Condorcet - Ranking vote for single-winner and proportional multi-winner election
Armando Pitocco
Public
Seen by 69
My question is: does Loomio community think to introduce ranking vote one day?
FOR WHO doesn't have any clue of what is ranking vote, Condorcet or Schulze completion, single and multiple-winner elections, go on. Excuse me for my rough explanation, I hope it could be useful.
First, here two excellent open source platforms that allow ranking vote:
https://modernballots.com and http://civs.cs.cornell.edu
(they work for single-winner and for multiple-winner elections, both let use Schulze completion)
How it looks like? Here a drag and drop interface (from Liquid Feedback) https://www.youtube.com/watch?v=4Nz0HdX7xRo
I'll try to point out some qualities of these method(S) in my opinion.
Ranking vote (through Condorcet methods) help people finding consensus.
It is useful when we have more than 2 options and we have to choose the best winner for everybody.
The present Loomio Poll is useful to understand which is the 1st choice of everybody.
But rarely in this kind of elections 1 option (among 3 or more options) reaches 50%+1.
So "who wins" is unclear.
"The majority want something Orange? Joke's on them. Blue wins"
Condorcet works in this way:
Imagine we have 3 options, A, B, C. if we compare couple of options (e.g. A vs B ) who wins is clear. So let's compare any possible combinations: e.g. A vs B, A vs C, B vs C.
Who wins every comparison is the *Condorcet winner*.
Sometimes even in this way is possible there is no clear winner (e.g. A wins B that wins C that wins A... Condorcet Paradox), so many methods have been invented to solve the paradox and pick the best winner as possible.
Actually finding the Condorcet's winner through a ranking vote is something like the iterative process that you could do with many Loomio "Proposals" in order to reach a larger consensus.
With Condorcet there are no binary decision, there is always place for other solutions: tertium semper datur.
E.g. in a Red vs Blue election, someone can propose a Purple solution. And there could be a "more reddish purple" and a "more bluish purple"... the existence of many variants doens't "split" the voters weakening the nearest proposals.
It favors the creative resolutions of conflicts, in my opinion.
Multiple-winner Condorcet
And what about if we have to choose more than 1 winner?
Some variations of Condorcet allow to get proportional results, that is if we have 3 winners they have to represent the largest part of voters and not just the majority.
Here some easy explanations https://modernballots.com/help/multiple-winners/
The basic idea is: instead of 1 winner we have a combination of winners. So let's take the combination of winners, among all the possible combinations, that wins against any other combination: i.e. the Condorcet winner combination
"Want to vote for Orange? Too bad. Vote for Teal or Blue wins"
[EDIT: images come from modernballots.com and try to criticise "traditional" voting methods]
Felipe Cavalcanti Sun 3 Sep 2017 3:35PM
Don´t yoiu think dot vote has similar features? https://www.loomio.org/p/eMW8VOTg/club-support?from=0&per=25
Rudolf Hornig Sun 3 Sep 2017 4:09PM
No. Dot voting is fine if I want to set a priority of items. Like in what order I want to execute some tasks. The main reason I prefer Condorcet voting is that it frees the voters mind about tactical considerations.
If I have 3 choices A,B,C and I'm absolutely fine with both A and B but strongly against C, in a Condorcet voting I would put both A and B as my first preference without any hesitation in a honest way.
On the other hand, with dot voting I would start immediately speculating whether splitting my dots (e.g. 8 dots) between A and B is a good strategy as it could probably help C to win. Like, if at the end the result is C:298, A:296, B:180 and I split my 8 dots between A and B (as 4:4) I would immediately feel the pain that I could help A to win if I had not split my vote, but instead used all my 8 dots to vote for A. In this case the result would be C:298 A:300 and B:176. And remember, I strongly disapproved choice C...
Long story short, dote voting is absolutely encourages tactical voting.
James Kiesel Sun 3 Sep 2017 11:28PM
I'll hop in here and say that we are considering developing out a ranked choice polling type. However, right now we're not considering something more complicated than 'give users a list of options, and let them order them by preference'... we're not going to develop scoring algorithms which declare winners the way Instant Runoff / Condorcet does.
The reason for this is that as a tool, I'd like Loomio to be most concerned with soliciting and tabulating opinions, while the actual decision-making is made by those who are best at it, namely people.
Just as it's left up to the group to decide what '5 agrees and 2 disagrees', or '10 yeses and 5 nos' means in the context of your group, it makes sense for us to tell you 'Option A got 4 first place votes and 2 second place votes, Option B got 1 first place vote and 6 second place votes', etc... and let you, the actual decision-makers, do with that information as you wish. :D
Rudolf Hornig Mon 4 Sep 2017 8:51AM
Thanks for the info.
As long as there is an easy way to export the results in a machine readable form, I would be fine with this. I can see the reasoning behind this decision. Only one request: please make the ranking UI in a way that users can specify two or more choices to be on the same rank. I.e. I prefer first A and C (whichever, I don't care) then as my second preference B and as third D.
Something like the liquid feedback UI would be cool: https://www.youtube.com/watch?v=4Nz0HdX7xRo
The result processing could be done off-line as you've suggested.
Armando Pitocco Mon 4 Sep 2017 12:19PM
I'd like Loomio to be most concerned with soliciting and tabulating opinions, while the actual decision-making is made by those who are best at it, namely people.
I agree with this point of view.
it makes sense for us to tell you 'Option A got 4 first place votes and 2 second place votes,
Option B got 1 first place vote and 6 second place votes'
Pay attention to just one point: in Condorcet methods what is most important is not the absolute rank of each option, but the relation among them. To let people recount ballots through an external machine you have to save these relations: not just 'Option A got 4 first place votes and 2 second place votes' but 'Option A wins B for 5-4, loses against C for 3-6".
That is, it should be possible to export every single ballot.
Connor Wed 6 Sep 2017 6:43AM
Hi Armando, Mathematician here. Before you go trying to convince Loomio to implement A Condorcet method, you should know that there is already a voting method that is proven to produce results as good or better than any Condorcet system, with much less complication. I suggest you take some time to explore rangevoting.org. On that site you will see how it solves not only the problems with Plurality (as your graphs helpfully demonstrate), but also the problems with ranked choice ballots and even Condorcet methods. Take a look and let me know if you have any questions.
Richard D. Bartlett Wed 6 Sep 2017 9:30AM
I really enjoyed this fun and clever tool for learning about different voting protocols: http://ncase.me/ballot/
Armando Pitocco Wed 6 Sep 2017 1:56PM
Connor @adroit and @richarddbartlett thank you so much for the reading you advised me, I'll study it deeply, hoping to understand.
Armando Pitocco · Sun 3 Sep 2017 1:13PM
Few days ago I talked by mail with @gdpelican about it, and he revealed me that Loomio is thinking to introduce some kind of ranking vote. I like very much this subject, the post is just an attempt to give some basic information, but probably many of you know this stuff better than me!