The dangers of analysis paralysis in voting reform
-
@Lime recently raised a point in another thread about how they were interested in variations on the STAR Voting theme, but not at the cost of real world implementation and adoption.
I'm happy to give up the half-star if it really does increase support in polling or experiments.
Interestingly, right after reading that I came across an old email with a staffer Dan with election vendor ClearBallot.
Dan writes:
I'll forward your message to our engineering team so that they can include it in their research for these types of elections. The main problem facing proponents of these types of elections is that there doesn't seem to be any one set of standards coming to the front. There are almost as many methods for determining winners as there are proponents, which makes it almost impossible to put real engineering resources into it yet. We are gathering as much information as we can, but I think it will be a while before we can build this into our certified software, which is what would be required to actually conduct an election this way. ...
I would suggest that one of the most important issues facing proponents of alternate voting methods is to work together to come up with a standard set of methodologies that can be given to government jurisdictions so that vendors have a common standard to work to. Until then, we're kind of shooting at a moving target, and it wouldn't be very smart to put too much effort into it. We want to be ready, but we can't support dozens of different algorithms for doing it.
The moral of the story is that the existence and promotion of too many voting methods is a barrier to the adoption of any of them.
-
@sarawolk There's always more work and analysis that can be done. So then at what point do you decide you have the method that should be promoted in full?
-
@toby-pereira Not before the invention of STAR Voting, but after.
In all seriousness, I'd say when we have a proposal that delivers on the most important goals and addresses people's most important concerns.
It's subjective and there's no hard line in the sand, but it's important for the research community to keep in mind what areas desperately need more research and in which areas that can be derailing or lead to diminishing returns.
There are a lot of places we desperately need more research and more high quality research.
-
@sarawolk said in The dangers of analysis paralysis in voting reform:
@toby-pereira Not before the invention of STAR Voting, but after.
STAR is great! I'm sure there's ways to slightly improve on it, but what matters now is probably public opinion. (Relatedly, has anyone done any head-to-head polling matchups of Score vs. STAR?)
-
@lime Among people who know about them or know about voting methods already, or among people who have never heard of them and get a simple explanation and form an intuitive opinion?
-
@sarawolk People who have never heard of them. (My main concern is that the automatic runoff might add complexity that keeps people from supporting STAR. It seems simple to me; then again, so does IRV, but the most common complaint people have about it is complexity.)
-
@lime Our experience has been that Score and Approval are non-starters for many of the most influential groups, like League of Women Voters, because of issues or perceived issues and incentives for strategic voting.
I think both are very, very good systems, but do think that these are valid concerns that STAR addresses and that need to be addressed for political viability's sake.
The STAR runoff changes the voter incentives so that they should show their honest preference order. Score voting encourages min-maxing. (All 5s and 0s.)
Worse, in Score Voting if a voter fails to use the full range of the ballot then their vote is less powerful than a voter who did. Since a rating is commonly used with the top score being a perfect/love rating, it's an expected behavior that some voters, especially people new to the system, will give their favorites lower scores because almost no politicians are actually loveable. This is a massive issue in practice, and also in terms of optics. The STAR runoff fixes this by effectively normalizing the ballots in the final round.
Approval is another great system, but the fact is that voters need to approve the front-runner on their side (and everyone they like better than them). That means that a voter who doesn't like the frontrunners has to lie and approve one anyways to have an effective voice. Alternative voting advocates in my experience (especially 3rd party voters) want to be able to show that they prefer their favorite over the frontrunner on their side.
So, yes simplicity is better if we can get it, but I think that STAR Voting is as simple as possible while still doing what people want voting reform to do. The system has to be able to inspire massive support to get passed and not every proposal has that kind of inspiring chemistry.
For the record, I would give Score Voting 4 stars, and I'd absolutely approve Approval Voting and most alternative voting methods. I would not approve IRV.
-
@sarawolk said in The dangers of analysis paralysis in voting reform:
Worse, in Score Voting if a voter fails to use the full range of the ballot then their vote is less powerful than a voter who did. Since a rating is commonly used with the top score being a perfect/love rating, it's an expected behavior that some voters, especially people new to the system, will give their favorites lower scores because almost no politicians are actually loveable. This is a massive issue in practice, and also in terms of optics. The STAR runoff fixes this by effectively normalizing the ballots in the final round.
STAR only fixes it for the final round, so the problem still exists when determining which candidates make the final.
