The Alaskan Top Four Model & iEBs
-
@gregw you are correct that STAR has clone issues, specifically it is clone positive (for paired clones) and hence it can in principle almost totally eliminate the purpose of the runoff and regress to an approximation of Score with roughly twice the number of candidates.
Score has issues with sensitivity to tactical voting, such as bullet voting and min-maxing, which can cause it to regress to an approximation of approval voting. The tactical concerns are valid, but in the end it’s a prisoner’s dilemma which is difficult to analyze. Let me put it this way: is traffic an issue?
Approval voting theoretically converges to a Condorcet method under perfect information and tactics. Condorcet methods are generally robust, but are also restrictive in the sense that they depend on rank (ordinal) information rather than score (cardinal) information. Approval is also pretty restrictive in that sense. In theory, the Condorcet winner (if they exist) may not be pleasing to the more passionate populations of voters, which can lead to substantial double-sided dissatisfaction. This is because Condorcet methods, if contorted into the framework of being a utilitarian model designed to minimize dissatisfaction—which they are not—effectively assumes that all marginal dissatisfactions between candidates are equal, both between any two candidates, and for any two voters. This clearly is not true, so the Condorcet winner may not be ideal in a utilitarian sense. (I’m not a utilitarian anyway, per se).
A new voting system we developed here called SQNV (symmetric quantile-normalized score) avoids many of these issues, but is also very experimental, and is likely not suitable for practical implementations because its mechanism is somewhat complex (for a layperson not as versed in voting theory/data science/mathematics) and would require the use of a computer program to run efficiently (which is a security issue and therefore also a legitimacy issue).
In my opinion, the nonlinear, symmetric, Borda-like methods probably constitute one of the best families of voting methods.
By this I mean, for each integer m (the potential number of candidates -1), take some corresponding increasing sequence q[0], q[1] …, q[m] of real numbers in [0,1] such that
q[m-k]=1-q[k]
and use q[k] for the Borda score value associated with rank k among the m candidates.
Choosing the score scheme in this kind of method introduces an arbitrary aspect. The choice is exactly analogous to selecting an activation function for a (very simple) neural network. SQNV is actually a nonlinear symmetric Borda system with a dynamic (rather than static) activation function. @Jack-Waugh gives an example of a ballot that could serve as a nonlinear symmetric Borda system with static activation in his initial post here. Static activation is more transparent and easier to tally without a computer program, but it is also arbitrary, and the question of “which voting system?” then gets deferred to “which activation function?” SQNV, on the other hand, constructs the activation function from the input ballots in a (hopefully) less arbitrary way.
If you’re interested in a pretty comprehensive list of voting systems to investigate, you can check certain users’ “signature” ballots, mine is just below
vvvvscore-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]
-
@gregw said in The Alaskan Top Four Model & iEBs:
what would be the best voting systems for rounds one and two?
My opinion is that for so long as the purpose is to elect a single winner, the best options for both rounds would be systems that conform to Wesberry vs. Sanders, which is to say, conform to Frohnmayer Balance. The simplest of these is Approval. More generally, the rest of the Score systems also conform.
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].
-
@cfrank said in The Alaskan Top Four Model & iEBs:
@Jack-Waugh gives an example of a ballot that could serve as a nonlinear symmetric Borda system with static activation in his initial post here.
Eh? I don't understand this characterization at all and in particular, I didn't think there was any similarity between my 2024-03-07 system and Borda, which converts rankings to ratings.
In any event, I would not recommend it for Alaska because I believe it does not support the strong security assurances needed in a public, political election.
-
@jack-waugh Borda is a positional score system, I didn’t mean your entire method, only the scores you used in the ballot. The scores you indicated were
{q[0],q[1],q[2],… q[m]}={0,1,10,50,90,99,100}
which, when normalized by dividing by 100, satisfies
- j>k implies q[j]>=q[k]
- q[k]+q[m-k]=1
and is also nonlinear. It’s like a discrete sigmoid activation function in a neural network.
-
@cfrank
Thank you for your insights! First, I should admit that compared to the everyone else on this forum I am mathematically challenged.The organization I am starting, Voters Take Charge, will advocate for specific (often with options) reforms. We will also be reactive; we will need to decide whether or not to support proposals made by other organizations, hence all my questions about Top Four primaries.
You have inspired a couple more questions:
Is resistance to clones important in any election with similar candidates, be it a primary or any election with a lot of candidates?
Would STAR be able to handle a four or five candidate general election?
-
@jack-waugh
Your point that every voter should have an equal vote is well taken, thank you.Concerning Approval Voting, I have written an article about it for the soon to exist VotersTakeCharge.us website. It needs to be evaluated. Could I post it on a new thread?
What voting system would you suggest for first round vote that selects four or five candidates to run in a general election? Assume that the political parties have already chosen their candidates and those candidates must participate in the first round.
-
@gregw said in The Alaskan Top Four Model & iEBs:
Could I post [a draft article on Approval Voting] on a new thread?
I think that's a good idea. Did you judge that simply citing one of the existing writings on the subject likely falls short of your goals of what you want to emphasize to your intended audience?
What voting system would you suggest for first round vote that selects four or five candidates to run in a general election?
That's a deeply interesting question. Does it make better sense to assign as the primary purpose of the first-round vote, simply to sideline the worst candidates, in direct service to the goal of electing the best in the final, or does it make better sense to assign it the purpose of providing to the voters in the final round, as wide a choice of ideologies and values as feasible, crossing political spectra? Maybe a halfway-decent PR or nearly-PR system would fit there.
