On 2018-12-04 08:11, Ted Stern wrote: > Hi Rob, > > What you're describing with your percent satisfaction sounds a lot like > Proportional Representation, though the context of a primary is quite > different. If you're doing what is in essence proportional representation, then the primary problem and the PR problem becomes the same - so use a proportional representation algorithm. (Something like PAV.) That's pretty much what James Green-Armytage concluded in his runoff paper ( http://jamesgreenarmytage.com/runoff.pdf) too. The difference seems to be that in a primary, you want to elect however many winners are required to keep voter satisfaction above a certain level, rather than elect a fixed number of winners. Furthermore, it might be necessary to adjust the outcome towards a proportional ordering (in practice, only considering outcomes that include the Approval winner). Using proportional representation also avoids having a bunch of bland center-seekers, because if the center only satisfies x% of the voters, it will only get x% of the seats, up to roundoff errors and the center-skewing effect of including the Approval winner. In proportional representation methods, a slight center-skew might not be too bad, as it helps kingmaker scenarios favor the center instead of a wing. In a primary, there's no kingmaker scenario that I can see, so the reason for including the Approval winner is mostly to not make the method any worse than if it had just been single-stage Approval. So, the above seems to suggest that the cutoff shouldn't be on support of the last candidate, but on satisfaction (what something like PAV is optimizing). On the other hand, just *explaining* what that score is might be hard... and the more seats you use PAV to allocate, the more opaque the workings of a global optimization method might become in the eyes of the voters.

Hello Rob (et al.), Returning to this after a while out of the loop. Two round methods have been on my mind while walking the dogs. After a lot of thought, I think that if you're going to just use Approval, the simplest method would be to have a runoff between the Approval Winner and the complementary Approval Winner ( "CAW", wins approval on all ballots that do not approve of AW), and leave it at that. And that would occur only if the AW had less than 50 percent of the vote. But that still seems unsatisfactory to me. I'm not happy with allowing the possibility of letting clones into the second round, but it also feels like there might be a compromise candidate who is missed. Maybe take the AW and AW-runner-up and the CAW plus CAW-runner-up. I'd be willing to go with this based on its simplicity. I've been looking again at Jameson Quinn's 3-2-1 voting method ( https://electowiki.org/wiki/3-2-1_voting). It has a number of attractive properties, but a plurality faction could defeat it with clones. It seems to me that one could make 3-2-1 cloneproof and get a 2-round method at the same time: - W321 = 3-2-1 voting winner according to Jameson's page above. - On all ballots that do not approve of W321, find the 3 candidates with highest top ("Good") rating sums (summed over only those ballots not approving W321). - Of those 3 candidates, drop the candidate with lowest top + compromise ("Good" + "OK") ratings (summed over only those ballots not approving W321). - Between those two remaining candidates, CW321 = the candidate most preferred on non-W321 ballots. In a single round election the winner would be the candidate most preferred between W321 and CW321 on all ballots. Up until the CW321 step above, all aspects of the method are summable in a single pass, but finding the CW requires another pass, thought that would be summable. Then the W321 vs. CW321 pairwise check would require another pass. For a two round election, you could have the W321 winner and the runner up, plus the two candidates from the CW321 step. And the 2nd round would only be necessary if the W321 candidate had less than 50% approval, or either of the CW321 pair were preferred to W321. This would be at most 4 candidates, and would require only a single summable pass through the ballots. For the final round, you would still have to do the clone-proofing step, but it would be much easier with at most 4 choices. On Wed, Dec 5, 2018 at 1:41 PM Rob Lanphier <robla@robla.net> wrote: > Hi Ted, > > This has been an incredibly helpful discussion! More inline... > > On Mon, Dec 3, 2018 at 11:11 PM Ted Stern <dodecatheon@gmail.com> wrote: > > What you're describing with your percent satisfaction sounds a lot > > like Proportional Representation, though the context of a primary is > > quite different. > > > > I come from the school of "challenge makes you stronger", so I would > > welcome