San Francisco election legal code challenge
-
@rob said in San Francisco election legal code challenge:
changes minimal
How are you going to measure "minimal"? Two points for every word added, one point for every word removed?
Approval-ordered Llull (letter grades) [10], Score // Llull [9], Score, STAR, Approval, other rated Condorcet [8]; equal-ranked Condorcet [4]; strictly-ranked Condorcet [3]; everything else [0].
-
@jack-waugh Well I'm not giving away a prize to the winner, so does it really matter?
-
Anyway, they say "In a Ranked-Choice Voting system, if no candidate receives over 50% of the vote, the least popular candidate – the one who was ranked 1st by the fewest people - is eliminated from the election." So, I think that preserves the worst aspect of IRV, no matter what else is changed.
Approval-ordered Llull (letter grades) [10], Score // Llull [9], Score, STAR, Approval, other rated Condorcet [8]; equal-ranked Condorcet [4]; strictly-ranked Condorcet [3]; everything else [0].
-
@jack-waugh maybe there is a way to slightly reword it in a sneaky way that allows Minimax to be used?
-
@jack-waugh If you really want me to come up with a way of measuring, Levenshtein distance works for me.
For example, the Levenshtein distance between "kitten" and "sitting" is 3, since the following 3 edits change one into the other, and there is no way to do it with fewer than 3 edits:
kitten → sitten (substitution of "s" for "k"),
sitten → sittin (substitution of "i" for "e"),
sittin → sitting (insertion of "g" at the end).However you measure it, I think minimal changes would make it far more attractive. Or maybe I should say, it should better avoid resistance. That's the idea of this post anyway, being able to say "here's a tiny change you can slide in to make it significantly better."
I have no doubt that the wording of that piece of legislation (which is from 2002, I think, and one of the earliest IRV implementations for US public elections) was hashed over a lot, to keep it both simple and unambiguous.
Obviously the changes should be clearly understandable (for instance you can't just use the term "Condorcet winner" or "Smith set" or the like... you have to actually describe what it means without sending people off to look things up elsewhere) Also obviously, that can't be objectively measured, so that's why I wasn't actually proposing this as something where there is an objective "winning answer."
So, I think that preserves the worst aspect of IRV, no matter what else is changed.
But, with the sort of changes I am suggesting (specifically to change it to be Bentham's method), that is only in the (presumably) rare case of there being no Condorcet winner.
Regardless it isn't a requirement to keep the "falls back to IRV" approach, but I think if you eliminate that, you make it a lot tougher sell because you probably are going to have to make it a bigger change. In this respect, I tend to be in the "perfect is the enemy of good" camp.
To me, the "worst aspect of IRV" is that it isn't Condorcet compliant, which would have avoided all these complaints about the IRV election in Burlington that supposedly elected "the wrong candidate."
Notice that even FairVote says this:
We also consider the Condorcet criterion to be important. This is the property that the candidate that would win a head-to-head race against every other candidate should always win. While RCV, approval, and score voting may fail to elect the Condorcet candidate, in practice RCV has done so in virtually every single election.
https://www.fairvote.org/how_is_rcv_better_than_approval_score_or_condorcet_voting_methods
-
maybe there is a way to slightly reword it in a sneaky way that allows Minimax to be used?
@brozai Is that possible? I'd love to see you try your hand at doing it.
I would have thought it would be a fairly substantial change. -
@rob I take it you are not trying to stay within the Forward Party's definition of RCV, then.
Approval-ordered Llull (letter grades) [10], Score // Llull [9], Score, STAR, Approval, other rated Condorcet [8]; equal-ranked Condorcet [4]; strictly-ranked Condorcet [3]; everything else [0].
-
I take it you are not trying to stay within the Forward Party's definition of RCV, then.
In a sense I am, at least according to this (emphasis mine) : "This is why Ranked-Choice Voting (or, at least, one variety of it) is sometimes referred to as Instant Runoff Voting." From this page: https://www.forwardparty.com/rankedchoice-voting
So, with that parenthetical, they acknowledge that the definition of Ranked Choice Voting isn't restricted to being IRV. While they do describe the IRV variety of Ranked Choice in some detail (suggesting that IRV is the variety they advocate), they never say it is better than any other variety.
I have heard Andrew Yang mention that he isn't really hard-core about it being IRV or even a ranked method --- just one that isn't subject to all the negatives of plurality. Strategically, though, they know that the IRV form of Ranked Choice has the most mindshare as well as being in use and more tested than any other alternative, and they don't want to complicate their messaging too much.
My suggestion (which isn't really just about the Forward Party) is that, given that Ranked Choice has mindshare -- as well as current legislation in place and so on -- we put something together that doesn't distract from that much, if any.
On a bit of a tangent, note that the San Francisco ranked choice legislation is flexible in one respect :
if the voting system, vote tabulation system or similar or related equipment used by the City and County cannot feasibly accommodate choices equal to the total number of candidates running for each office, then the Director of Elections may limit the number of choices a voter may rank to no fewer than three.
And that is how it was until a couple years ago... you could only rank your top three. Many people see things like those things as minor details. Ranking only 3 is actually pretty bad, but better than plurality.... so you could consider the initial implementation somewhat of a "baby step" toward something even better. Now that they have upgraded the system, we can rank the candidates. The change from 3 to nearly unlimited was not even something people worried about or talked about, it was just introduced with zero fanfare or controversy.
The point is, altering it to be Condorcet compliant might be seen by a large number of people similarly. Like "ok, yeah in a perfect world we'd like Condorcet compliance, but for now we appreciate that at least we get to rank them." To people like us (well, me anyway), it is important. But if those resistant to change see the distinction between RCV-IRV and RCV-Bentham as a minor detail, that is a plus for getting it adopted.
For the Forward Party, allowing for a Condorcet compliant version of Ranked Choice would certainly involve a bit of a change of wording on their web site, but not a lot. Honestly, it could be simply a footnote or short parenthetical expression.
-
@rob So Benham's method is actually Condorcet compliant. It only differs from other Condorcet methods in case of a cycle, right? And a cycle is fairly unusual, right?
Approval-ordered Llull (letter grades) [10], Score // Llull [9], Score, STAR, Approval, other rated Condorcet [8]; equal-ranked Condorcet [4]; strictly-ranked Condorcet [3]; everything else [0].
-
Correct. It just falls back to regular IRV in the case of a cycle. And yes, cycles are expected to be very rare.
It isn't what I'd consider the best Condorcet method, but I think that is probably outweighed by the practical / strategic value of making it seem similar to a method that is already being used and otherwise has some momentum. I think it would be very easy to sell as simply being a variation of Ranked Choice.
I think if any municipality were to adopt it, it would then make it a lot easier to propose to the next municipality moving to RCV some sort of better Condorcet method. But really, I don't care, Bentham is fine as far as I'm concerned. I believe, in real world elections, it would be as immune to strategic voting and strategic nomination as any method.
