Proposed options for "voting on voting methods"
-
@rob I would put them in the same class as STLR and STAR.... I think. They are all directed towards trying to reweight the scores as you eliminate losers.
STAR and STLR do it for the final round while several others do it at each round. The cost of doing it a each round is that you lose monotonicity and there is a fair bit of added complexity. In my mind they all have the same intent so it would be fair to put them together. "reweighted cardinal run-off methods" or something like that would be a reasonable name.
All that said, It would be a shame to lump all such systems of varying quality together with a frontrunner like STAR. Could I propose that commonly used systems like Approval, STAR and IRV stand alone while other systems with no lobby behind them get grouped?
-
@keith-edmonds Totally agree on keeping STAR, Approval and IRV separate (only lumping together minor variations of them).
-
Great idea! Some thoughts:
I would probably mention that IRV is also known as Alternative Vote. In the UK, we had a referendum on using it for our parliamentary elections, and that was the name it went under.
For ranked-choice Condorcet, for the examples I'd probably include some of the better known methods like ranked pairs or Schulze, so it's clearer at a glance what the category is. Otherwise it might confuse people into thinking it's some esoteric group rather than what it actually is.
For median methods, I would exclude the word "cardinal". As you say, Majority Judgement differs from your description anyway because it uses letter grades rather than cardinal scores but that it falls into this category anyway. I would just make the category wider so that it is included without having to add a caveat.
At the risk of incurring the wrath of @Keith-Edmonds I wouldn't have STLR. As far as I know (and it's possible I'm wrong), it is just his pet method and has gained no wider traction. And if you include that method, then there are dozens of others you could add too. It is also based on what I would consider to be the misguided notion of utility ratios. In any case, it does seem to stick out as the most esoteric method in the list.
-
@toby-pereira On the contrary. I agree with all your points and you will get no wrath. I have purposely not attempted to gain traction for STLR since STAR is the better choice for lobbying.
I think it would be best to have a class containing IRNR, Cardinal Baldwin, STLR and other reweighting methods. Brian Olson, Aldo Tragni, myself and some others have put a lot of time into the development of this class. STAR is technically in this class but it is on the edge of it. For that and reasons I mentioned before STAR should stand alone.
While we are on the topic, if you consider the reweighting method of STLR to be misguided then by extension you would also consider cardinal Baldwin to be misguided. Is there a reweighting system which you think if better?
-
@keith-edmonds said in Proposed options for "voting on voting methods":
While we are on the topic, if you consider the reweighting method of STLR to be misguided then by extension you would also consider Cardinal Baldwin to be misguided. Is there a reweighting system which you think if better?
If I've understood the methods correctly, then I think Cardinal Baldwin normalises in a different way from STLR and preserves relative score distances (affine transformation I think it's called) rather than ratios.
E.g. if a voter score candidate A 5 and B 4 (out of max 5) and these are the last two remaining candidates, then STLR will leave the scores alone, whereas Cardinal Baldwin will turn them into 5 and 0 respectively.
Whereas if A scores 1 and B 0 (still out of 5), then both methods would normalise to 5 and 0.
I would prefer Cardinal Baldwin's handling of this.
-
@toby-pereira OK, makes sense. The one round run off version of that is STAR and the every round run off is Cardinal Baldwin. Presumably you prefer this because it maximizes voter influence. I invented STLR specifically to avoid that as it leads to majoritarianism.
It seems there are (at least) three types. This one. The one STLR uses. And the one for IRNR and Distributed Voting
I propose that these should all be consider them all the same class for this poll. Do you agree? Wanna suggest a name? I gave one above.
-
@keith-edmonds said in Proposed options for "voting on voting methods":
I propose that these should all be consider them all the same class for this poll. Do you agree? Wanna suggest a name? I gave one above.
Well in terms of the poll, I think if they were all to be included, some of them might different enough methods to be kept separate, even if they have the similarity of some sort of score normalisation. I think lumping together should only happen if you'd expect basically the same results, and I'm not sure if that's the case with all these methods.
As for the name, I'm not too bothered especially since I'm not invested in these methods, so I wouldn't disagree with your "reweighted cardinal run-off methods". But as I'm here, Cardinal Automatic Run-off methods could be abbreviated to CAR methods!
-
What about this.... I'm thinking I should use longer names than the abbreviations I previously suggested (so you can tell what they are without looking them up) but still relatively short so it doesn't make textual ballots too bulky.
I was thinking just "cardinal alt" for ones that aren't condorcet compliant and don't fit into other categories, so cardinal baldwin, STLR, and a few others would be members of the class. If we get enough people voting for them, we should break it down further in future votes.
ranked-irv
ranked-condorcet
ranked-borda
cardinal-condorcet
cardinal-median
cardinal-alt
star
approval
score
choose-one -
@rob I have to say I think I agree with @Keith-Edmonds that STAR should probably be in a class of its own, especially because of its much more prominent presence in voting milieu than other `cardinal-alt' methods.