more contrasting voices into the general election. > > Yeah, I think we may have been educated in the same school. ;-) I > would welcome contrasting voices too, to a point. An approval-based > primary should only advance candidates who are trying to ultimately > get a high approval rating in the general election. The primary > election should select candidates with the desire and ability to earn > the approval of a majority of voters during the general election > cycle, rather than wasting the time and energy of the general election > electorate who may not have the time or patience to evaluate > candidates who aren't serious contenders. > > Ted also wrote: > > I think that including the complementary opposition winners against your > > "truly viable candidates" would be a way to do that. > > > > You're more likely to get engagement and consequent turnout when more > > people feel like their voices are being heard in the debate. A slate of > > blandly similar center seekers would be a recipe for voter apathy. > > My hope is that after a few years, primary election voters become > accustomed to the system. They might then stop approving bland > center-seekers, and start approving exciting center-seekers. Or maybe > bland but incredibly competent center seekers. Voters would hopefully > be at liberty to stop being overly-focused on the left-right > continuum, and start looking at other dimensions (e.g. > competence-incompetence, honesty-corruption, > charisma-unpersuasiveness). If elections became a little less > exciting, that might actually be a good thing. > > Moreover, voters would hopefully become more generous about approving > candidates in the primary that they're not yet ready to vote for in > the general election, as a way of saying "hey, having Katie > McKrazypants in the general election debates ought to make things > interesting! Let's vote for her!". Over time, I suspect that > candidates would get good at positioning themselves as good > safe-to-fail experimental candidates. Then the general election > becomes the safeguard where many of the same voters might then cast a > more critical vote. Given that we'll have over seven months(!) to vet > the general election candidates, we should assume that some voters > will change their mind, and/or Katie McCrazypants would get thoroughly > vetted and found to be not-so-crazy after all. > > A close candidate who gets 45% approval (e.g Cassie McCloseypants) but > doesn't get quite enough approval to advance to the general election > still sends a pretty strong signal. It would seem that at least some > of the advancing general election candidates (e.g. Larissa > McLeftypants and Ronald McRightypants) would come to understand that > the 45% candidate (Cassie McCloseypants) was trying to advance some > issues that are worthy of further discussion during the general > election cycle. Both general election candidates would be well > advised to seek out an endorsement from Cassie McCloseypants, and > would have plenty of time to broker for support from the Cassie > McCloseypants campaign (e.g. recruiting volunteers, getting email > lists, getting contact databases, etc) > > Ted also wrote: > > If the viable candidates are A_1 (= Approval Winner), A_2 (Approval > > runner up), etc., with complementary opponents B_1, B_2, etc., then I > > think it would be appropriate to add them in (A_i, B_i) pairs until your > > desired representation level is met. > > > > Actually, I don't know if I would put the truly viable cutoff at 50%. In > > a true jungle primary, you might end up with only 40% winners at the > > highest. I might go down to 33. 3% A_i candidates if that's what it > > takes to get at least 66.6% voter representation. > > The tradeoff that we're up against: providing genuine choice for a > large part of the electorate vs properly culling the list of primary > contenders so as not to overwhelm voters with options. Though I > suspect your intuition is correct about almost all candidates scoring > well under 50% in the first few elections, my fear is that you're > being too generous for the long-term. The benefit of having a 50% > threshold for _automatic_ advancement to the general election is that > it makes the barrier high enough that it becomes difficult for rabid > extremists to game the system. > > What I fear: as the system becomes routine, sophisticated candidates > will learn to optimize their primary candidacies to achieve whatever > the minimum percentage required to guarantee them a spot on the > general election ballot. 50% is a respectable target. 