4 ms·

I've been a proponent of STAR for some time, but today I'm happy to learn about 321. It sounds even easier to explain and implement than STAR (though approval i

by kortex 5y ago

I've been a proponent of STAR for some time, but today I'm happy to learn about 321. It sounds even easier to explain and implement than STAR (though approval is likely the method with the least friction using existing equipment). I really don't see the appeal of IRV, other than momentum. It's harder to implement. It's not really intuitive why you have to rank them. (I belive) it has higher chance of screwing up ballots (e.g. Forgot to rank a candidate). And I don't like how it forces you to split hairs on two equally liked candidates. Minimizing mental friction is important when we struggle to get high voter turnout as-is. I think 321 most closely matches the political climate in America. People tend to like a candidate, hate them, or "eh, they're fine." Also, not totally sure (maybe someone can chime in), but I think 321 is "stateless" or "commutative" in that the counts of Goods, Oks, and Bads are independent. With any runoff methods, each ballot has a "state", because after a knockout, you need to apply the runoff votes based on other information on that ballot. This makes it hard to compose ballots, e.g. for our goofy ass electoral college system. With 321, (I think) you can decompose each "channel" total for each candidate, and aggregate them at a higher level.

crazygringo 5y ago
> I think 321 most closely matches the political climate in America. People tend to like a candidate, hate them, or "eh, they're fine." I couldn't disagree more. I voted in yesterday's NYC Democratic mayoral primary (13 candidates on the ballot), and there were three candidates I considered "like", but with a very strong preference for the first over the second, as well as the second over the third. In today's political climate, people tend to be very passionate about their #1 choice, even if there are 3 they "like". That passion matters.
- godelski 5y ago
  This is an argument for cardinal systems though... (more score/star rather than approval or 321). When ranking you encode in such a way that your preference from one candidate over another is uniform. But in scoring you don't have this limitation. You can express that you REALLY are passionate about 2 candidates but have a slight preference of one over the other, are lukewarm about a few, and really dislike others. Ordinal systems do not allow you to express this. You're right, passion matters.
  - crazygringo 5y ago
    Score voting does solve a lot of problems, you're right. Unfortunately it also introduces a ton of new ones, because while it's easy to compare ordinal preferences, there's no single good intuitive way to compare cardinal preferences. If you give voters a scale from 0 to 10, one person's 3 is another person's 5. It's not even remotely clear that one person's vote of 8 should count the same as two other people's vote of 4, or even that the same person's 8 should count double their 4. Is the preferences scale linear? Logarithmic? Exponential? Other? The arbitrariness and therefore unknown meaning of cardinal voting scores, and cardinal preferences generally, is the reason why philosophers mainly abandoned cardinal measures of utility a long time ago, and most utilitarian philosophy is based on knowledge of ordinal preferences only.
    - godelski 5y ago
      > there's no single good intuitive way to compare cardinal preferences. What? You can literally rank them if you're confused. Better yet, rank them with unequal spacing (A:5, B:4, C:2). It is hard to take your claim seriously because it has been tested in real world settings. This isn't a theoretical proposition (and you're arguing over theory). Also did you really just say "logarithmic vs exponential?" The reality doesn't match your hypothesis and that is testable. You're claiming philosophers abandoned it and I don't see evidence for this. Also I'm not sure why philosophers matter here, this isn't philosophy. Cardinal methods are the preference of most researchers that study voting. I don't know about you, but I tend to go with expert opinion (not absolutely, but it is a strong signal that something is the right path to follow. Then you just vet the work).
