New polling type: Ranked Choice
Hello friends!
Over at Loomio HQ we've been working on a new polling type per request from several of our paying customers, and we're ready to start rolling it out to a few groups to test it.
We're calling it 'Ranked Choice', and it's a similar to a Concorcet or Instant Runoff style poll, where each participant is able to select their top N options. Here's how it works:
create a poll which has a series of options, similar to our existing polling types
Select how many choices each participant should make (So, if '3' is entered here, we'll count the top 3 choices for each participant)
When you vote, drag and drop your selections in order of preference. Again, only the top 3 choices will count here
When the results are compiled, we present a table displaying each option and how many 1st, 2nd, 3rd, etc. place votes they received.
These types of polls are especially useful when you have a large number of options, or in an election setting where you're concerned with 2 somewhat similar options 'stealing' votes from one another, resulting in a less popular option gaining the plurality.
Feel free to check out an example poll to try it out for yourself, or if you'd like to give this a try with your community, please shoot an email to contact@loomio.org saying so, and we'll flip on this polling type for your group. As ever, excited to hear your feedback and refine this feature further so that we can release it to the world.
Happy voting!
Excellent!
I dare to point some questions/issues, but this new feature is wonderful!
1. in column Total options are scores as in a sort of Borda method, isn't it?
2. Is it possible to export ballots data, in order to allow people to recount with their favourite voting system (e.g. Schulze method or other Condorcet)?
3. As asked by @rhornig, ex-aequo voting would be great, the only missing feature to make it fully compatible with condorcet methods.
4. Make ranked voting compatible with range/score voting: https://modernballots.com for example let use 5 stars ⭐ system to vote in place of ranking with drag and drop. From number of stars you can both give points (range voting) and reorder for preference (ranked voting) if options are 5 or less. It could be an optional feature, in addition to drag and drop option. (@adroit @richarddbartlett as range voting supporter maybe maybe are interested?).
5. Order of options is randomizable? It seems to influence voting beahaviours.
@gdpelican can we enable experiments in this group?
Experiments are now enabled in this group :)
What is your favourite vegetable?
ranked choice by Robert Guthrie Closed Thu 5 Oct 2017 8:41PM
A silly ranked choice to demonstrate the feature. There can only be 3 winners!
Results
| Results | Option | Rank | % of points | Points | Mean | Voters | ||
|---|---|---|---|---|---|---|---|---|
| Artichoke | 1 | 38 | 9 | 3 | 3 | |||
| Potato | 2 | 21 | 5 | 2 | 3 | |||
| Carrot | 3 | 17 | 4 | 1 | 3 | |||
| Cabbage | 4 | 17 | 4 | 2 | 2 | |||
| Celery | 5 | 8 | 2 | 2 | 1 | |||
| Not voted | 0 | 0 | 282 |
| 1 - Artichoke | ||
| 2 - Potato | ||
| 3 - Carrot |
I love Artichoke but you can't deny that Potatoes rule the world!
| 1 - Artichoke | ||
| 2 - Celery | ||
| 3 - Cabbage |
Being a vegan I like them all! However I appreciate the ranking process.
why are results in reverse order?
gives the impression (to me) that 3 is top (maybe add st nd rd?)
Oh, I see what you mean.. in the individual stances... they're reversed.. That's a bug.
It is unfortunate to see support for a Ranked Choice Method before Range Voting. Not only is it more complicated, it's less accurate and as Arrow's Impossibility theorem tells us, has unavoidable flaws that are particularly clandestine because most people are only familiar with Plurality :/
I'm not that into Ruby on Rails, but I would definitely be willing to help implement Range in any way I can.
@adroit - Sounds interesting. Can you do some mockups of how that might look within Loomio?
I'll see what I can do. For a reference implementation that's really quite good, check out Voteupapp.com. VoteUp seems to work really well.
As Connor mentioned it would be great to have a score based voting option. RCV as you are doing it here is better but as it's done by Govt. it has some major issues that cause some voters rankings to be ignored, leading to unrepresentative outcomes.
STAR Voting is a great alternative that is highly accurate and rewards honest and expressive voting, breaking the historical trade-off. We're actually running twin ballot measures now to implement it in Oregon election. Having STAR on Loomio would really help spread the word!
S.T.A.R. stands for Score, then Automatic Runoff. Check out this sample ballot below. In the first round you score the candidates. Those that you leave blank get a zero. If you don’t have a preference between candidates you can give them the same score.