33% is > horrifyingly low target. With FPTP, we live in a world where each > party optimizes for 50.1% inclusion by estranging the other 49.9% that > aren't part of the tribe. I shudder to think about what tactics would > emerge to optimize political tribes for 33.4% in/66.6% out. > > Mind you, I'm still proposing we develop a small loophole for allowing > "marginally viable candidates" to advance, though only in service of > increasing overall ballot satisfaction. It seems too large of a > loophole to allow *every* complementary candidate to the runners-up to > advance, doesn't it? Granted, it does at least provide disincentive > for the dominant political party to flood the field with clones, but > it also seems like it would be possible for the complementary > candidates to be fringe, extremist candidates with very low approval > scores. Perhaps the answer to that is to just set the floor for > "marginally viable" to be 40%, and then only advance marginally viable > candidates who increase the ballot satisfaction score. We can also > still limit the number of marginally viable candidates who advance > (the "opposition candidate pool") by maintaining the rule that the > opposition candidate pool will be no larger than the majority > candidate pool. > > I suspect that it would take a few years of usage before any > second-place candidate gets above 50% approval. In fact, it may be > common for even the Approval Winner (AW) to get less than 50%. > > With all this in mind, it may be worth reading Clay Shentrup's > responses to my blog post: > <https://medium.com/@ClayShentrup/i-challenge-this-858b5ab0a09a> > <https://medium.com/@ClayShentrup/way-too-much-complexity-b4b2c8d6ee6e> > > In short, he's suggesting to just use the top two Approval winners. > That seems likely to motivate candidates of the dominant party to > unite as clones of one another, but maybe I'm missing something. > Still, the KISS principle of his proposal is appealing. > > I think I'm pretty close to being able to write up draft 3 of MAF, > which should be simple enough. What I have in mind now is basically: > "allow all candidates who (usually) get over 50% approval, and allow a > (usually) equal number of opposition candidates who get over 40% > approval". The first "(usually)" acknowledges the possibility that > there may not be any candidate that gets over 50% approval, and the > second "(usually)" acknowledges that there might not be any candidates > who qualify as opposition candidates, or that the opposition > candidates may be outnumbered. > > Rob >

- W321 = 3-2-1 voting winner according to Jameson's page above.
- On all ballots that do not approve of W321, find the 3 candidates
with highest top ("Good") rating sums (summed over only those ballots not
approving W321).
- Of those 3 candidates, drop the candidate with lowest top +
compromise ("Good" + "OK") ratings (summed over only those ballots not
approving W321).
- Between those two remaining candidates, CW321 = the candidate most
preferred on non-W321 ballots.

Thinking some more about Rob's MAF proposal, I think I have it boiled down to a simple method: First round: Total Approval Winner (AW) plus any other candidate with >= 50% approval, plus the Complementary Approval Winner (CAW) -- the candidate who is approved most on ballots that do not approve AW. Second round Approval Winner is the overall winner. Rationale: If AW and other candidates have >= 50% approval, the sincere Condorcet winner is probably among them. If AW has < 50% support, the sincere Condorcet Winner might not be included as the Approval runner up if the AW is part of a clone set. In addition, having a contrasting candidate to the AW, no matter what the level of approval, means that the sentiment space in which an approval cutoff is measured is broadened. This method is clone-resistant, summable, and resistant to pushover. I think that it may also be resistant to burying. I would like to see a Yee plot of this method. I may eventually get around to programming that but it could be a while. On Mon, Mar 18, 2019 at 2:41 PM Ted Stern <dodecatheon@gmail.com> wrote: > Hello Rob (et al.), > > Returning to this after a while out of the loop. > > Two round methods have been on my mind while walking the dogs. > > After a lot of thought, I think that if you're going to just use Approval, > the simplest method would be to have a runoff between the Approval Winner > and the complementary Approval Winner ( "CAW", wins approval on all ballots > that do not approve of AW), and leave it at that. And that would occur > only if the AW had less than 50 percent of the vote. > > But that still seems unsatisfactory to me. I'm not happy with allowing > the possibility of letting clones into the second round, but it also feels > like there might be a compromise candidate who is missed. Maybe take the > AW and AW-runner-up and the CAW plus CAW-runner-up. I'd be