Forum OpenACS Q&A: Proposal for election process

Collapse
Posted by Joel Aufrecht on
1) Nominee statements
Each nominee may provide up to 256 bytes of HTML text to be displayed
in the candidate list.  Each nominee is provided an ETP page hosted on
openacs.org.  Nominees must provide final ballot statements within 6
hours of the end of the nomination process.  Nominees may modify their
ETP pages at any time.

2) Ballots
A ballot consists of an email to the election mail alias.  A ballot
template, consisting of a list of all candidate names, separated by
line breaks, in random order, will be provided on openacs.org.  A
voted ballot body is a list of up to nine names from the original
list, separated by line breaks.  A ballot with more than nine names is
a spoiled ballot.  A line which does not exactly match one of the
original names is null, but does not spoil the ballot.  A line which
repeats a name on the same ballot is null, but does not spoil the ballot.

3) Election period
The election will commence 24 hours after the close of the nomination
process, and last for 48 hours.  Ballots (ie, emails) recieved before
or after this period are not considered.  Ballots from email addresses
not on the voter pool list are void.  If more than one ballot is
received from a single email, the last one is counted.

4) Counting
Each time a candidate's name appears on a ballot, the candidate
recieves one vote.  The nine candidates with the most votes win.

5) Special terms
The four winning candidates with the least votes are elected to
six-month, rather than one-year, terms.  This is establish the
staggered election of the OCT; in the future, all terms are intended
to last for one year.

6) One winner per organization
If more than one person from one organization (company, foundation,
university) is elected, all votes for all such persons except the one
with the most votes within that set are nullified.

Collapse
Posted by Tom Jackson on

If you provide a tenth 'alternate' slot, you can handle the case of a tie. In the case of a tie of 2 or more, use the alternate slot to count which of the tiees get the most votes. In this case you have several places where ties make a difference.

Btw, why are you not using cumulative voting? Why can't someone vote more than once for the same person? This is the most democratic method of voting, no? It supports minority viewpoints best as a small group could insure a voice on the committee.

Collapse
Posted by Tom Jackson on

Sorry, here is a better way to do instant runoff voting: you rank your choices on the ballot. In each counting round, one candidate is eliminated. You stop when you get to nine remaining candidates. It shouldn't be too hard for a bunch of programmers to figure out how to do this.

Also the idea of splitting the vote in half every 6 months has the same effect of concentrating power into fewer hands. Essentially only the top four or five groups will be supported, after the first six months. If you really want nine voices, vote in nine equal seats. After all, they are all equal.

Collapse
Posted by Joel Aufrecht on
IRV is "A Fairer Way to Conduct Single-Winner Elections". It's not immediately applicable to multi-winner elections. If you have a hundred ballots, each with somewhere between 1 and 23 ranked candidates, how do you eliminate one candidate per round?

I think we should ulmitately go to Condorcet or something, but even normal Condorcet is for single-winner. Which systems are best for multi-winner elections?

Collapse
Posted by Tom Jackson on

Joel, in the case of using IRV for multi-winner elections, you just stop eliminating when you get to the number of winners you want.

Anyway IRV seems way too complicated. My points are these:

  • Staggered elections = dilution of minority votes, this is the same thing, but not quite as bad as voting via precinct instead of voting at large. There is also no justification for this. Why vote every six months?
  • One person One vote also isn't completely fair. If forces you to vote for people you don't want to, or wastes at least part of your vote. For a nine seat race, each person gets nine votes, why can't they just vote for one person (they get nine) or two (each gets 4.5) etc.? (That is one way) The other way is to count each vote: 5 for her 4 for him. This allows, again minority voice, otherwise I can only vote for my candidate once for 9 seats, and my chances of helping him/her and getting a voice are diluted to 1/9.

I think if we are going to all this process, with all these rules, why not choose some more modern method, one more likely to get a variety of voices?

Collapse
Posted by Joel Aufrecht on
The upside to staggered votes is that some membership of the OCT team is guaranteed to remain constant over any period less than a year. Do you think that dilution of minority votes is likely to be a serious problem? If so, how could we address it without losing continuity? Three staggers of 3 people each, and elections every 4 months?

Fair, as I've seen it best defined, is "the choice that the most people want, wins," which in a situation with more than two choices means, "the choice that wins more than other choices in pairwise comparison, wins." And the mathematicans report that Condorcet is the best way to determine this for many-choice, single-winner. But I haven't seen any research on how that applies to many-choice, many-winner. Can you point me to some, or better yet, summarize? If we stack votes (5 votes for a top choice, 4 for a second choice), my intuition is that we are then open to more gaming, and potentially less fairness by the above definition. And we wouldn't be able to produce a count that says, 55 people out of 180 wanted X as a team member.

Collapse
Posted by Tom Jackson on

In a nine person race, fair means that a constituency which is composed of only a little more than 1/9th the voting population has a 'fair' chance of getting one seat. By dividing the election into a 5 person race and a 4 person race, the constituency has to be composed of at little more than 1/5 or 1/4 the voting population. Dividing the race into three would make matters worse.

You are always going to have continuity. What is the chance that everyone will be voted out? And if they are, maybe that is also a good thing? Continuity is something that can be decided by a vote, it doesn't have to be enforced by process.

Again, my points are not that we should go with the 'most provably fair' system possible. But the two details I have pointed out are provably unfair to minority voices. Note that one, that is 1/9, vote of 'no' on any TIP, forces 2/3 voting. That means the intent it to support minority voices down to this level, and the two voting issues I am pointing out work against that intent.

Collapse
Posted by Carl Coryell-Martin on
I am fond of approval voting myself.
http://alum.mit.edu/ne/whatmatters/200211/

and for multi-candidate elections we should take the top N candidates.

Collapse
Posted by Tom Jackson on
That sounds like a pretty good system. Easy to do.
Collapse
Posted by Joel Aufrecht on
I have added approval voting to the TIP as an option for the OCT members to select when approving the TIP.
Collapse
Posted by Joel Aufrecht on
Note also: "While AV is a strikingly simple election reform for finding consensus choices in single-winner elections, in elections with more than one winner -- such as for a council or a legislature -- AV would not be desirable if the goal is to mirror a diversity of views, especially of minorities; for this purpose, other voting systems should be considered." (from http://alum.mit.edu/ne/whatmatters/200211/).
I don't think there's an obvious right answer, and I'm no longer worried, if I ever was, about getting this part exactly right.  I'm more interested in having high participation, a simple and comprehensible result, and an active new OCT.  Tom, I think these your points are very good and we should keep working on the issues and seeing what works.
Collapse
Posted by Tom Jackson on

Joel, I think that is the impression I got. It is simple, mostly directed at getting centrist representation. It isn't really geared for including minority voices in an at large multi winner race.

Probably the simplist way to get half way there is to replace all members at the same time. This also may not be necessary for the following reason: our community is not very large. There are also organizational limits, which otherwise might lead to some kind of ballot box stuffing.

I'm still not conviced that having an election every six months is worth the extra effort. In the end everyone will serve one year terms, but we will have to expend half the effort. Working relationships will need to be adjusted every six months as new members join and have to be 'educated'. But again, this is a small community. Probably anyone elected will already have working relationships with the other members.

Collapse
Posted by Mark Aufflick on
fwiw - i like the idea of the staggered elections. in governmental politics we regularly see the disruption to projects and people through the radical shift in policies when a party replaces it's ideological opponent.

i think this idea will force an increase of harmony in policies, since everyone will have an overlap with half of the "old guard", and is also useful to help train up new members (in the unlikely event that no existing OCT members are returned).