      - crazygringo 5y ago
        > What? You can literally rank them if you're confused. That's ordinal, not cardinal then. Your example "(A:5, B:4, C:2)" gives zero indication as to what 5 vs 4 means, or 4 vs 2, beyond simply being "greater than". That's the entire point. > It is hard to take your claim seriously because it has been tested in real world settings. I have no idea what that even means. Any voting system can be "tested". It doesn't mean it produces desirable results. The philosophies behind different voting methods aren't "testable". Political philosophy is philosophy because it isn't testable -- it's about what's desirable from some ethical viewpoint. E.g. we don't "test" majoritarian voting, we defend it from principles of equality. > Also did you really just say "logarithmic vs exponential?" Yes I did. y=2^x and y=log(x) have different shapes qualitatively. In other words, is voting 10 twice as strong as voting for 9, or only a teensy bit stronger? Not sure what's confusing you here, or what words you think would express that difference more clearly. > You're claiming philosophers abandoned it and I don't see evidence for this. It's just the basic history of utilitarianism or consequentialism. Any survey on that can point you in the right direction as to why cardinal preference have been largely abandoned. (I'm obviously not talking about economics where "utility" is commonly measured in dollars. Most analyses of voting methods are concerned with what should be the "right" outcome, not the dollar-maximizing outcome. Social choice theory has been the attempt by economists to essentially measure votes in dollars, but even that still works almost entirely with ordinal preferences that are then analyzed in dollar outcomes.) > Also I'm not sure why philosophers matter here, this isn't philosophy. Voting methods are a huge part of political philosophy. Of course it is. It basically straddles political philosophy and political science in the literature. (Then there are some additional contributions by economists, especially when it gets deeper into the mathematics of it, mainly in social choice theory.) > Cardinal methods are the preference of most researchers that study voting. That's absolutely the opposite of my experience and knowledge. There are absolutely some people who study them, but they're absolutely not the focus of "most" researchers, at least not in the realm of political philosophy and political science. Voting with cardinal preferences is also barely used in real-world political governance. Again, the fundamental problem with cardinal voting is simply that there's no intuitive, easily agreed-upon meaning for what cardinal values even mean. Most succinctly: on a 10-point scale, define the difference between a "4" and a "5", and argue why your definition is superior to 10 other definitions that are trivial to come up with. Also explain how you'll get voters to reliably understand and vote according to that same definition as well. On the other hand, ordinal preferences are entirely intuitive and unambiguous to express and interpret. Which isn't to say they're perfect, but they give researchers something they can actually work with, both theoretically and practically.
- kortex 5y ago
  Wow, I envy you. We don't get a lot of choice in Upstate NY. You're lucky if you even get 2 candidates for a lot of the smaller races. I'm probably over-generalizing but I still think this disparity is a result of low/med density vs high density like NYC. You're always gonna have more choices in high density areas, and I'm definitely in favor of more granular signals whenever possible. I'd love to see an IRV where you don't have to obey any strict ordering rules (which I guess is like score voting where the max score is the number of candidates) which I guess might be a thing?
godelski 5y ago
Also it should be noted that IRV actually increases election spoilage, which is ironically one of its main selling points.

