Evaluating Single-winner Systems From 2021-10-18 Until the Next Major Discovery
-
I suggest that at this point in the history of our discussions, Score Voting has earned such a special place that the most economical way to evaluate other systems is to start by showing a case where they outperform Score. This is because of Score's expressivity, simplicity, and balance. And by "outperform", I mean that in the context of the quest to defeat two-party dominance (2PD). And all of this assumes that the proposals are balanced, because otherwise, by definition, they are antidemocratic.
Both @rob and @cfrank are proposing systems that although they collect Score-style ballots, they depend mainly on the ordinal distillation of the scores. What case shows that one of these outperforms Score, when the voters are in each fork of the trial (Score and system under test) using optimal strategy for the given system?
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 For one, defeating two-party dominance doesn't happen just by creating a great system. You also have to sell that system to the public. I'd take Score in a heartbeat over our current system, but I don't think it is marketable due to its strong incentive to be strategic. Especially with its "dirty little secret" which is that to vote most effectively under Score, you simply ignore its costly feature, which is to vote with "in between" values. The smart voter under Score will typically give either a minimum or maximum value to each candidate. When people figure this out -- and it would become quickly apparent to everyone as soon as Score starts to gain momentum -- they aren't going to like it. "Why pay more for a dimmer switch when the light bulb really only works when it's all the way on or all the way off?"
So my goal is to offer a system that is as predictable as possible. And that means, make it so hard to gain an advantage by being strategic, that the vast majority people won't bother. I think any Condorcet system accomplishes this, and STAR does as well (although I think STAR may still incentivize strategic nomination, which is a problem).
I don't know what you mean by "what case shows...?" Are you looking for an example, a study, or for us to "make a case" for our particular system?
Regardless, for the most part, I am not really trying to make a case against Score, because I don't think Score has many people still arguing for it. Most of its fans have either joined the STAR camp or the Approval camp, or are possibly doing what I am doing and seeing if there might be something yet considered, that is more likely than either of those to gain widespread traction.
-
@rob Let me revise my original suggestion in that where I said 'defeat 2PD', what I should have said is 'defeat the oligarchic control mechanism discerned in Warren D. Smith's "Cash3" screed'. That is, take away the incentive to chase, to the exclusion of candidates who show merit, a bandwagon as indicated by money support (or fame).
You mention Approval. Approval is a further specification of Score, to the variant of it whose range has just two values. So, it's included. As for STAR, it is likely to get results that are almost as resistant to dominance by money as those from Score. You just support your favorite pair of clones and to a slightly lesser extent, your compromise pair of clones.
And yes, by "case," I meant an example starting with true voter sentiments, then proceeding through strategy.
Marketability is important, but I think we have to predict performance first. IRV has been marketed with more success to date than any other reform, but we all have seen the grounds to expect it not to perform adequately. I think that once we think we have good information about the systems that will perform, we can look at how most effectively to market them. For example, in a recent post, I suggested marketing Approval as Disapproval to forestall the charge of non-OPOV.
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 said in Evaluating Single-winner Systems From 2021-10-18 Until the Next Major Discovery:
And yes, by "case," I meant an example starting with true voter sentiments, then proceeding through strategy.
I stand by the example I provided in the other thread. I think many voters on the left, with imperfect knowledge but behaving reasonably and realistically, would have been much more likely to lower their vote for Gore in 2000 due to Nader entering the race under both Score and STAR. Under Score, that would have significantly hurt Gore, under STAR, it wouldn't have. To the extent that Warren Smith's short article discusses money's influence (which I personally think is a stretch to say that better voting methods address that), I stand by that Score would be worse than STAR, since under Score a candidate like Nader would be more discouraged from running for reasons I noted.
My priorities are a good bit different than yours. Polarization is the main issue that I think is problematic under our current system, and I am genuinely scared that we are on a path toward another civil war in the US due to polarization caused by both the voting system and by social media so quickly emerging as a way people get their information about the world.
