3 ms·

Ranked Choice does suffer from some mathematical irregularities, see http://zesty.ca/voting/sim/ http://zesty.ca/voting/sim/ (2006) Ranked Choice has had some

Ranked Choice does suffer from some mathematical irregularities, see http://zesty.ca/voting/sim/ http://zesty.ca/voting/sim/ (2006) Ranked Choice has had some notable unexpected outcomes, see https://bolson.org/~bolson/2009/20090303_burlington_vt_mayor.html https://bolson.org/~bolson/2009/20090303_burlington_vt_mayor... See also https://www.starvoting.us/ https://www.starvoting.us/ and https://electionscience.org/library/approval-voting/ https://electionscience.org/library/approval-voting/

mindslight 4y ago
Approval voting still suffers from the tyranny of the mainstream parties. There will be conservatives who only vote for Democrats or Republicans, making it so that people who want real change by voting Green/Libertarian would have to not approve-vote their fallback D/R to have any chance of G/L winning. The main utility would be the public referendum on how much support each party has. Score voting collapses to approval voting under strategy. The right answer is "Ranked choice voting", but not with runoff but rather Condorcet. Condorcet specifies that the winner is the candidate who beats every other candidate in a pairwise matching. The single drawback of Condorcet - the possibility of having no Condorcet winner - is overblown. It's the equivalent of a tie in a plurality election. If the current standard tiebreaker mechanism is not good enough (eg another election), then the tie could be broken by taking the same ranked choices and translating them into a heuristic based on approval voting, score voting, etc.
- obelos 4y ago
  Bullet voting is an overblown “weakness” of Approval and Score. “I only voted for my honest favorite” is not a huge drawback, IMO. Nevertheless, in the actual municipal elections using Approval Voting in St. Louis and Fargo, voters on average approved of more than one candidate. And in Score there is incentive to use middle ratings because not doing so increases the probability that a popular candidate you're lukewarm on gets overtaken by a candidate you deeply dislike. The value of STAR voting is that it reduces even further the incentive to bullet vote. By introducing a Condorcet-esque runoff of the top two finalists, every voter is strongly incentivized to use meaningful middle scores in order to have a say in the outcome of the runoff. IRV/RCV, on the other hand, can punish voters for voting honestly. The more viable candidates there are in the field, the more likely your top ranking for your honest favorite will result in benefitting your least preferred. Given that one of the qualities most every election reform advocate wants is to see a larger field packed with viable candidates, this drawback of IRV is especially detrimental.
  - mindslight 4y ago
    > “I only voted for my honest favorite” is not a huge drawback It's a drawback because the input model prevents voters from expressing their actual preference. If individual voters' preferences cannot be meaningfully expressed, then the outcome isn't valid regardless of how reasonable the tallying looks. (see also: Plurality) For example, let's say that I want L/G to win, but will begrudgingly accept a D win to prevent an R win at all costs. With Approval or Score voting, I am unable to simply express this preference! With Approval voting, my two options are to leave out D and create a higher chance of R winning, or include D and have a higher chance of L/G losing. And Score voting just allows me to create a weighted blend of the two - how much do I want to hurt L/G to help D beat R? I don't see how STAR doesn't similarly fall to the two party tyranny. The top two scores will be the two popular parties, which votes then choose between. Putting "Condorcet" after another system seems pointless. > IRV ... The more viable candidates there are in the field, the more likely your top ranking for your honest favorite will result in benefitting your least preferred I wholeheartedly agree about IRV (which describes a tallying method). I'm talking about reframing "RCV" (which describes the input method, rather than the tallying method) to be tallied with Condorcet rather than greedy-algorithm runoffs. We can then debate which heuristics or procedures are appropriate for breaking any Condorcet ties. The only voter preference RCV+Condorcet prevents is the ability to express an individual circular preference (A>B>C>A). But I've seen no arguments about why that would even be a reasonable preference to want to express.
    - obelos 4y ago
      Instant Pairwise Runoff and similar Condorcet methods seem like they would produce good results, though I'm averse to unsummable methods for practical reasons. But I think there's more nuance than saying a ranking method better captures voter preference than Approval because there's numbers. An ordinal method loses information about the distance between ranks. A voter could strongly prefer two candidates and then despise the rest of the pack pretty equally, and much of that distinction is lost in ranking. And the overall end result, even if it doesn't end up electing a despised candidate, is making a despised candidate look like they're the third most popular. That information helps shape who's involved in future elections.