-
In any case, I would prefer a Cardinal Condorcet method to STAR. By asking for scores rather than ranks, it may protect against fairly weak Condorcet winners.
E.g.
49: A>>>>C>B
49: B>>>>C>A
2: C>A>BWith rated ballots, many of the 98 that don't have C as their favourite would likely just score C zero.
I suppose my problem with STAR is that it's indecisive about whether score or head-to-head is more important. Head-to-head but only for the two candidates with the highest scores. It's a bit arbitrary.
I tend to think that Condorcet is probably better if done right, but as a class of methods rather than a single method, there isn't universal agreement and a co-ordinated push behind any particular method.
-
@toby-pereira It's important to have explicit instructions.
-
@toby-pereira said in The dangers of analysis paralysis in voting reform:
STAR only fixes it for the final round, so the problem still exists when determining which candidates make the final.
As long as the instructions are clear and explicit there's an argument that this is a feature, not a bug. Some people actually don't have strong preferences and don't love or hate any of the options, even relative to each other. This comes up in polls more than in elections, but we do see voters argue that the nuance between candidates is only one kind of useful nuance. They also like being able to show the difference between a strong preference and a weak one or an abstention.
-
@toby-pereira I really like Condorcet methods as well, 5 stars. Condorcet cardinal methods are best in class too, but I don't like that it's a bit misleading for voters. It looks like the level of support you give will matter, not just the preference order, but in fact the scores are almost never used in practice (unless there's a Condorcet cycle).
STAR has a more even balance between Condorcet and Cardinal and I like that. I think both approaches have strong benefits. In plant breeding and natural selection there's a concept called hybrid vigor. I think STAR has that.
-
@sarawolk As for why two. Two candidates is the number that guarantees a majority preferred winner whenever possible, and two is the number that eliminates spoilers. Top two runoffs have huge advantages, both mathematically and in terms of perception, and so STAR leverages that.
-
@sarawolk said in The dangers of analysis paralysis in voting reform:
@lime Our experience has been that Score and Approval are non-starters for many of the most influential groups, like League of Women Voters, because of issues or perceived issues and incentives for strategic voting.
That might be the case, but I'm guessing there's a very big difference between what LWV thinks and what the average person thinks.
@sarawolk said in The dangers of analysis paralysis in voting reform:
Worse, in Score Voting if a voter fails to use the full range of the ballot then their vote is less powerful than a voter who did. Since a rating is commonly used with the top score being a perfect/love rating, it's an expected behavior that some voters, especially people new to the system, will give their favorites lower scores because almost no politicians are actually loveable.
I'm not actually convinced that's a problem. If someone wants to cast an honest ballot that says they're unsure, or that they think all the candidates are equally bad, that should be up to them. This is an honest vote expressing how strongly they care. Some people will be upset or happy with the result, no matter who wins.
There's two possibilities here: either the voter wanted to cast an honest (unnormalized) ballot, or they wanted to cast a dishonest ballot but didn't know how.
If they want to be honest, that's fine, and that's their right. We shouldn't stop them.
If they don't, that seems like two strikes against them: they're not honest, and they're not smart enough to strategize. Why give their ballot a higher weight?
I'm generally opposed to "automatic strategy" in cases where voters have chosen not to use it. Some kinds of automatic strategy can be good, if they make use of information that's not available to the average person. Otherwise, why are we using score or STAR instead of approval?
-
@lime But what is honest in this context? On a 0 to 5 scale, how much do you have to like your favourite candidate to "honestly" give them 5 stars?
-
@toby-pereira said in The dangers of analysis paralysis in voting reform:
@lime But what is honest in this context? On a 0 to 5 scale, how much do you have to like your favourite candidate to "honestly" give them 5 stars?
What's perfect honesty? I don't know. Interpersonal utility comparisons are hard.
But I can still do better than assuming everyone has equally-strong preferences; sometimes I just don't care that much about an election.
-
@sarawolk Yes, & so the obvious solution is to propose the absolute MINIMAL multicandidate voting-system. The simplest one:
Approval.
If you want something more elaborate, then you can always ask for it later. But, for now, first, get voting-system reform at all, by proposing & enacting the absolute minimal Approval.
It's the least to ask for. ...&, as you know, it's minimalness makes it the completely un-arbitrary method, & the most easily described, ,defined, proposed, enacted, implemented, administered, & audited for error & count-fraud.