@Jack-Waugh said in Evaluating Single-winner Systems From 2021-10-18 Until the Next Major Discovery:
Marketability is important, but I think we have to predict performance first
Ok. As I said in another thread, systems that require or strongly incentivize strategy are very hard to predict due to the "hall of mirrors" effect, with each voter trying to guess what other voters are going to do. STAR and especially Condorcet methods address this. I don't think you can easily "predict performance" with Score. The system seems to be designed for you to vote in a way that is counter to good strategy, leaving it as an exercise in psychology to predict -- far more so than on better systems.
And STAR has more momentum than Score, the latter of which has been around for a long time and to my eyes, is dead in the water.
-
I tend to agree with @rob, I think STAR is a great system, and that political polarization is the main issue in the U.S. I am less keen on Condorcet methods because, in my opinion, they are too rigid/majoritarian, whereas standard Score is too elastic and subject to the negative effects of strategic voting. I believe STAR does a great job of finding a middle ground, my qualm with it has to do with the somewhat arbitrary nature of the cardinal score values, and the fact that rather than reducing the magnitudes of the majoritarian and strategic force vectors individually, it simply pits them against each other. But you can't have it all, and that's probably close to the best we can hope to accomplish, although it would be nice to reduce their magnitudes more directly as well. I hope my analogy makes sense, I have already described this in the previous forum but I'd love to continue the discussion.
In any case I think adopting any one of the voting systems we are talking about here, whether in federal, state, or local elections, would constitute a giant leap forward for democracy. Score is definitely dead in the water, especially compared with STAR. My intuition is that the framework of STAR voting could be easily applied to an ordinal score system (for example something along the lines of "SP Voting") and that the combination would greatly reduce the effects of majoritarianism and of strategic voting.
But that's a lot of armchair philosophizing. If we want to know how and when and why certain systems perform better than others, we will need to do an in-depth comparative analysis, ideally one that is both experimental and theoretical. I think it would be a great but ambitious project to organize a series of game elections, where there is a certain number of candidates, and at each election, each player is made aware of the number of points they will receive according to the winner(s) of the election. Each "board" would be a different voting system. Somehow voters would need real incentives to accumulate points. I mentioned using gambling/stock markets before, I think in a controlled environment it could be a very effective tool for understanding how voting systems operate, especially if we could analyze the ballots of the electorate given their point incentives and also analyze the distribution of points as the system stabilizes. I only mention this because the topic is concerning the evaluation of systems, and I think any evaluation that matters for practical purposes needs to be holistic. I think we should follow suit with the behavioral economists, theorizing and developing special cases seems too much like wheels spinning in mud to me.
-
Whether plain Score with more than two options is "dead in the water" with respect to public appeal or not is irrelevant to the question of whether it has earned a place as the baseline against which to compare all proposed systems as a first step to determining whether they are worth any further attention. If you think the power relations created by Score are poor from the viewpoint of resisting control by money, you should find it that much easier to show that your shiny system outperforms it so as to get past this first gate.
I am not impressed by the hall-of-mirrors problem and my tentative belief is that an adequate strategy with Score is to support ones favorites fully and ones compromise candidate (or clone set) at about .99 if you think your favorite is only 1% popular. I predict that this strategy will work better than voting Approval-style, which I think @rob is saying in another thread is the top strategy. I base this opinion on having played around a very little bit with a few examples. Also, it's just obvious that always voting in Approval style will at best elect the compromise.
Simulation may cast light on a comparison of strategies and I am working on a simulator. Existing simulators are great, but I have the idea of making the some of the parameter values for the problem under simulation follow a slider control. I want a researcher to be able to use the slider to control the proportion of small-d democrats in the electorate. I will have the simulator give updated results as quickly as possible whenever the user moves the slider. This should make it possible by titration to see where each combination of voting system and strategy in the test transitions from electing plutocrats to electing small-d democrats.
-
@rob said in Evaluating Single-winner Systems From 2021-10-18 Until the Next Major Discovery:
My priorities are a good bit different than yours. Polarization is the main issue that I think is problematic under our current system, and I am genuinely scared that we are on a path toward another civil war in the US due to polarization caused by both the voting system and by social media so quickly emerging as a way people get their information about the world.
There are two kinds of fascist in the US. One kind only wants to continue with corporatism while maintaining a façade of at least a civil society, where decisions are made by some kind of peaceful interaction (even if they result in warlike results like the deaths of poor people from air pollution, starvation, and so on). Then we have the growing faction of the other type of fascist, the type who wants to simply rule by force, as though might makes right.
