4 ms·

I haven't reviewed the json in depth but I wonder if the data is sufficient to also test against other vote-counting methods such as Condorcet. BTW, I've alway

I haven't reviewed the json in depth but I wonder if the data is sufficient to also test against other vote-counting methods such as Condorcet. BTW, I've always found the descriptions of Condorcet to be self-defeating, even on Wikipedia (and my edits have been rejected). I've always considered there to be a conceptual separation between the Condorcet method itself, and the various "tie-breaking" schemes that are bolted on if the Smith Set has more than one member. But in the voting literature, they're always communicated together. The reason this causes confusion is that then a "Condorcet Method" can then be described as having this flaw or that, when those flaws almost entirely come up only in those tie-breaking scenarios. So then Condorcet itself gets painted with the same broad brush. But an election with a single-member Smith Set is free of all of those flaws that only apply to the tie-breaker methods. And personally, I don't see why we should be so scared of a multi-member Smith Set. The presence of which is an accurate indicator that the voting population is actually unable to come to a clear decision - that is valuable information! In a Democratic election, I'd honestly be more interested in having a runoff election at that point, rather than trying to bolt on some sort of flawed tie-breaking vote-counting mechanism.

KerrickStaley 4y ago
One could advocate for the adoption of Smith-IRV. It's easy to explain (it's the same as IRV, we just filter it down to the Smith set first waves hand) and it satisfies the Condorcet criterion. (I really like this framing BTW, separating the concepts of the Smith set and the tie-breaking process).
- tunesmith 4y ago
  I'd go for that. As far as I'm concerned, Smith-IRV is better than IRV in every circumstance. The only counterpoint I've seen are twisty arguments about a modified "Majority Criterion", where a single-member Smith Set might not (arguably) be someone that a majority "voted for". Which would seem to apply just as much to IRV as well, but IRV adherents that make this point seem to think that the mechanics of IRV somehow overcome that.
- fdsjgfklsfd 4y ago
  Alternatively: "It's the same as Condorcet, we just add multiple rounds for no reason." Why combine two systems to make a more complicated system when you could just use Condorcet directly?
fwip 4y ago
I think the fact that voters knew they were participating in an instant-runoff-vote would make it difficult to draw strong conclusions about what the results would be under other voting systems.
- tunesmith 4y ago
  The contention that being aware of IRV-vs-Condorcet materially matters in terms of strategic voting is also a pretty unprovable contention. People should be instructed to vote their honest preferences in IRV anyway.
jsmith45 4y ago
I see little point in forcing a separate run-off of Smith Set candidates if there are multiple (i.e. there is no Condorcet winner). After all, we just had everybody indicate their preferences in a ranked list. A runoff could only give new useful information if: 1. one or more voters change their minds. 2. one or more voter gave misleading preference information. 3. The voter's true preferences are not representable as a ranked list (e.g. they Prefer Mr Rock to Ms Scissors, prefer Ms Scissors to Mx Paper, and prefer Mx Paper to Mr Rock.) So i think the tiebreaking methods are reasonable. It is probably also worth noting that for some Condorcet methods, attempting to vote dishonestly to exploit some flaw in the system is has great risk of backfiring. The impact that the dishonest vote will have when a Condorcet winner does exist can differ from the impact it will have if a tiebreaking method come into play, such that it is only safe to try to strategically vote like that if you already know there will be no Condorcet winner. (Which would be very hard to know. Polling can't even reliably predict who will win with single vote elections. Existing polling methodology would be way too under-powered to detect if there will be a Condorcet winner or not).
- jefftk 4y ago
  In addition to voters changing their minds (which is possible based on additional time, media coverage, research, debates, etc) there are other ways a runoff could change the results: * People who previously didn't vote learn about the situation and participate. * People who previously didn't indicate a preference between two candidates think more and vote accordingly.
godelski 4y ago
I personally don't care much about the strong condorcet criteria as long as the weak one is met. Condorcet methods (RP and Schulze are most common) typically get higher voter satisfaction efficiency under ideal conditions (voters being completely honest) but not under strategic conditions. They also aren't winning my much. This[0] gets shared a lot but often people don't pay attention to the spread in VSE over different strategies, which is just as important as max VSE. Though, of course, voting requires much more than just getting a good VSE. If you're a VSE maximizer, than yeah, Condorcet methods make the most sense, but it is naive to approach it as the only criteria. [0] https://electionscience.github.io/vse-sim/ https://electionscience.github.io/vse-sim/
- tunesmith 4y ago
  Let's say that 49% "strongly" support A, and 51% "weakly" support B. If VSE means A has a higher score, I think it's meaningless in terms of democratic elections that follow the one-person-one-vote principle, where B should definitely win. There's a whole field of "voter satisfaction" and "utility" that doesn't really apply to democratic elections where we have to worry about populations bullying other populations.
  - godelski 4y ago
    > Let's say that 49% "strongly" support A, and 51% "weakly" support B. This example does not have enough information to have a optimal VSE solution or even a good guestimate. We don't know the distance between A and B nor percentages of A that are okay with B and vise versa. So essentially we don't know what our latent landscape even looks like. But if you can add to that and figure it out then we can actually capture information about population bullying. The issue with voting is that it appears simple but is rather complex. But complex problems also require nuanced solutions (not necessarily complex, but can't be solved by first order terms).
    - tunesmith 4y ago
      Distance between A and B are subjective according to each voter. If you are (hypothetically) a binary thinker and think A and B are far apart and absolutely prefer A and hate B. And if I am (hypothetically) a nuanced thinker and can see plusses and minuses to both, and after careful consideration I choose B. Then should your vote matter more than mine?