willing to go > with this based on its simplicity. > > I've been looking again at Jameson Quinn's 3-2-1 voting method ( > https://electowiki.org/wiki/3-2-1_voting). It has a number of attractive > properties, but a plurality faction could defeat it with clones. > > It seems to me that one could make 3-2-1 cloneproof and get a 2-round > method at the same time: > > > - W321 = 3-2-1 voting winner according to Jameson's page above. > - On all ballots that do not approve of W321, find the 3 candidates > with highest top ("Good") rating sums (summed over only those ballots not > approving W321). > - Of those 3 candidates, drop the candidate with lowest top + > compromise ("Good" + "OK") ratings (summed over only those ballots not > approving W321). > - Between those two remaining candidates, CW321 = the candidate most > preferred on non-W321 ballots. > > In a single round election the winner would be the candidate most > preferred between W321 and CW321 on all ballots. Up until the CW321 step > above, all aspects of the method are summable in a single pass, but finding > the CW requires another pass, thought that would be summable. Then the > W321 vs. CW321 pairwise check would require another pass. > > For a two round election, you could have the W321 winner and the runner > up, plus the two candidates from the CW321 step. And the 2nd round would > only be necessary if the W321 candidate had less than 50% approval, or > either of the CW321 pair were preferred to W321. This would be at most 4 > candidates, and would require only a single summable pass through the > ballots. > > For the final round, you would still have to do the clone-proofing step, > but it would be much easier with at most 4 choices. > > > > > > > > On Wed, Dec 5, 2018 at 1:41 PM Rob Lanphier <robla@robla.net> wrote: > >> Hi Ted, >> >> This has been an incredibly helpful discussion! More inline... >> >> On Mon, Dec 3, 2018 at 11:11 PM Ted Stern <dodecatheon@gmail.com> wrote: >> > What you're describing with your percent satisfaction sounds a lot >> > like Proportional Representation, though the context of a primary is >> > quite different. >> > >> > I come from the school of "challenge makes you stronger", so I would >> > welcome more contrasting voices into the general election. >> >> Yeah, I think we may have been educated in the same school. ;-) I >> would welcome contrasting voices too, to a point. An approval-based >> primary should only advance candidates who are trying to ultimately >> get a high approval rating in the general election. The primary >> election should select candidates with the desire and ability to earn >> the approval of a majority of voters during the general election >> cycle, rather than wasting the time and energy of the general election >> electorate who may not have the time or patience to evaluate >> candidates who aren't serious contenders. >> >> Ted also wrote: >> > I think that including the complementary opposition winners against your >> > "truly viable candidates" would be a way to do that. >> > >> > You're more likely to get engagement and consequent turnout when more >> > people feel like their voices are being heard in the debate. A slate of >> > blandly similar center seekers would be a recipe for voter apathy. >> >> My hope is that after a few years, primary election voters become >> accustomed to the system. They might then stop approving bland >> center-seekers, and start approving exciting center-seekers. Or maybe >> bland but incredibly competent center seekers. Voters would hopefully >> be at liberty to stop being overly-focused on the left-right >> continuum, and start looking at other dimensions (e.g. >> competence-incompetence, honesty-corruption, >> charisma-unpersuasiveness). If elections became a little less >> exciting, that might actually be a good thing. >> >> Moreover, voters would hopefully become more generous about approving >> candidates in the primary that they're not yet ready to vote for in >> the general election, as a way of saying "hey, having Katie >> McKrazypants in the general election debates ought to make things >> interesting! Let's vote for her!". Over time, I suspect that >> candidates would get good at positioning themselves as good >> safe-to-fail experimental candidates. Then the general election >> becomes the safeguard where many of the same voters might then cast a >> more critical vote. Given that we'll have over seven months(!) to vet >> the general election candidates, we should assume that some voters >> will change their mind, and/or Katie McCrazypants would get thoroughly >> vetted and found to be not-so-crazy after all. >> >> A close candidate who gets 45% approval (e.g Cassie McCloseypants) but >> doesn't get quite enough