My Marxist informant used to say that WWII was just a conflict between different elements of the same ruling class, and the narrative promoted at the time that it was a fight against fascism was just false and misleading propaganda crafted for the purpose of organizing the cannon fodder. Nonetheless, from the point of view of the typical Allied soldier, the point of the war was to prevent the fascists from taking over the world, by any means necessary, including, if necessary, putting many of them to death. They did not have the analysis I gave above about two kinds of fascist, but their understanding of what a fascist was conformed to what I now call the second kind.
I think it may well be the case that the only way to counter fascism of the second kind in the US is to return to armed violence. Do they understand any other language? When I ask Trump supporters on antisocial media whether they have the courage to accept OPOV, no one responds. What strategy involving voting systems will have an effect toward quelling their movement? For sure, it has to be as transparent and simple as possible. These anti-intellectual know-nothings will never follow a complicated line of reasoning. It's more than hard enough just to get small-d democratic dedicated political volunteers to follow one.
-
@Jack-Waugh are you referring to my system as shiny or are you talking to @rob, or both of us, or speaking generally? Not impassioned here, but I’m definitely happy with SP Voting on a theoretical level and also believe it could be implemented in a way that is almost as simple and straightforward as Score or STAR. It definitely outperforms Score, I am certain of that, I can try to organize some simulations to convince you but I’m not that tech savvy or idle. I can send you the code and elaborate on it or share it publicly, etc. If large factions of the electorate consistently split-bullet vote, then split-bullet voted candidates become run-of-the mill noise to the system, and room is made in the middle for people who want to express their preferences with more nuance to be heard.
The theoretical backing of the system is superior to Score. I am not just saying that because I developed it, although I did develop it with that specific purpose in mind—I’m saying that because it is based first and foremost on the concept of a candidate being consensually evaluated on a global scale, rather than on a candidate locally amassing score points with arbitrarily assigned numerical values that gamify what is supposed to be a workable substitution for large-scale social agreement.
In SP Voting, a voter doesn’t know the quantitative value his ballot will contribute to the candidate, because that depends explicitly on how everybody else votes as well. In fact, in SP Voting, individual voters have no ability to add explicit points to the candidate’s score. Only the whole collective electorate can do that. They can certainly choose to form coalitions, or be bribed or what have you, but those broader issues aside there is never really any conflict of interest that the voter faces in casting his ballot, other than the question of how they should express their preferences using the ordinal scores. They know only that scores get better going from left to right, and the question of, for example, “How should I rate my second choice?” becomes totally subjective, as it should be.
Like I said, together with a STAR framework, I think it would be an excellent system. I don’t really care about eliminating voter regret or whatever, I think people will obviously feel dissatisfied to an extent because that’s always going to happen with a successful compromise, but obviously reducing it all else equal is a plus. It’s better to evaporate and spread voter dissatisfaction thin than to let it conglomerate into angry mobs (especially two big Red and Blue ones) that can be easily manipulated by opportunistic parasiticians (am I saying that right?). The STAR framework seems to reduce voter regret and encourage honest preference indication on ballots, which I do believe to be indicative of compromise.
-
@cfrank said in Evaluating Single-winner Systems From 2021-10-18 Until the Next Major Discovery:
are you referring to my system as shiny or are you talking to @rob, or both of us, or speaking generally?
I am referring to every system that isn't Score Voting.
But when I proposed Score as a baseline for this kind of comparison, I was assuming the criterion all of us care about is how effective a proposed system is at resisting plutocratic control that happens via vote splitting -> inequality -> bandwagon chasing -> money chasing. Rob has said that maximizing the criterion of effectiveness against money control gives different results from maximizing some sort of qualities that will forestall a hot civil war. And he says he thinks the anti-civil-war criterion is more important than the anti-control-by-capital criterion, so it should be the central motivation in designing and choosing systems. I am waiting for him to explain what properties he expects to have this effect, and his grounds for thinking they do.