by kortex 5y ago

crazygringo 5y ago

> I think 321 most closely matches the political climate in America. People tend to like a candidate, hate them, or "eh, they're fine." I couldn't disagree more. I voted in yesterday's NYC Democratic mayoral primary (13 candidates on the ballot), and there were three candidates I considered "like", but with a very strong preference for the first over the second, as well as the second over the third. In today's political climate, people tend to be very passionate about their #1 choice, even if there are 3 they "like". That passion matters.

godelski 5y ago
This is an argument for cardinal systems though... (more score/star rather than approval or 321). When ranking you encode in such a way that your preference from one candidate over another is uniform. But in scoring you don't have this limitation. You can express that you REALLY are passionate about 2 candidates but have a slight preference of one over the other, are lukewarm about a few, and really dislike others. Ordinal systems do not allow you to express this. You're right, passion matters.
- crazygringo 5y ago
  Score voting does solve a lot of problems, you're right. Unfortunately it also introduces a ton of new ones, because while it's easy to compare ordinal preferences, there's no single good intuitive way to compare cardinal preferences. If you give voters a scale from 0 to 10, one person's 3 is another person's 5. It's not even remotely clear that one person's vote of 8 should count the same as two other people's vote of 4, or even that the same person's 8 should count double their 4. Is the preferences scale linear? Logarithmic? Exponential? Other? The arbitrariness and therefore unknown meaning of cardinal voting scores, and cardinal preferences generally, is the reason why philosophers mainly abandoned cardinal measures of utility a long time ago, and most utilitarian philosophy is based on knowledge of ordinal preferences only.
  - godelski 5y ago
    > there's no single good intuitive way to compare cardinal preferences. What? You can literally rank them if you're confused. Better yet, rank them with unequal spacing (A:5, B:4, C:2). It is hard to take your claim seriously because it has been tested in real world settings. This isn't a theoretical proposition (and you're arguing over theory). Also did you really just say "logarithmic vs exponential?" The reality doesn't match your hypothesis and that is testable. You're claiming philosophers abandoned it and I don't see evidence for this. Also I'm not sure why philosophers matter here, this isn't philosophy. Cardinal methods are the preference of most researchers that study voting. I don't know about you, but I tend to go with expert opinion (not absolutely, but it is a strong signal that something is the right path to follow. Then you just vet the work).
    - crazygringo 5y ago
      > What? You can literally rank them if you're confused. That's ordinal, not cardinal then. Your example "(A:5, B:4, C:2)" gives zero indication as to what 5 vs 4 means, or 4 vs 2, beyond simply being "greater than". That's the entire point. > It is hard to take your claim seriously because it has been tested in real world settings. I have no idea what that even means. Any voting system can be "tested". It doesn't mean it produces desirable results. The philosophies behind different voting methods aren't "testable". Political philosophy is philosophy because it isn't testable -- it's about what's desirable from some ethical viewpoint. E.g. we don't "test" majoritarian voting, we defend it from principles of equality. > Also did you really just say "logarithmic vs exponential?" Yes I did. y=2^x and y=log(x) have different shapes qualitatively. In other words, is voting 10 twice as strong as voting for 9, or only a teensy bit stronger? Not sure what's confusing you here, or what words you think would express that difference more clearly. > You're claiming philosophers abandoned it and I don't see evidence for this. It's just the basic history of utilitarianism or consequentialism. Any survey on that can point you in the right direction as to why cardinal preference have been largely abandoned. (I'm obviously not talking about economics where "utility" is commonly measured in dollars. Most analyses of voting methods are concerned with what should be the "right" outcome, not the dollar-maximizing outcome. Social choice theory has been the attempt by economists to essentially measure votes in dollars, but even that still works almost entirely with ordinal preferences that are then analyzed in dollar outcomes.) > Also I'm not sure why philosophers matter here, this isn't philosophy. Voting methods are a huge part of political philosophy. Of course it is. It basically straddles political philosophy and political science in the literature. (Then there are some additional contributions by economists, especially when it gets deeper into the mathematics of it, mainly in social choice theory.) > Cardinal methods are the preference of most researchers that study voting. That's absolutely the opposite of my experience and knowledge. There are absolutely some people who study them, but they're absolutely not the focus of "most" researchers, at least not in the realm of political philosophy and political science. Voting with cardinal preferences is also barely used in real-world political governance. Again, the fundamental problem with cardinal voting is simply that there's no intuitive, easily agreed-upon meaning for what cardinal values even mean. Most succinctly: on a 10-point scale, define the difference between a "4" and a "5", and argue why your definition is superior to 10 other definitions that are trivial to come up with. Also explain how you'll get voters to reliably understand and vote according to that same definition as well. On the other hand, ordinal preferences are entirely intuitive and unambiguous to express and interpret. Which isn't to say they're perfect, but they give researchers something they can actually work with, both theoretically and practically.
kortex 5y ago
Wow, I envy you. We don't get a lot of choice in Upstate NY. You're lucky if you even get 2 candidates for a lot of the smaller races. I'm probably over-generalizing but I still think this disparity is a result of low/med density vs high density like NYC. You're always gonna have more choices in high density areas, and I'm definitely in favor of more granular signals whenever possible. I'd love to see an IRV where you don't have to obey any strict ordering rules (which I guess is like score voting where the max score is the number of candidates) which I guess might be a thing?

godelski 5y ago

Also it should be noted that IRV actually increases election spoilage, which is ironically one of its main selling points.