[mnomitch]

Hi! I actually made a similar site for my own purposes - https://poller.io - and I've been using it with friends for a while.

One interesting finding. I was expecting that Instant Runoff would be the method we all preferred. I also included plurality winner, Borda Count and the Condorcet winner (if there was one) in the list of results. After using this with friends for decisions, I realized that a lot of choices that did very well in borda count were getting eliminated in the IRV. If you have a lot of options in the poll, there's a good chance that an option that will make the majority happy gets eliminated because it has very few first place votes. Academically, I knew this was a possibility, but in practice it happened a ton. This made us change to borda counts as our method of choice for things (and then we promised to not game the system). But... if there's a Condorcet Winner, we always go with that.

I also thought that having a borda count as the method of eliminating the "last place" choice in each IRV round would be kinda nifty. This still lets people game the system though, and is kind of a gross hack.

Just letting you know our experience in case it is helpful for future options on polls! Good luck!

[DominikPeters]

The resulting voting rule (Borda except if there is a Condorcet winner) is commonly known as Black's rule (https://en.wikipedia.org/wiki/Black%27s_method).

[tunesmith]

KUDOS to you for respecting the Condorcet Winner.

After spending a notable portion of my life diving deeply into voting systems, their side effects, and the various poorly-thought-out arguments that people make, I basically only feel wincing pain when a lot of these projects come out. So I'm glad that the concept of the Condorcet Winner hasn't been forgotten.

The IRV zealots have basically won the messaging war, using what I believe to be unethical tactics. IRV is not the only way to count RCV, but they act like it is. But power is power, so until society suffers enough with lousy results, it's just going to be the FPTP pain all over again, dressed up in different clothing.

[ClayShentrup]

cordorcet methods are overly complicated and don't optimize the right thing. the goal is maximal social utility in the face of strategic behavior. cardinal methods are simpler and generally better at this.

https://electionscience.github.io/vse-sim/VSEbasic/

[tunesmith]

> cordorcet methods are overly complicated

https://news.ycombinator.com/item?id=33111997

> and don't optimize the right thing

Apparently "candidate that would beat all others head to head should win" isn't the right thing? You'll need to explain that.

> maximal social utility

Some folks define social utility as erring toward more passionate voters, which is a potential problem with all cardinal methods. If they don't elect the Condorcet Winner, they are fatally flawed for any election that purports to be democratic.

[clolege]

Isn't Borda Count just Score voting where you enforce that voters can't give two options the same amount of support?

I would be curious why people would prefer Borda over Score if so.

[wnoise]

Not at all. Score Voting as usually defined is continuous on [0,1]. Rescaling to e.g. [1,5] doesn't change this. Restricting to integers on that does; but if there are fewer than 5 candidates, equivalent rankings can still differ for total scores. It's only equivalent if you pick an integer scale [K, K+N] with N candidates, K arbitrary.

That said, I prefer score, or even approval, as that's what strategic score devolves to, but without huge amounts of information loss.

[clolege]

> if there are fewer than 5 candidates, equivalent rankings can still differ for total scores

I’m having a hard time following, can you give an example?

I’m saying that Borda allows voters to give one candidate 5 points, one other 4 points, etc. and score would allow voters to give as many candidates 5 points, as many candidates 4 points, etc.

[wnoise]

I'm objecting to the statement that they're equivalent if you forbid ties.

This is only true when you have as many candidates as score slots.

[clolege]

> It's only equivalent if you pick an integer scale [K, K+N] with N candidates, K arbitrary.

or [0, K], K arbitrary.

because you can give multiple candidates the same amount of support in Borda if it is 0.

[mnomitch]

I'm not super familiar with score, but the advantage in this scenario is that voters call all input the same way and get borda + IRV + a potential condorcet winner.

If you allowed same scores then you can't do both simultaneously and/or the UX gets really funky.

[clolege]

you can do both simultaneously if you count ranked ballots like an election office would: you throw away all tied rankings, and rankings that come after :)

[awb]

> and then we promised to not game the system

That's the key point. Unfortunately, all known voting systems can be gamed.

Here's an interesting video that explains how they can be manipulated: https://www.youtube.com/watch?v=yhO6jfHPFQU

[tadmilbourn]

Thanks! And I hadn't come across Poller before. Pretty cool. FYI - RankedVote actually uses a type of Borda Count as its tiebreaking mechanism. Not quite what you're saying, but is better than picking a name out of a hat!

You bring up a lot of interesting points on what people feel happy with after voting. None of these systems is perfect. Each can be picked apart in various ways. That's why I tend to frame things in terms of "does this move things forward from the status quo?"

[mnomitch]

Oh interesting. If you don't mind me asking, what is the tie breaking mechanism?

And totally agreed on the framing. Also, prepare yourself for the feedback of "we should have a runoff election with the top two" and then people completely rejecting the concept that you've already done the runoff. :)

[tadmilbourn]

Hehe...a bunch of Nate Cohn's in your friend group? :-)

In the event multiple choices have the same number of votes and those choices are the lowest vote getters in the round, RankedVote calculates a Borda Score for each of those candidates across all ballots that were cast. Basically just saying "Ok, these three candidates were tied, which one was ranked lowest on average across the entire group of voters?" And then eliminate that one.

[RosanaAnaDana]

I'll be using yours over the OP's.

Theirs has a sign in/ sign up wall; yours I get to use right away.

[elihu]

> "I also thought that having a borda count as the method of eliminating the "last place" choice in each IRV round would be kinda nifty."

That is a real thing. I don't remember the name for it though.

[skybrian]

Did you ever try star voting?

[mnomitch]

No, but only because I have been too lazy to update the code.

If I had more free time, I would probably add Star and Approval and then try to use Star with my group of friends. Maybe OP can save me the trouble by adding to his site ! :)

[godelski]

Honestly, for things like "where should we eat" Approval is the method I do with my friends. It's quick and easy. No need to write anything down. Just thumbs up if you like the suggestion and then we take the place with the most thumbs up.

[nerpderp82]

Shouldn't there be some Coq or Agda or gasp Rust code that implements all these voting algos that we can agree on and share?