-
@Jack-Waugh that’s reasonable. I think both are important considerations, but while there is always going to be conflict I do think avoiding a bloody conflict would be ideal. I am of the opinion that tamping down on capital control is a much loftier and less realistic consideration than quelling political divisions. Even Marxists are selling books. Of course a Marxist would view that only as postponing the inevitable. But maybe postponement will give more time to let the passions diffuse in a steadier and more controlled way rather than as an explosion.
Any system that encourages compromise and empowers the electorate over the politicians would probably have beneficial effects in terms of reducing political tensions. Easier said than done, politicians will fight tooth and nail to hold on to any power they can keep. I think most of the tensions are fabricated intentionally via identity politics anyway.
-
@Jack-Waugh I also want to make one point about systems being “balanced.” If you are talking about a formal property like Frohnmeyer balance as a criterion by which to judge the democratic quality of a system, I explained in the previous forum how that quality can be artificially induced into a slightly modified version of literally any voting system whatsoever without significant alterations in the election results. So it’s actually a bit of a red herring argument against a system anyway. The only somewhat legitimate but hazy argument people seemed to have against my construction before was that it somehow went against the “spirit” of balance, whatever that means. To me it means the Frohnmeyer formalism is a failed one and perhaps a new one is necessary. Until then it’s a fuzzy concept that’s up to interpretation, and because of that I can’t take it as more than fluff used to support the biases in favor of systems that are already preferred for other reasons.
-
@cfrank, would you cite that, please?
-
@Jack-Waugh I will try, I’m on my phone and will try to link the past discussion here soon.
-
-
@Jack-Waugh a 100% correct guess lol. I stand by my reasoning there, I hope you find it sensible. Anyway, it’s sort of neither here nor there, just a small point of conversation. I personally think the very concept of Frohnmeyer balance casts voting systems into a combative rather than a cooperative framework. I honestly see no reason why ballots should be required as a prerequisite to conform to that artificial structure, it seems to me at worst almost like an ad hoc dogmatism invented to justify supporting Score over other systems, and at best an example of paradigmatic confirmation bias.
-
I'm pretty sure that where @rob writes as though your modification if applied to FPtP would produce "vote for and against" (F&A), it doesn't.
Anyway, your proposed transform can be made more extreme to illustrate your point even better.
Suppose a vote in the modified system consists of these parts:
- a vote in the original system (the vote I'm trying to cast),
- another vote in the original system (the vote I'm trying to counter),
- a polynomial of degree 8,
- the factors of another polynomial of degree 8
The tally looks for matching pairs where each one is trying to counter the other and each one factors the other's polynomial. The matching pairs are eliminated. The remaining votes are counted as just their first part, as in the original system.
Before seeing this kind of example, I wanted to say that a decent single-winner voting system would meet Frohnmayer balance (F-balance) and also would be as expressive as Approval. But you have shown that F-balance doesn't mean anything without that Abelian stuff you cited (I don't live and breathe group theory, so am not, off the top of my head, sure what laws are implied). What about that, @SaraWolk? All of equal.vote got shot down.
Anyway, I stand by my proposal that the first step of evaluating a single-winner system that isn't Score should be to give an example, starting with true voter sentiments and proceeding through strategy, where the system outperforms Score. I don't ask for a proof that the strategy used with Score is the best. It will suffice to use my strategy. Favorite (and clones) gets top score, lesser evil (and clones) gets almost the top score, and everybody else gets the bottom score. "Almost" means as close as the popularity of the favorite, as a proportion.
But again, this is conditioned on defeating the plutocracy as the top goal. If @rob explains what constraints and criteria would describe a system that should be judged as most suitable for defusing a hot civil war, everything could change.
-
@Jack-Waugh you’re right about For-and-Against, the modification I suggested is much less significant because it requires the pairs to match exactly, whereas For-and-Against counts all the positives and all the negatives. Just another example of how arbitrary the concept of “balance” in this formalism is. I liked your strengthening of the argument, you’re absolutely right about that as well, because it virtually eliminates any level of information voters might have about other ballots.
Maybe there is something more tangible that F-balance is trying to point to, but it isn’t clear what it might be. In my opinion, with respect to the concept of balance, the strengthened construction should by all reason send equal.vote back to the drawing board, especially if they can’t even formalize a valid counter-argument.