approval to advance to the general election >> still sends a pretty strong signal. It would seem that at least some >> of the advancing general election candidates (e.g. Larissa >> McLeftypants and Ronald McRightypants) would come to understand that >> the 45% candidate (Cassie McCloseypants) was trying to advance some >> issues that are worthy of further discussion during the general >> election cycle. Both general election candidates would be well >> advised to seek out an endorsement from Cassie McCloseypants, and >> would have plenty of time to broker for support from the Cassie >> McCloseypants campaign (e.g. recruiting volunteers, getting email >> lists, getting contact databases, etc) >> >> Ted also wrote: >> > If the viable candidates are A_1 (= Approval Winner), A_2 (Approval >> > runner up), etc., with complementary opponents B_1, B_2, etc., then I >> > think it would be appropriate to add them in (A_i, B_i) pairs until your >> > desired representation level is met. >> > >> > Actually, I don't know if I would put the truly viable cutoff at 50%. In >> > a true jungle primary, you might end up with only 40% winners at the >> > highest. I might go down to 33. 3% A_i candidates if that's what it >> > takes to get at least 66.6% voter representation. >> >> The tradeoff that we're up against: providing genuine choice for a >> large part of the electorate vs properly culling the list of primary >> contenders so as not to overwhelm voters with options. Though I >> suspect your intuition is correct about almost all candidates scoring >> well under 50% in the first few elections, my fear is that you're >> being too generous for the long-term. The benefit of having a 50% >> threshold for _automatic_ advancement to the general election is that >> it makes the barrier high enough that it becomes difficult for rabid >> extremists to game the system. >> >> What I fear: as the system becomes routine, sophisticated candidates >> will learn to optimize their primary candidacies to achieve whatever >> the minimum percentage required to guarantee them a spot on the >> general election ballot. 50% is a respectable target. 33% is >> horrifyingly low target. With FPTP, we live in a world where each >> party optimizes for 50.1% inclusion by estranging the other 49.9% that >> aren't part of the tribe. I shudder to think about what tactics would >> emerge to optimize political tribes for 33.4% in/66.6% out. >> >> Mind you, I'm still proposing we develop a small loophole for allowing >> "marginally viable candidates" to advance, though only in service of >> increasing overall ballot satisfaction. It seems too large of a >> loophole to allow *every* complementary candidate to the runners-up to >> advance, doesn't it? Granted, it does at least provide disincentive >> for the dominant political party to flood the field with clones, but >> it also seems like it would be possible for the complementary >> candidates to be fringe, extremist candidates with very low approval >> scores. Perhaps the answer to that is to just set the floor for >> "marginally viable" to be 40%, and then only advance marginally viable >> candidates who increase the ballot satisfaction score. We can also >> still limit the number of marginally viable candidates who advance >> (the "opposition candidate pool") by maintaining the rule that the >> opposition candidate pool will be no larger than the majority >> candidate pool. >> >> I suspect that it would take a few years of usage before any >> second-place candidate gets above 50% approval. In fact, it may be >> common for even the Approval Winner (AW) to get less than 50%. >> >> With all this in mind, it may be worth reading Clay Shentrup's >> responses to my blog post: >> <https://medium.com/@ClayShentrup/i-challenge-this-858b5ab0a09a> >> <https://medium.com/@ClayShentrup/way-too-much-complexity-b4b2c8d6ee6e> >> >> In short, he's suggesting to just use the top two Approval winners. >> That seems likely to motivate candidates of the dominant party to >> unite as clones of one another, but maybe I'm missing something. >> Still, the KISS principle of his proposal is appealing. >> >> I think I'm pretty close to being able to write up draft 3 of MAF, >> which should be simple enough. What I have in mind now is basically: >> "allow all candidates who (usually) get over 50% approval, and allow a >> (usually) equal number of opposition candidates who get over 40% >> approval". The first "(usually)" acknowledges the possibility that >> there may not be any candidate that gets over 50% approval, and the >> second "(usually)" acknowledges that there might not be any candidates >> who qualify as opposition candidates, or that the opposition >> candidates may be outnumbered. >> >> Rob >> >