W3C home > Mailing lists > Public > public-w3process@w3.org > May 2014

RE: Voting again…

From: Michael Champion (MS OPEN TECH) <Michael.Champion@microsoft.com>
Date: Sun, 4 May 2014 19:38:13 +0000
To: Charles McCathie Nevile <chaals@yandex-team.ru>, "public-w3process@w3.org" <public-w3process@w3.org>
Message-ID: <1399232292588.8988@microsoft.com>
Hmmm...  The turnout in these elections is something like 20%. Most AC reps have  barely enough information about the candidates to make binary decisions about them, much less rank them. If one wants to ensure that their preferred candidate is elected, it would be much more effective to help inform the AC about his/her qualities  and turn out the vote of like-minded people than to cast a single "strategic" vote that is statistically quite unlikely to change the outcome.

I don't object to using a ranked voting system even if I  don't care enough to invest any energy in making it happen.  I do care that decisions in W3C are made by an informed consensus, which is why I keep asking for solid information and analysis of why using a more complex system would actually benefit W3C.

Could you remind us (with links) about the various AB and AC discussions of this topic over the last couple of years?  I vaguely recall there were a handful of people who care deeply about this and a much larger group who were open minded but wanted to see some research/analysis and consensus building on which of the various alternatives would work best for W3C.   Wasn't someone going to start a CG or an AC task force to research this and bring back recommendations?   

________________________________________
From: Charles McCathie Nevile <chaals@yandex-team.ru>
Sent: Sunday, May 4, 2014 10:04 AM
To: public-w3process@w3.org
Subject: Voting again…

TL;DR: AB (and TAG) elections are broken, if we find out exactly how
broken it won't really affect the ability of each group to do their work,
but it isn't hard to fix this and we should. A simiarly broken system is
being proposed for Webizens, whereit is likely to be even more important
to have a decent system.

There is an election (for the Advisory Board), where there are a dozen
candidates. As far as I can tell all of them are qualified for the
position, although of course Ihave preferences for whom I would prefer to
see on the AB.

But the existing system for voting makes a total mess of this. Given
candidates of roughly equal quality, more than half the votes cast will be
completely wasted. If we see (as we have in the last two TAG elections and
the last AB election) a campaign for a slate of voters, even if they are
running on a single "party" platform and proposing exactly the same
things, the system is likely to elect the entire slate, rather than a
diverse group of people representing the diverse preferences of the whole
AC.

The only sensible way to vote, then, is to vote for one candidate. I waste
80% of my right to vote, but increase the odds that my vote is actually
useful in helping elect one of the people I *really* want on the AB. I do
this repeatedly, but reluctantly, as the least-worst of a set of bad
options in how to vote under the current system.

This is why for the last few years (and more so since the beginning of
last year when it started being an issue in practice as slates of
candidates were run in elections) I have been trying to get some traction
on a change in the voting procedures.

There have been a number of arguments raised against such a change, and I
would like to address them:

== It ain't broke, why fix it?

When a slate of candidates standing on a single position ("reform all the
thingz") gets elected in each of the last 3 elections it may mean that we
face only one important problem. I doubt that - and as a candidate (and
someone many other candidates have turned to for advice) I understand the
pressure to align oneself with that "platform". I think this is an
anti-democratic result, and that we are seeing an artificial reduction in
the diversity of the AB, and how representative they really are of the
membership.

In addition I *believe* that an increasing number of votes are "strategic"
- people voting for far fewer candidates than they could, in order to try
and maximise the chances that their vote counts at all. Without an open
count it is hard to know for certain, but the W3C team at least has those
numbers.

== Don't deligitimise the AB / TAG

I find this argument illogical. The voting system we use effectively
deligitimises the results as being representative of us, the AC. This is a
simple mathematical property of a choice made a decade or so ago, based on
a aprticular view of a century and a half of research into and deployment
of voting systems designed to ensure fair outcomes (and also those
designed to favour incumbents or other sectional interests).

While running an experimental vote system that showed the results we get
in our elections are not a very good reflection of the real intentions of
the community would indeed suggest that we have "the wrong people" on the
TAG or AB, I don't think this is a huge problem in practice. First,
because while I believe the AC gets misrepresented by election results, I
also believe that the candidates who are elected but would not have been
under a fairer system are actually widey regarded as well-qualified for
the task.

In other words, what we have now is a somewhat random system for selecting
representatives, that doesn't actually choose what we want, but whose
impact is balanced by the fact that at least most candidates are
reasonable. In which case, why bother with the election instead of a
random selection?

== No voting system is perfect, so why swap?

Arrow's theorem is one of those that demonstrates that it is effectively
impossible to get a system that cannot be gamed (e.g. allowing a slate of
candidates to effectively shut out diversity), and that entirely reliably
represents the interests of the voters.

However, it is one piece of a large body of work going back to at least
the mid 19th century on how to provide *better* systems. And it is
generally accepted that some systems are significantly better than others
- and we happen to have on of the worst.

== STV is really complicated…

Well, it does require the ability to rank candidates in order of
preference. Which in turn means being able to count perhaps as far as 12.
I believe that all members of the AC can manage that task without getting
terribly confused.

Even the most complex voting systems could be hand-calculated for the AC -
but luckily we also have computers that can simplify this process,
providing a result effectively instantaneously for an election the size of
those we run in W3C.

== How do we select a new system?

Obviously, unless you have a decent voting procedure in place, you're
unlikely to get one by voting on it! However, while I think the level of
candidates for W3C elections is generally high enough that it isn't a big
deal whether we vote or pull names out of a hat, that isn't true of voting
systems.

Pretty much any STV or ranked-pair voting system is so much better than
the current one that it would be a big improvement.

However, David Baron did some research a while ago. He expressed interest
in approval voting[0], which unfortunately is still subject to most of the
problems of our current system. His eventual conclusion (which matches
mine, unless I have misinterpreted him as merely proposing a strawman) was
that there are a couple of systems seriously worth further consideration.

Of those David suggested were most worth further investigation, the
"qualification" is that they are highly resistant to managed voting,
ensuring that the best voting strategy is to say exactly what you want.
And of them, the Shulze system[1] has a couple of very interesting
properties:
  - You can try it on the Web[2], where there is an explanation of how it
is a big improvement, and
  - there is open-source code[3] to implement the counting, which means we
would require a minimal effort to implement it in e.g. WBS (assuming we
don't 'outsource' our voting to the open web).

Since David's suggestion and subsequent discussion a year ago, I haven't
seen *any* suggestion that we should not use the Schulze system in any
discussion. I would like to know if there are any alternative candidates.

== Conclusion

There should be a call for consensus on whether we should adopt an STV
system for voting. At this stage, I am thinking of simply an up/down vote
on the Schulze system (primarily to help defuse the "but it's all so
complicated - and which system would we choose" argument that is
occasionally raised), but before I do anything I welcome further thoughts,
suggestions, ideas and opinions.

[1] http://en.wikipedia.org/wiki/Schulze_STV
[2] http://www.modernballots.com/
[3]
https://github.com/bradbeattie/python-vote-core/blob/master/pyvotecore/schulze_stv.py

cheers

Chaals

--
Charles McCathie Nevile - Consultant (web standards) CTO Office, Yandex
       chaals@yandex-team.ru         Find more at http://yandex.com
Received on Sunday, 4 May 2014 19:38:45 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 20:35:10 UTC