Sure, the infinitely-many elaborations, such as rank- balloting methods, can be a convenience-luxury for voters. Rank-balloting, for example is an automatic-machine that takes your preferences as input, & outputs a result for you.
Whereas Approval is a handtool that requires something from the voter, a rank-method does it all for you. But the cost of that is the loss of all of the abovementioned Approval-advantages. I suggest that the problem of whether the voter will use Approval optimally is nothing compared to the problem of failure to enact the method, or being able to afford its implementation, or prevent count-fraud.
I should add that, though it's often pointed out that Approval's implementation requires only the deletion of the software code that prevents "overvotes"...event that isn't needed:
Say it's an Approval election between 5 candidates. Tell the count program that it's five 2-candidate races. The ballot-card only has a candidate name & a pencil-bubble or punch-place on every other line...the line for each actual candidate, On the fictitious candidate's line, the ballot card just has blank-space.
One each ballot, the count-software records your vote for the actual-candidate, & zero vote for the fictitious one. When the count is concluded, the software reports the vote-totals for all the actual candidates, & zero votes for each of the fictitious candidate.
Thereby is Approval implement4d without any software modifications. ...at absolutely zero cost.
So, forget all the elaborations that we might like, & just propose the method that's completely un-arbitrary because it's absolutely minimal: Approval.
Do you want reform, or do you want to diddle-debate forever about which elaboration is better?
-
[quote]
Our experience has been that Score and Approval are non-starters for many of the most influential groups, like League of Women Voters, because of issues or perceived issues and incentives for strategic voting.
[/quote]
…
When they express those issues, remind them that Approval, & not Vote-For-1, allows rating of each candidate, & thereby the expression & counting of support for as many as one wants.
…
Remind them that when you ask for too much, you often don’t get anything. Remind them that we don’t have time to debate forever about the best elaboration.
[quote]
Approval is another great system, but the fact is that voters need to approve the front-runner on their side (and everyone they like better than them). That means that a voter who doesn't like the frontrunners has to lie and approve one anyways to have an effective voice.
[/quote]
…
No, you don’t have to approve anyone that don’t like. Never approve someone that you don’t like.
…
…& if you believe that the frontrunners are both people you don’t like, then you’ve probably been very deceived by the mass-media. Don’t believe everything that you hear in the mass-media.
…
The Democrat & the Republican aren’t “the two choices”.
…
[quote]
Alternative voting advocates in my experience (especially 3rd party voters) want to be able to show that they prefer their favorite over the frontrunner on their side.
[/quote]
…
Condorcet (as RP(wv) ) would be the way to achieve that, but the complicated contraptions that can achieve that have too many problems, lacking Approval’s easy description, definition, proposing, enactment, implementation, administration, & auditing against error & count-fraud.
…
STAR? It’s in-between the extremes of Approval & the rank-methods…a compromise between them. Most importantly, it doesn’t have Approval’s absolute minimalness & resulting unique un-arbitrariness of choice. Its score-count & runoff require added software, where Approval doesn’t require any new software, or even any modification of current count-software (as described later in this post)
…
As I was saying before, yes STAR has some voters-helping features that Approval doesn’t have. But I claim that it’s better to trust to voters to do well with Approval than to take on the disadvantages of the more complicated methods, including any in-between compromise.
…
I want STAR to succeed in Eugene, & I’d vote for STAR’s enactment in an already-existing initiative or referendum if it’s too late to start an initiative or referendum for Approval.
…
But I suggest that all electoral-reformers, including CES & Equal-Vote, should, right now, be contacting the Oregon State Legislature & proposing that, in November, it shouldn’t just be an “RCV” referendum. It should instead be competing referenda for both “RCV” & Approval.
…
…or maybe they should just drop the “RCV” referendum due to its fraudulent promotion,
…
But, not only as a competing parallel referendum: If “RCV” is being voted-on for all Oregon elections other than the Legislature, then why shouldn’t Approval be voted for in a referendum for enactment for voting for state Legislators? (…or at least one house of the Legislature).
…
So, two possibilities: 1) Competing parallel referenda (but maybe just drop “RCV”); & 2) A referendum for Approval for election of state Legislators.
…
Ideally, RP(wv) & the Finnish system of open-list proportional-representation should be offered too, just in case people might prefer rank-voting or PR…but it’s probably a bit late in the year to propose something that isn’t minimal, as Approval is.