You turn in your ballot and everybody’s scores are totaled. The two highest scoring candidates are finalists who advance to the automatic runoff. The runoff is one-person-one-vote. Your ballot already shows which finalist you scored higher, and the runoff throws your full weight behind the finalist you preferred.
Is it possible to let export voters' ballots?
In this way for who needs Condorcet method is possible to upload in an external tool in order to find the condorcet winners.
Each voter's ballot should save relationship among ranked options (1st A, 2nd C, 3rd B and so on)
We get asked for this all the time, heh. I reckon it'll get done very soon, possibly over the holidays. 🎄
If you're going to use a ranked ballot you need to make sure the ballots are counted in a way that actually uses all the info off of each ballot. (Instant Runoff Voting ignores some of people's rankings and can have skewed results.)
It's misleading to have a system that makes it seem like you are getting to vote on more than one option but that may not even count your down ballot rankings. :(
Ranked Pairs is a really good ranking system! Star Voting is a really good scoring system.
I'm a bit confused about this; I think there's a perfectly valid use case for having, say, a poll with 12 options where you pick your favorite three... it's a bit unreasonable to expect a voter to truly be able to rank that many options, and if we count down-ballot votes in that case you're likely to end up with aberrations where people don't really care about the distinction between, say, their 8th and 10th choice, but the way that those votes are arranged can have a significant impact on the final results.
Of course, this setting is up to the poll creator, so if you have a situation where you expect your voters to rank all of the options and have them counted, you're perfectly able to do so, and we make it super-clear in the interface how much of the ballot we're actually counting.
for "lazy voters" (or, otherwise, too verbose poll) a good feature could be allow ex-aequo, i.e. equally ranking of more options.
In this case a manner to simplify the ranking could be to put a limited number of ranks with a score voting style.
For example you can realize with 5 stars: rank #1 is 5 stars, last rank is 0 stars. If you have 6 or less options, actually you are allowing people to accurate rank all their options in different ranks, if tey want. If you have 7 or more you are forcing people to vote some options with the same rank. But it gets far more user-friendly.
Example
* 1 Option I love ⭐ ⭐ ⭐ ⭐ ⭐
* 2 Option I can tolerate ⭐
* 3 Indifferent to this option ⭐ ⭐
* 4 Really bad option (0 star)
* 5 Very good option ⭐ ⭐ ⭐ ⭐
* 6 Normal ⭐ ⭐ ⭐
* 7 I don't know ⭐
* 8 Not so bad ⭐ ⭐ ⭐
The machine could easily translate it in an accurate ranking:
* 1st op.1
* 2nd op. 5
* 3rd op. 6 and 8
* 4th op. 3
* 5th op. 2 and 7
* 6th op. 0
The way we do it for STAR Voting (Score-Then-Automatic-Runoff with a 5 star ballot,) is to always count all the data from every ballot. Otherwise you'd be wasting people's votes! Eek. Any candidate left blank gets a 0. You can give a score or rating to as many or as few candidates as you like. If you don't like a candidate, don't know about them, or don't care, it's totally fine to leave that candidate blank, but if you do have an opinion or preference STAR Voting lets you express it.
Even if only some voters chose to use the full expressive potential it will improve results overall. This system really incentivizes honest and well educated voting. Contrast that with other systems that don't let you show a full opinion or that don't necessarily use all the info you put on your ballot... why even bother? No wonder voter turnout is so low.
I applaud the introduction ranked voting. The drag and drop interface and layouting still needs a bit fine tuning. Dragging looks a bit strange, especially if I grab the option on the left side, but other that that, nice.
If I could have a wish, it would be useful to have the ability to rank an option to the same priority, meaning that I do not have preference between the two. If I want to eliminate tactical voting, I need to allow this otherwise if I don't have a preference between A and B I will always start wondering which should be placed as first and which as second. Will the A,B or B,A help better for A or B to win? How will other people vote?
With this approach I'm adding and artificial difference between A and B which is not there in reality.
I see the cases when exclusive ordering is required. For example if I have to do A and B and C tasks in some order, it makes no sense to place A and B at the same priority level. Probably and additional checkmark could be used to allow placing different option to the same preference level.
Additionally I would love to see an easy way to export the whole dataset (per voter) in CSV file so it could be processed in an external too, to evaluate it and get a real condorcet/sultze result (even if loomio itself does not support it directly).
Thanks a lot for implementig this.