mindslight 4y ago

Approval voting still suffers from the tyranny of the mainstream parties. There will be conservatives who only vote for Democrats or Republicans, making it so that people who want real change by voting Green/Libertarian would have to not approve-vote their fallback D/R to have any chance of G/L winning. The main utility would be the public referendum on how much support each party has. Score voting collapses to approval voting under strategy. The right answer is "Ranked choice voting", but not with runoff but rather Condorcet. Condorcet specifies that the winner is the candidate who beats every other candidate in a pairwise matching. The single drawback of Condorcet - the possibility of having no Condorcet winner - is overblown. It's the equivalent of a tie in a plurality election. If the current standard tiebreaker mechanism is not good enough (eg another election), then the tie could be broken by taking the same ranked choices and translating them into a heuristic based on approval voting, score voting, etc.

obelos 4y ago
Bullet voting is an overblown “weakness” of Approval and Score. “I only voted for my honest favorite” is not a huge drawback, IMO. Nevertheless, in the actual municipal elections using Approval Voting in St. Louis and Fargo, voters on average approved of more than one candidate. And in Score there is incentive to use middle ratings because not doing so increases the probability that a popular candidate you're lukewarm on gets overtaken by a candidate you deeply dislike. The value of STAR voting is that it reduces even further the incentive to bullet vote. By introducing a Condorcet-esque runoff of the top two finalists, every voter is strongly incentivized to use meaningful middle scores in order to have a say in the outcome of the runoff. IRV/RCV, on the other hand, can punish voters for voting honestly. The more viable candidates there are in the field, the more likely your top ranking for your honest favorite will result in benefitting your least preferred. Given that one of the qualities most every election reform advocate wants is to see a larger field packed with viable candidates, this drawback of IRV is especially detrimental.
- mindslight 4y ago
  > “I only voted for my honest favorite” is not a huge drawback It's a drawback because the input model prevents voters from expressing their actual preference. If individual voters' preferences cannot be meaningfully expressed, then the outcome isn't valid regardless of how reasonable the tallying looks. (see also: Plurality) For example, let's say that I want L/G to win, but will begrudgingly accept a D win to prevent an R win at all costs. With Approval or Score voting, I am unable to simply express this preference! With Approval voting, my two options are to leave out D and create a higher chance of R winning, or include D and have a higher chance of L/G losing. And Score voting just allows me to create a weighted blend of the two - how much do I want to hurt L/G to help D beat R? I don't see how STAR doesn't similarly fall to the two party tyranny. The top two scores will be the two popular parties, which votes then choose between. Putting "Condorcet" after another system seems pointless. > IRV ... The more viable candidates there are in the field, the more likely your top ranking for your honest favorite will result in benefitting your least preferred I wholeheartedly agree about IRV (which describes a tallying method). I'm talking about reframing "RCV" (which describes the input method, rather than the tallying method) to be tallied with Condorcet rather than greedy-algorithm runoffs. We can then debate which heuristics or procedures are appropriate for breaking any Condorcet ties. The only voter preference RCV+Condorcet prevents is the ability to express an individual circular preference (A>B>C>A). But I've seen no arguments about why that would even be a reasonable preference to want to express.
  - obelos 4y ago
    Instant Pairwise Runoff and similar Condorcet methods seem like they would produce good results, though I'm averse to unsummable methods for practical reasons. But I think there's more nuance than saying a ranking method better captures voter preference than Approval because there's numbers. An ordinal method loses information about the distance between ranks. A voter could strongly prefer two candidates and then despise the rest of the pack pretty equally, and much of that distinction is lost in ranking. And the overall end result, even if it doesn't end up electing a despised candidate, is making a despised candidate look like they're the third most popular. That information helps shape who's involved in future elections.