…
(…but it should be noted that Finnish open-list PR wouldn’t require any new balloting-equipment or count-software modifications. When the vote-count has been reported, the seat allocations to the parties & to their candidates could be determined at any kitchen-table where there’s a hand-calculator.)
…
BTW: The Greens aren’t 3rd-Party. If everyone who would prefer the things that they offer that polls say most people want, knew what is offered & voted honestly, the Greens would be 1st-party. …& one of the 2 frontrunners.
…
As Robert Bristow-Johnson points-out, yes Approval voting can be tactical. So what. You don’t know whether you should approve your 2nd-choice? Neither do the other voters, so don’t worry about it.
…
Which feels more important?: The danger that your favorite will lose to your 2nd choice, or the danger that your 2nd choice will lose to worse candidates? You could just vote according to which danger feels worse.
…
Yes Approval is a different kind of voting, compared to rank-balloting. Tactical, sure, when you aren’t using a rank-method to do it all for you. There are many ways of voting, & many ways of making the choice. But that’s a good thing for the voter, because s/he can do it any way s/he likes, any way s/he feels like.
…
How complicated the choice is, is up to you. Often we might have the goal of maximizing our expectation. Expectation & probability are subjective. A shuffled & cut deck of cards is on the table. I take the top-card & look at it, not showing it to you, & put it in my pocket, & ask you what is the probability that the top-card is now the ace-of-spades. To you, that probability is 1/52. To me, it’s either zero or 1/51. Neither of us is wrong.
…
Expectation depends on what information you have & choose to use.
…
Situations differ in what kind of information we have. Maybe the frontrunners are the most reliable information.
…
Best-Frontrunner is a valid expectation-maximizing strategy in Approval. But don’t take the mass-media’s word for who the frontrunners are. The Democrat & Republican aren’t really “the two choices”. If you use Best-Frontrunner don’t be a sucker about who the frontrunners are.
…
If you don’t know or feel who the frontrunners are, then that strategy isn’t for you. There are other methods. They’re versions of voting honestly.
…
There’s the Better-Than-Expectation strategy: Vote for all the candidates who are better than what you expect from the election. Which candidates would you rather install instead of holding an election? Which candidate is so good that hir election would pleasantly surprise you? Approve hir & all that are better.
…
When you approve above your perceived expectation (& others do too), you raise your expectation.
…
Never approve below your perceived expectation. That would lower your expectation.
…
Or maybe your best information is about who’s the likely CW.
…
One way to judge expectation-maximization is the reasonable assumption that the CW is the best that each voter can get, & therefore hir best choice for bottom-approval: Approve down to the CW.
…
So, if you have a good estimate of roughly where the CW is, in the candidate-lineup, that can be a good way of voting: Approve the likely CW & everyone better.
…
In the Internet polls I’ve seen, the Green is usually the CW. Sometimes Bernie, but usually the Green.
…
Yes that’s right: No need to vote for anyone who isn’t as good as the Green. …in Approval or Plurality. The media have deceived you about what the other voters want. They want a lot better than what the Republocrats allow. Universal medical care, higher taxes on the rich, more equality, not always being in wars. Polls & surveys indicate that the public are incomparably more progressive than their Republocrat “representatives”.
…
Also, notice that in Burlington & Alaska, when the CW was eliminated, the transfers & the win went in the progressive direction. …because the median voters chose someone more, instead of less, progressive. …further confirming that the voter-median is progressive, & that the CW is among the progressive candidates.
…
Don’t let anyone tell you that the Democrat & the Republican are “the two choices”.
…
Another way of identifying the CW is: Who’s the best candidate such that the people preferring hir to the worse ones amount to a majority? Maybe that’s the strongest or perceived-most-reliable information that you have.
…
Or: Approval is Set-Voting: Ask yourself if there’s a set that you like better than the rest, & approve them.
…
Or just approve the ones that you like. That maximizes the likelihood of electing someone you like. What’s wrong with that?
…
Or, if you feel like it, vote for the ones you feel to be above the candidates’ average merit. Or, if you don’t have that good an estimate of merits, then the most likely guess for who’s above-average is that the best half of the candidates are most likely the above-average ones.
…
Or sometimes there’s a particularly large merit-gap somewhere among the candidates. Among these possibilities, that’s probably the most likely merit-fact that we perceive among the candidate-merits. So approve the candidates above that largest gap.
…
That’s justified by the fact that, though Approval doesn’t allow you to vote all of your preferences, it allows you to vote the ones that you consider the most important. That largest merit-gap is a good measure of that.
