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

Re: Condorcet Voting and Supermajorities (Re: [CONSTITUTIONAL AMENDMENT] Disambiguation of 4.1.5)

On Sun, Dec 10, 2000 at 07:48:07AM -0500, Raul Miller wrote:
> On Sun, Dec 10, 2000 at 08:16:13PM +1000, Anthony Towns wrote:
> > It's not like your interpretation (of supermajorities in particular,
> > but also of cyclic tie-breaking) has ever actually been used before,
> > either within Debian or without.
> On the "different winners of the same vote" issue:
> [1] I don't see that you've shown my ideas to be flawed in this fashion.

Of three possible interpretations of the constitution (one clearly valid,
two of questionable validity, labeled "mine" (the 6 repeated final
ballots) and "yours"), the former and mine end up with the same results
given the same opinions of voters, yours ends up with different results
in a number of cases, without introducing any (further) ambiguities.

No, your system isn't random: it won't come up with different winners
to the same vote; but that's a pretty pointless criteria. Choosing the
first option alphabetically will manage that.

> [2] If my interpretation of the constitution is equivalent to your
> interpretation of the constitution where there's no cyclic tie,

It isn't. It's at best equivalent when there are no cyclic ties and no
differing supermajority requirements. There may also be other differences
that we never got around to looking at.

> On the philosopic side of things:
> It's not like anyone else has designed a real voting system to be used
> for an email participation context.

I'm not sure how much overemphasis you're putting on the word "email"
in the above, but there certainly are plenty of other voting systems
out there.

> If you want us to fall back on the sort of voting systems used in more
> traditional contexts: Supermajority would be described as a different
> kind of ratio: how much of the entire body must be in favor of the option
> before it can pass.

ie, n:m vecomes n/(n+m). That's, err, not a terribly fundamental change.

And it's yet again being inexplicit about what "in favour of the
option" means, although since we're apparently talking about a general
principle or something, that's somewhat understandable.

A more useful description of supermajority might be by reversing the
ratio, and saying that `a 2/3 supermajority means that a 1/3 minority
can veto the resolution'.

> What I see is that debian's system is designed to work on the idea that
> the people interested in a subject are the people who are competent
> to decide on that subject,

No one ever suggested making voting compulsory for developers.

> That's funamentally different from the usual sort of voting system,
> which assumes universal participation of some sort, and doesn't presume
> that anything other than voter opinion indicates that one idea is better
> than any others.

And that doesn't follow at all. What, other than voter opinion, would
you like to base a vote's results on?

> I'll agree that my interpretation of supermajority is different from the
> tradition, but I think that, philosophically, mine aligns closer to the
> way debian has made choices in the past.

That's possible. ``If the majority are being stupid, then a minority
shall form a Cabal, manipulate the system to their own ends, and get
what they want anyway.''

> Also, the weights assigned to votes at different preferences is
> fundamentally arbitrary:

No, it's not. It influences how people have to vote in order to get what
they want. To take a topical (and extreme) example, the US presidential
elections give a weight of 1 to the first preference, and 0 to all other
preferences, so someone with since preferences:
	1: Nader, 2: Gore, 3: Bush
is obliged to think carefully about their vote, and probably change it to
	1: Gore, ...
depending on whether he or she thinks Nader will get any use out of a vote
for him (since Gore almost certainly will).

The point of selecting a good voting system is to make sure voters get
an optimal compromise without having to think carefully about whether
they should lie about what option they'd really prefer to win.

> > It's not like it's the end result of months of discussion by
> > experts, and years of electoral research. It's not even like it's
> > the widely held consensus of a bunch of non-experts. 
> Ok, point me at some references to expert discussion on voting systems
> with Debian's philosophic base.

``Look, all these docs tell you how to create secure temp files in /tmp,
while I want to create them in /var/tmp! What good are they? What a
waste of time. I'll just use fopen(), it's easy and understandable.''

Whatever. There's plenty of actual criteria you can actually test a
voting method against in previous messages. If you want to ignore those
and instead base your measure of how good a voting system is solely on
how closely it matches your reading of the constitution, you're both
welcome and deluded.


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

     ``Thanks to all avid pokers out there''
                       -- linux.conf.au, 17-20 January 2001

Attachment: pgpZiKmYmpN8n.pgp
Description: PGP signature

Reply to: