pass an emergency
resolution suspending the election and asking you to restart it with a
more sensible interpretation of ballots.

Ian.

Previously Ian Jackson wrote:
> I am very disappointed, Wichert, that despite detailed explanations,
> and clear consensus to use a more sensible system, you haven't heeded
> everyone's wishes.

I refuse to change a voting system after the voting processes has
already started. We are using the same tested and proven system as
used for the previous election and changing at at the last moment is in
my opinion not a wise decision. I have no problem with discussions for
a different system, but that should be done seperately (as part of
the by-laws revision perhaps) and not be rushed.

> I'm now wondering whether the Board should pass an emergency
> resolution suspending the election and asking you to restart it with a
> more sensible interpretation of ballots.

If the board really wants to do that I'll reconsider my position as
secretary and board member.

Wichert.

--
Wichert Akkerman <wichert(at)wiggy(dot)net> It is simple to make things.
http://www.wiggy.net/ It is hard to make things simple.

On Fri, Nov 07, 2003 at 01:05:01PM +0000, Ian Jackson wrote:
> Wichert Akkerman writes ("Call for votes for SPI board of directors membership election"):
> > As such, vote preferences should be made by ranking the candidates. A
> > vote that simply specifies "X" is does not imply preference. [...]
>
> I am very disappointed, Wichert, that despite detailed explanations,
> and clear consensus to use a more sensible system, you haven't heeded
> everyone's wishes.
>
> I'm now wondering whether the Board should pass an emergency
> resolution suspending the election and asking you to restart it with a
> more sensible interpretation of ballots.

I disagree. You yourself wrote the resolution that specified the
"Condorcet election method". Sorry to tell you this, but Debian's method
is not Condorcet, but modified Condorcet. Real Condorcet doesn't have a
way of specifying equal preferences, and before we decided to use the
modified method I'd want to see some mathematically rigorous comparison
of our modified version and the standard version to see how the
modifications can affect the outcome and what effect that has on Kenneth
Arrow's famous criteria for election fairness.

It also is a very bad precedent for the board to discard all votes that
have been cast so far and restart the election, forcing people who have
voted early to revote. Restarting the election is not much better of a
precedent, since it causes people to wonder if the board will change its
mind yet again and make their second vote useless. It will cause
significant confusion among SPI members.

Furthermore, it contributes to a general impression of the board as
indecisive, which is fatal to an organization like SPI that is at
present run entirely by the board. You yourself say in your platform
that "SPI needs to regularise its own internal structure." Why, then, do
you want to change the structure of a currently ongoing election whose
rules you yourself wrote as recently as last month?

I strongly urge the board to let the election conclude as currently
planned.

- Jimmy Kaplowitz
jimmy(at)debian(dot)org

P. S. - Ian, your MTA is rejecting mails that I send to your chiark
address from my machine, due in part to an absence of reverse DNS that
is beyond my immediate control. Your anti-spam measure on your Debian
address rejected my mail sent to that address, advising me either to use
your chiark address or change my From: line to my Debian address,
despite the fact that my From: line WAS my Debian address. (Maybe it
meant the envelope sender? This is never my Debian address, due to the
way I send mail. Or possibly it was confused because I was using mutt's
bounce feature?) Finally, your SPI address is unroutable. This is really
unacceptable; my only options to contact you are either to send to a
mailing list or to bounce my mail to my mailbox on master.d.o, then log
into master and bounce my mail to your chiark address. I understand your
measures are trying to fix a problem of too much spam, but your
solutions are broken. Fix them, please.

On Fri, Nov 07, 2003 at 08:29:18AM -0500, Jimmy Kaplowitz wrote:
> Real Condorcet doesn't have a way of specifying equal preferences, and
> before we decided to use the modified method I'd want to see some
> mathematically rigorous comparison of our modified version and the
> standard version to see how the modifications can affect the outcome
> and what effect that has on Kenneth Arrow's famous criteria for
> election fairness.

Clarification: true Condorcet does allow you to say "I have no
preference between A and B." (That's what a vote of CD or DC would say.)
It does not allow you to say "I prefer D equally much over A, B, and C."
Also, Kenneth Arrow's theorem doesn't apply to systems like Condorcet
where partially completed ballots are accepted. But, the Condorcet
method has been well and rigorously studied by many people worldwide,
and there have been evaluations of its fairness (by several different
criteria, I'm sure), and among the election methods cognoscenti it is
the most recommended method for fair elections. The same is not true for
Debian's method; it has been discussed by laymen who haven't really done
a rigorous or detailed study of all the ramifications, but it has only
been used in one organization with one set of requirements, and may skew
the results from what would be desired in some as yet unforseen way.
This is what I'm pointing out in the above paragraph.

- Jimmy Kaplowitz
jimmy(at)debian(dot)org

On Fri, Nov 07, 2003 at 08:29:18AM -0500, Jimmy Kaplowitz wrote:
> I disagree. You yourself wrote the resolution that specified the
> "Condorcet election method". Sorry to tell you this, but Debian's method
> is not Condorcet, but modified Condorcet.

As is SPI's method, which uses Cloneproof Schwartz Sequential Dropping,
which was invented by Mike Ossipoff in 2001 [0], some 207 years after
the Marquis de Condorcet's death in 1794.

The "questionable" additions Debian has made to Condorcet are those
that affect supermajority and quorum, not the interpretation of unranked
options.

Cheers,
aj

[0] http://groups.yahoo.com/group/election-methods-list/message/6376

--
Anthony Towns <aj(at)humbug(dot)org(dot)au> <http://azure.humbug.org.au/~aj/>
I don't speak for anyone save myself. GPG signed mail preferred.

Australian DMCA (the Digital Agenda Amendments) Under Review!
-- http://azure.humbug.org.au/~aj/blog/copyright/digitalagenda

On Fri, Nov 07, 2003 at 08:42:34AM -0500, Jimmy Kaplowitz wrote:
> On Fri, Nov 07, 2003 at 08:29:18AM -0500, Jimmy Kaplowitz wrote:
> > Real Condorcet doesn't have a way of specifying equal preferences, and
> > before we decided to use the modified method I'd want to see some
> > mathematically rigorous comparison of our modified version and the
> > standard version to see how the modifications can affect the outcome
> > and what effect that has on Kenneth Arrow's famous criteria for
> > election fairness.
>
> Clarification: true Condorcet does allow you to say "I have no
> preference between A and B." (That's what a vote of CD or DC would say.)
> It does not allow you to say "I prefer D equally much over A, B, and C."
> Also, Kenneth Arrow's theorem doesn't apply to systems like Condorcet
> where partially completed ballots are accepted. But, the Condorcet
> method has been well and rigorously studied by many people worldwide,
> and there have been evaluations of its fairness (by several different
> criteria, I'm sure), and among the election methods cognoscenti it is
> the most recommended method for fair elections. The same is not true for
> Debian's method; it has been discussed by laymen who haven't really done
> a rigorous or detailed study of all the ramifications, but it has only
> been used in one organization with one set of requirements, and may skew
> the results from what would be desired in some as yet unforseen way.
> This is what I'm pointing out in the above paragraph.

I completly agree with the above.

I would just add that in Debian election only one option wins whereas
in SPI 3 options over 4 win. That make the process more likely to
end in a corner case for the algorithm if partial vote are accepted.

Also, it very important to teach voters that Condorcet system do not
allow for strategic voting and that they should vote sincerly instead
of strategically.

Cheers,
Bill

On Fri, Nov 07, 2003 at 08:42:34AM -0500, Jimmy Kaplowitz wrote:
> [...] and among the election methods cognoscenti it is
> the most recommended method for fair elections. The same is not true for
> Debian's method; it has been discussed by laymen who haven't really done
> a rigorous or detailed study of all the ramifications, [...]

I love it when free software hackers explain how a group of amateurs
can't possibly contribute anything of useful quality to the state of the
art in some given field, and that their work is inherently inferior to
that of professionals.

Cheers,
aj

--
Anthony Towns <aj(at)humbug(dot)org(dot)au> <http://azure.humbug.org.au/~aj/>
I don't speak for anyone save myself. GPG signed mail preferred.

Australian DMCA (the Digital Agenda Amendments) Under Review!
-- http://azure.humbug.org.au/~aj/blog/copyright/digitalagenda

On Fri, Nov 07, 2003 at 08:42:34AM -0500, Jimmy Kaplowitz wrote:
> Clarification: true Condorcet does allow you to say "I have no
> preference between A and B." (That's what a vote of CD or DC would say.)
> It does not allow you to say "I prefer D equally much over A, B, and C."

Hm, I might be wrong about this, and votes for one candidate might be
possible to interpret like that. I'm really not sure; Condorcet is
complicated in its excellence. But in any case, the instructions are
quite clearly stated on the voting page, and in my first reply to Ian I
list a large number of reasons why an action on the part of the board to
restart the election under different rules would be a huge mistake. I
stand by those comments.

- Jimmy Kaplowitz
jimmy(at)debian(dot)org

On Fri, Nov 07, 2003 at 01:05:01PM +0000, Ian Jackson wrote:
> I am very disappointed, Wichert, that despite detailed explanations,
> and clear consensus to use a more sensible system, you haven't heeded
> everyone's wishes.

> I'm now wondering whether the Board should pass an emergency
> resolution suspending the election and asking you to restart it with a
> more sensible interpretation of ballots.

I do not consider the current system to be unfair, and I think that the harm
caused by delaying *yet again* this election would far outweigh any
benefits -- given that ths system is not unfair.

-- John

On Sat, Nov 08, 2003 at 12:13:12AM +1000, Anthony Towns wrote:
> On Fri, Nov 07, 2003 at 08:42:34AM -0500, Jimmy Kaplowitz wrote:
> > [...] and among the election methods cognoscenti it is
> > the most recommended method for fair elections. The same is not true for
> > Debian's method; it has been discussed by laymen who haven't really done
> > a rigorous or detailed study of all the ramifications, [...]
>
> I love it when free software hackers explain how a group of amateurs
> can't possibly contribute anything of useful quality to the state of the
> art in some given field, and that their work is inherently inferior to
> that of professionals.

They can possibly contribute if they try, of course, but so far I have
seen no one trying to provide a rigorous proof that the Debian's method
is not flawed.

This is just a matter of independent interest: Debian is interested in
finding a method that fit its need whereas ``professionals'', as you
call them, are interested in certifying election methods.

Cheers,
Bill

John Goerzen writes ("Re: Call for votes for SPI board of directors membership election"):
> On Fri, Nov 07, 2003 at 01:05:01PM +0000, Ian Jackson wrote:
> > I'm now wondering whether the Board should pass an emergency
> > resolution suspending the election and asking you to restart it with a
> > more sensible interpretation of ballots.
>
> I do not consider the current system to be unfair, and I think that the harm
> caused by delaying *yet again* this election would far outweigh any
> benefits -- given that ths system is not unfair.

You (and others who have said similar things) are right; you've
answered my question for me by convincing me the answer is no.

We should revisit this question before the next time we have an
election.

Ian.