KerrickStaley 4y ago

One could advocate for the adoption of Smith-IRV. It's easy to explain (it's the same as IRV, we just filter it down to the Smith set first waves hand) and it satisfies the Condorcet criterion. (I really like this framing BTW, separating the concepts of the Smith set and the tie-breaking process).

tunesmith 4y ago
I'd go for that. As far as I'm concerned, Smith-IRV is better than IRV in every circumstance. The only counterpoint I've seen are twisty arguments about a modified "Majority Criterion", where a single-member Smith Set might not (arguably) be someone that a majority "voted for". Which would seem to apply just as much to IRV as well, but IRV adherents that make this point seem to think that the mechanics of IRV somehow overcome that.
fdsjgfklsfd 4y ago
Alternatively: "It's the same as Condorcet, we just add multiple rounds for no reason." Why combine two systems to make a more complicated system when you could just use Condorcet directly?

fwip 4y ago

I think the fact that voters knew they were participating in an instant-runoff-vote would make it difficult to draw strong conclusions about what the results would be under other voting systems.

tunesmith 4y ago
The contention that being aware of IRV-vs-Condorcet materially matters in terms of strategic voting is also a pretty unprovable contention. People should be instructed to vote their honest preferences in IRV anyway.

jsmith45 4y ago

I see little point in forcing a separate run-off of Smith Set candidates if there are multiple (i.e. there is no Condorcet winner). After all, we just had everybody indicate their preferences in a ranked list. A runoff could only give new useful information if: 1. one or more voters change their minds. 2. one or more voter gave misleading preference information. 3. The voter's true preferences are not representable as a ranked list (e.g. they Prefer Mr Rock to Ms Scissors, prefer Ms Scissors to Mx Paper, and prefer Mx Paper to Mr Rock.) So i think the tiebreaking methods are reasonable. It is probably also worth noting that for some Condorcet methods, attempting to vote dishonestly to exploit some flaw in the system is has great risk of backfiring. The impact that the dishonest vote will have when a Condorcet winner does exist can differ from the impact it will have if a tiebreaking method come into play, such that it is only safe to try to strategically vote like that if you already know there will be no Condorcet winner. (Which would be very hard to know. Polling can't even reliably predict who will win with single vote elections. Existing polling methodology would be way too under-powered to detect if there will be a Condorcet winner or not).

jefftk 4y ago
In addition to voters changing their minds (which is possible based on additional time, media coverage, research, debates, etc) there are other ways a runoff could change the results: * People who previously didn't vote learn about the situation and participate. * People who previously didn't indicate a preference between two candidates think more and vote accordingly.

godelski 4y ago

I personally don't care much about the strong condorcet criteria as long as the weak one is met. Condorcet methods (RP and Schulze are most common) typically get higher voter satisfaction efficiency under ideal conditions (voters being completely honest) but not under strategic conditions. They also aren't winning my much. This[0] gets shared a lot but often people don't pay attention to the spread in VSE over different strategies, which is just as important as max VSE. Though, of course, voting requires much more than just getting a good VSE. If you're a VSE maximizer, than yeah, Condorcet methods make the most sense, but it is naive to approach it as the only criteria. [0] https://electionscience.github.io/vse-sim/ https://electionscience.github.io/vse-sim/

tunesmith 4y ago
Let's say that 49% "strongly" support A, and 51% "weakly" support B. If VSE means A has a higher score, I think it's meaningless in terms of democratic elections that follow the one-person-one-vote principle, where B should definitely win. There's a whole field of "voter satisfaction" and "utility" that doesn't really apply to democratic elections where we have to worry about populations bullying other populations.
- godelski 4y ago
  > Let's say that 49% "strongly" support A, and 51% "weakly" support B. This example does not have enough information to have a optimal VSE solution or even a good guestimate. We don't know the distance between A and B nor percentages of A that are okay with B and vise versa. So essentially we don't know what our latent landscape even looks like. But if you can add to that and figure it out then we can actually capture information about population bullying. The issue with voting is that it appears simple but is rather complex. But complex problems also require nuanced solutions (not necessarily complex, but can't be solved by first order terms).
  - tunesmith 4y ago
    Distance between A and B are subjective according to each voter. If you are (hypothetically) a binary thinker and think A and B are far apart and absolutely prefer A and hate B. And if I am (hypothetically) a nuanced thinker and can see plusses and minuses to both, and after careful consideration I choose B. Then should your vote matter more than mine?