ICANN/DNSO
DNSO Mailling lists archives

[ga-full]


<<< Chronological Index >>>    <<< Thread Index >>>

Re: [ga] Voting rules, take 4


	I've been letting my email slide, but I sat down today and read the entire
[ga] voting thread.  Some thoughts:

	1. It's not well taken, I think, to tell Eric that we can ignore his
proposals aside in passing version 1.0 now, and that he should put them
forward as proposals for version 1.1.  The reason is that Harald's ruleset
includes a provision designed to make the rules *hard* to change in
subsequent versions -- it provides that the rules cannot be changed, once
adopted, without a supermajority and a special quorum.  The point of
Harald's Rule 5 is to promote stability by requiring that even proposed
changes *with the support of a majority of the membership* cannot be
enacted unless they surmount the 2/3 hurdle.  So this is the last time that
competing proposals will contend on a level playing field; once we first
adopt a ruleset, whatever we have already adopted will have a huge
procedural advantage over any proposed changes.  As a result, it's worth
focusing on Eric's proposal now.

	2. I think people aren't paying enough attention to the differences
between candidate elections and issue elections.  There are good arguments
for having fancy voting rules in candidate elections.  The argument for
having elaborate procedures to compute preferences in issue ballots is much
weaker, b/c the prime source of manipulation of those ballots has nothing
to do with vote counting -- it comes in the decisions as to how to present
the various alternatives for voting.  It's silly to give the chair broad
power to word the proposals being presented to the electorate and then
adopt elaborate procedures for counting the votes.  To address the fear
that the chair will manipulate issue ballots, you need something like
Roberts' Rules (which have their own problems, and which I'm not
advocating), not vote-counting mechanisms.  In any event, I don't think IRV
is especially useful for issue ballots, in part because it's designed to
produce a "winner" no matter how splintered voter preferences are.

	3. So what's the best approach for single-member candidate elections?
Harald has proposed that if a candidate gets 50%, he or she wins;
otherwise, there will be a runoff with a smaller number of candidates, so
that someone can get 50% in the smaller field.  Eric has proposed that if a
candidate gets 50%, he or she wins; otherwise, the IRV vote-counting
mechanism will simulate a runoff with one fewer candidate, so that someone
can get 50% in the smaller field (and it will iterate this until the
simulated runoff really does give somebody 50%).

	The most striking thing about these proposals is how alike they are.  They
differ in two ways.  [1] Under Harald's proposal, the chair has more
discretion as to how to structure the runoff.  For example, he might choose
to eliminate several candidates at once, rather than holding sequential
runoffs eliminating them one at a time.  The IRV mechanism, by contrast, is
mechanical.  [2] Under Harald's proposal, we actually have to hold repeated
elections.  IRV, on the other hand, requires only one ballot, and simulates
all necessary runoffs.  I think each of these proposals has advantages.  I
expect that either would probably be fine.

Jon


Jonathan Weinberg
weinberg@msen.com
--
This message was passed to you via the ga-full@dnso.org list.
Send mail to majordomo@dnso.org to unsubscribe
("unsubscribe ga-full" in the body of the message).
Archives at http://www.dnso.org/archives.html



<<< Chronological Index >>>    <<< Thread Index >>>