-
@andy-dienes oh my bad I had star in there but somehow lost it while sorting things around. Fixed.
-
@rob said in Proposed options for "voting on voting methods":
What about this.... I'm thinking I should use longer names than the abbreviations I previously suggested (so you can tell what they are without looking them up) but still relatively short so it doesn't make textual ballots too bulky.
I was thinking just "cardinal alt" for ones that aren't condorcet compliant and don't fit into other categories, so cardinal baldwin, STLR, and a few others would be members of the class. If we get enough people voting for them, we should break it down further in future votes.
ranked-irv
ranked-condorcet
ranked-borda
cardinal-condorcet
cardinal-median
cardinal-alt
star
approval
score
choose-oneI think longer abbreviations are good. The problem with cardinal-alt is that it would be difficult to know how to score it if you like some that would fall under the category but not others. I would probably only include options that are specific enough so that someone would likely have a similar opinion of all the methods it would cover.
-
@toby-pereira Point taken. Just keep in mind that this isn't just going to be a one time vote. If enough people vote for the cardinal-alt to justify it, we can break it down into several for the next vote. Or have a separate vote for "best way to tabulate cardinal ballots?" or something,
Notice that other ones can be broken down further as well. For instance ranked condorcet has various cycle breaker mechanisms, ranked irv can or cannot allow for equal rankings, score can be 0-5 or 0-100, and so on.
Anyway, I remain open to suggestions, but I'd like to keep it around 10-12 options, and account for STLR one way or another (since it is the favorite method of at least one of the top contributors here, and doesn't fit into other categories.).
-
There's a category of first-step-eliminations before other main popular methods that I would like as a seperate category or vote in another thread.
As an example I would like to eliminate 35% of the candidates with the lowest (average) scores.
ranked-irv[1] ranked-condorcet[2] ranked-borda[0] cardinal-condorcet[10] cardinal-median[4] cardinal-alt[10] star[6] approval[5] score[4] choose-one[1] minimax-TD[10]
-
@multi_system_fan Is there a system that meets this description that has been described/documented in some detail?
-
@rob no, I read about it somewhere but can't find it....
-
Can we have Reverse STAR as an option?
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 This one? https://www.votingtheory.org/forum/topic/130/star-like-method-reverse-star
That falls under "cardinal-condorcet." (Condorcet compliant methods with score ballots)
It's basically Copeland//score ... since it does pairwise wins first, of those tied for first place, elect the one with the highest score tally. But if there is a Condorcet winner, that will be the winner and the score tally is never looked at.
I think it is a good method, if cardinal ballots are possible and if precinct summability is important. (if precinct summability is not important, and a little complexity is ok, I think it would be better if it normalized the ballots after eliminating the copeland-losers).
Are you ok with it just being a subset of cardinal-condorcet (which also includes Score Sorted Margins)? I'll bet is is really, really hard to find a scenario where the results would differ between various cardinal-condorcet methods.
-
@rob I think this collection of voting systems would be great as the candidate pool for the vote on voting methods, although personally I will have to do some reading about the distinctions between some of the different Condorcet methods. We may also include distributed voting because it is another distinct framework that increases the diversity of the candidate pool, and another system I would want to include is Bucklin voting, which is a non-Condorcet method somewhat similar to ranked-IRV but without vote transfers. I also think including some method that produces a metric combining score/rank and statistical dispersion (variance/IQR) to choose the winner would be good, and possibly for/against because again it increases diversity.
I think a separate vote on preferred Condorcet tie-breaking methods (given that a Condorcet method is being used, of course) would be interesting as well, since this is also a central topic in voting theory literature whether or not a Condorcet method is preferred in the first place.
Less importantly maybe, we could have a vote about tie-breaking methods in general. The book "Economics and Computation" (edited by J. Rothe) suggests that it "may also be appropriate to not implement a specific tie-breaking rule, but to randomly choose one from several 'reasonable' tie-breaking rules" (p. 250).
score-stratified-condorcet [10] cardinal-condorcet [9] ranked-condorcet [8] score [7] approval [6] ranked-bucklin [5] star [4] ranked-irv [3] ranked-borda [2] for-against [1] distribute [0] choose-one [0]
-
@cfrank I'll make sure to include Bucklin.
@cfrank said:
The book "Economics and Computation" (edited by J. Rothe) suggest that it "may also be appropriate to not implement a specific tie-breaking rule, but to randomly choose one from several 'reasonable' tie-breaking rules
Ewwww.
Not a fan of random anything.My current thinking on Condorcet is that, preferable to finding the Condorcet winner directly and then doing something else if there is none, is to have a singular mechanism that determines the winner (always, unless there is a "true tie" which becomes less and less likely with large numbers of voters), that just happens to be Condorcet compliant. BTR-IRV is a good example. My intuition is that it is less likely to provide an incentive to strategically try to create a cycle, if a cycle isn't really a major "thing".
Anyway, that's probably just semantics (as to whether it is a tie breaker) and straying a bit from voting on voting methods.
