3 ms·

Ranked choice voting systems (e.g. single transferable vote) have a lot of "game theory" advantages over e.g. first-past-the-post, but are hard to implement wit

Ranked choice voting systems (e.g. single transferable vote) have a lot of "game theory" advantages over e.g. first-past-the-post, but are hard to implement with manual counting. Edit: I suppose that doesn't mean voting machines are required though, just that the results on each ballot paper need to be inputted to a computer to determine the outcome. A lot of data inputting required, but still auditable.

eru 4y ago
First-past-the-post is pretty horrible. Alas, experience from Australia shows that single transferable vote isn't any better in practice, either. It's more complicated with no benefit in practice. (See https://rangevoting.org/rangeVirv.html https://rangevoting.org/rangeVirv.html for more details on Australia.) If traditional voting is required, I have higher hopes for approval voting or the slightly more complicated range voting. These are quite easy to evaluate completely by hand. Especially the former. If you are happy to go a bit off the beaten track, using sortition amongst volunteers might be a good way to fill up a parliament. (Or any other body big enough for the law of large numbers to kick in.) As a slightly more complicated system, I would have every voter indicate their favourite candidate on a ballot sheet, and then draw enough votes at random to fill up parliament. (Use your favourite mechanism to handle the same candidate being voted for multiple times in your sample. Perhaps give them more weight, or re-draw, or have people write down an ordered list of preferred candidates, and admit the highest ranked one who isn't already in parliament, etc.) Random sampling is surprisingly powerful. It also completely bypasses Arrow's impossibility theorem.
- BlueTemplar 4y ago
  Does it ?! Wow, that is tempting...
  - eru 4y ago
    Well, Arrow's theorem only applies to deterministic procedures.
  - Vecr 4y ago
    Yes, but there's literally no way to prove that a random number that you "generated" is actually random. The best idea would be to write people's names on balls, put them in a clear funnel thing, and blow them around or something like that on live TV, but for sortition, I'm not sure how that would scale. In the movie Contagion they did that for each day of the year someone could be born on, and that's how they distributed the vaccines. Not sure how that would scale to millions of balls and how you would audit that each person's name is on one ball and only one ball.
    - eru 4y ago
      Yes, building a purely mechanical trustworthy source of randomness is hard. However, you can look at how the lottery does it, perhaps? For example, they don't write the players names on the balls. They use more indirection. (And lots of other cleverness.) Also keep in mind that I suggest to use this mechanism for something like filling up a parliament. The German Bundestag has about 600 members. The UK has 650 MPs in their House of Commons. I posit that getting into parliament is perhaps comparable to winning the lottery, ie something people might trust lottery equipment and procedures to handle. Becoming the president of the US is a bigger deal than winning the lottery, so we can't naively expect lottery equipment and procedures to be above suspicion.

eru 4y ago

First-past-the-post is pretty horrible. Alas, experience from Australia shows that single transferable vote isn't any better in practice, either. It's more complicated with no benefit in practice. (See https://rangevoting.org/rangeVirv.html https://rangevoting.org/rangeVirv.html for more details on Australia.) If traditional voting is required, I have higher hopes for approval voting or the slightly more complicated range voting. These are quite easy to evaluate completely by hand. Especially the former. If you are happy to go a bit off the beaten track, using sortition amongst volunteers might be a good way to fill up a parliament. (Or any other body big enough for the law of large numbers to kick in.) As a slightly more complicated system, I would have every voter indicate their favourite candidate on a ballot sheet, and then draw enough votes at random to fill up parliament. (Use your favourite mechanism to handle the same candidate being voted for multiple times in your sample. Perhaps give them more weight, or re-draw, or have people write down an ordered list of preferred candidates, and admit the highest ranked one who isn't already in parliament, etc.) Random sampling is surprisingly powerful. It also completely bypasses Arrow's impossibility theorem.

BlueTemplar 4y ago
Does it ?! Wow, that is tempting...
- eru 4y ago
  Well, Arrow's theorem only applies to deterministic procedures.
- Vecr 4y ago
  Yes, but there's literally no way to prove that a random number that you "generated" is actually random. The best idea would be to write people's names on balls, put them in a clear funnel thing, and blow them around or something like that on live TV, but for sortition, I'm not sure how that would scale. In the movie Contagion they did that for each day of the year someone could be born on, and that's how they distributed the vaccines. Not sure how that would scale to millions of balls and how you would audit that each person's name is on one ball and only one ball.
  - eru 4y ago
    Yes, building a purely mechanical trustworthy source of randomness is hard. However, you can look at how the lottery does it, perhaps? For example, they don't write the players names on the balls. They use more indirection. (And lots of other cleverness.) Also keep in mind that I suggest to use this mechanism for something like filling up a parliament. The German Bundestag has about 600 members. The UK has 650 MPs in their House of Commons. I posit that getting into parliament is perhaps comparable to winning the lottery, ie something people might trust lottery equipment and procedures to handle. Becoming the president of the US is a bigger deal than winning the lottery, so we can't naively expect lottery equipment and procedures to be above suspicion.