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].
-
Thank you for your interest in my Approval Voting article. I posted the article a couple of minutes ago. I posted under Advocacy.
The first goal for Voters Take Charge is to prove that American voters want proportional representation. Second, we will inform voters that there are better options than plurality and ranked choice voting.
My article (indeed most of my articles) is an attempt to distill the wisdom of leading election scientists into a clear and maybe concise document for my website visitors.
God willing, Round 1 of a two round system will eliminate the worst candidates and give voters a five distinct (nonclone?) candidates. To make such a proposal fly, we would need to use a system that is fair and understandable. Not easy.
BTW No labels announced that they are giving up their 2024 Presidential run for lack of a candidate. RFK jr. on the other hand is collecting nomination signatures.
-
@gregw said in The Alaskan Top Four Model & iEBs:
God willing, Round 1 of a two round system will eliminate the worst candidates and give voters a five distinct (nonclone?) candidates.
I hope some of the participants who are into comparing PR alternatives will take an interest in this question.
-
Yes. The various organizations promoting Top Four Primaries are proposing a "choose one" vote for the first round. I would like to offer something better.
-
@gregw said in The Alaskan Top Four Model & iEBs:
Yes. The various organizations promoting Top Four Primaries are proposing a "choose one" vote for the first round. I would like to offer something better.
Approval, for example, would be better than Choose-one Plurality. And it would be easier to explain to people than a PR system. It could be improved later, after people have become used to equality.
-
@cfrank said in Before a Real Runoff:
@jack-waugh I don’t know much about PR, but this is an attempt to balance seats by considering party affiliations without stuffing clones. Can PR be improved with “Cake Cutting” incentives?
I should think that a system that does not refer to parties would work better. Simplicity is probably necessary for selling it.
Maybe Reweighted Range Voting (RRV) using the same kind of ballots (maybe Approval) as the final will use. In PR discussions, when I have brought up RRV, the PR mavens have told me that it's not one of the best options. But they don't seem to be here today, so I'm hoisting RRV up the flagpole.
-
Another option for the first round of balloting and tallying is Asset Voting. Each candidate gets points for how many voters approved hir. The candidate having the least support gets to distribute it to whichever other candidates se chooses. Repeat this until the number of candidates is the number for the final plus one. Send the ones with the most support then on to the final.
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].
-
Your advocacy is elegant, I will change the opening as follows but leave the rest alone.
In support of replacing Plurality Voting with Approval Voting, eminent election scientist Jack Waugh asserts our right to better voting systems:
A voter should be the one to determine which candidates her or his vote supports and opposes. Plurality Voting does not respect this right. In an N-candidate election for a single seat or office, voters who want to oppose fewer than N - 1 are told to lump it. They are denied the right to cast a vote that reflects their political judgment.
I came up with a short version I will use on my plurality voting page:
As voters, we should have the right to choose which candidates we support or oppose. Plurality Voting does not respect this right; we can only support one candidate and oppose the rest.
Some voters are happy to support only one candidate, they will choose the lesser of two evils and call it democracy. They have the right to vote as they will, but the rest of us suffer under their limit. Should we not have the same right to vote as we will?
Voters Take Charge was created to demand that our right to vote as we choose is fulfilled. We are even more subversive; we want an election system that puts voters in control of public policy. Therefore, we demand Proportional Representation and better voting systems for single-winner elections.
Lime also suggest RRV, definitely better than STV.
Ambassador quotas and asset voting are both fascinating ideas. I can see running both up the flagpole in the future.
I am also tempted to propose something akin to DPR voting (Direct Party and Representative Voting) on steroids. Take a PR election in a five-member district; at stake are the district’s votes in the legislature, which are equal to the population of the district. The top five candidates split the district's votes based on the percentage of the vote each candidate received. The legislature would need to rewrite its rules of operation. That is a feature, not a bug.
For now I will go with your suggestion of Approval Voting for the first round of a Top Five election. Simple and fair.
I recently voted in a nonpartisan mayoral election with seventeen similar candidates and a top two runoff. Top five with approval voting would have helped. But next time I'm just going to vote for the candidate that uses the phrase "multi-modal transportation" least often. A multi-modal transportation boondoggle exists about 150 feet from my back door.
-
I'm not an election scientist, let alone an eminent one, more of a voting-systems enthusiast. Anyway, I don't need you to credit me for those words of advocacy.
Regarding "Plurality", note that Approval is decided by plurality. That's why I call FPtP "Choose-one Plurality Voting".
-
In regard to control of pubic policy, look up "Liquid Democracy".
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].
-
Liquid Democracy would be exciting and much more direct than ballot initiatives. It would drive party leaders nuts.
I would love for a few small but real governments to give it a try.
-
@gregw said in The Alaskan Top Four Model & iEBs:
small but real governments
Smaller than the State of Rhode Island And Providence Plantatations?
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 was thinking in terms of a small municipality, but if Rhode Island wants to go for it, great!
-
Before I look into the rest of the stuff in this thread, I notice that RRV has been mentioned a few times. This is a score conversion of Thiele's Proportional Approval Voting, or PAV. However, I do not think it is the best conversion. The method known as Sequential Proportional Score Voting (SPSV) has better criterion compliance (it passes the additive and multiplicative versions of scale invariance*), and is no more complex.
*Basically if you multiply all scores by a constant and/or add a constant to all scores, the result remains unchanged.