…
…& it’s supported by the Set-Voting mentioned above.
…
But all of those strategies are moot for me, because I claim that our elections are merit-dichotomous: Two-valued merit. There are completely unacceptable candidates.
…
So, for me, Approval voting is really simple, & is the perfect fit to how I judge the candidates:
…
Just approve all the acceptables & none of the unacceptable.
…
But, if you don’t perceive it that way, you could just approve the ones that you feel like approving. If there’s no other information, then that maximizes your expectation based on what you know or feel.
…
It could be argued that there’s an objective optimal way of voting: The one that would give the highest total utility if used in a long sequence of elections. …if you knew what it is. If you want to pursue that, then you use the maximum amount of available information, the most detailed information, to calculate which vote would maximize that objective expectation.
…
For one thing that would require a lot of really unreliable probability-guesses, & difficultly-quantified numerical utility-estimates for the candidates. The great unreliability of that input-information is obvious, & it surely isn’t worth the great amount of work in calculating an optimal vote based on it. So forget that kind of strategy. Forget about “objective optimal vote”.
…
The voter needn’t have all these possibilities explained to hir. S/he’ll already like & use one of them. This multiplicity of ways of voting isn’t a complication. The beauty of Approval voting is that one can use any of those ways of choosing (& maybe various others too). Whichever one feels the best, is the best for you. It maximizes your expectation based on what you know or feel.
…
Lastly, it should be pointed out that Approval’s Myerson-Weber equilibrium is at the voter-median. Approval will soon home in on the CW, at the voter-median. People who criticize Approval because they want it all done for them by a rank-method forget about that Myerson-Weber equilibrium.
…
A Myerson-Weber equilibrium is an outcome that confirms the voters assumptions that led voters to vote as they did.
…
Our “Plurality” Vote-For-1 system can keep on electing ANY pair of candidates, however disliked, forever at Myerson-Weber equilibrium.
…
…which of course is what it’s been doing, & will continue to do unless & until we replace it with Approval.
…
As I said, Approval is a completely different kind of voting, compared to rankings. Approval is a handtool that leaves it to you. …& thereby it’s the absolute minimal method, & therefore it’s the unique completely un-arbitrary method-choice.
…
But what if, despite these considerations, progressives are hopelessly enamored of & wedded to rankings, due to the heavily-funded “RCV” (Hare) promotion? Then:
…- Offer them a better rank-method. Yes, IRVists point out that there’s much to be said for not electing a middle CW compromise who is unliked. Yes, when there’s a Mutual-Majority, Hare is guaranteed to choose the favorite candidate of the largest faction of a Mutual-Majority. But the IRVists are forgetting something when they use an argument about an unfavorite compromise:
…
They’re forgetting that Hare is strategy-ridden & sucker-vulnerable. To avoid strategy-problems, it’s necessary to elect the CW, even if s/he might be unfavorite.
…
So then, if progressives demand rankings, then offer them Condorcet, in its best version. Its best version is Ranked-Pairs(wv).
…
Among the people who discuss voting-systems & compare rank-methods, Schulze & Ranked-Pairs (RP) are the popular methods. …& are regarded as the kings of criteria compliance.
…
…&, additionally, RP(wv) is very strongly probabilistically autodeterrent for burial strategy, & invulnerable to offensive-truncation strategy.
…
So, not only does RP(wv) meet the Minimal-Defense Criterion (Just don’t rank anyone you don’t like, & if the CW’s voters all refuse to do so, then no one they don’t like can take the win from the CW via burial). …& also, RP(wv) is invulnerable to offensive-truncation strategy.
…
But, additionally, even if no one uses any defensive-strategy against burial, RP(wv) automatically probabilistically deters burial. I call that “autodeterrence”. RP(wv) is very strongly autodeterrent.
…
If there’s a circular-tie in Condorcet, it will almost surely be a strategic one. That’s the only thing that can prevent Condorcet from electing the CW, & thereby delivering on the promise of counting & acting-on every preference voted by every voter.
…
RP(wv), by thoroughly deterring burial, & thwarting offensive-truncation is the most strategy-free Condorcet. Yes, MinMax(wv) & Smith//MinMax(wv) share RP(wv)’s strategy-deterrence, but RP (along with the more complicated Schulze) is one of the two most favorite & prestigious methods among the people most familiar with voting-systems.
RP(wv) definition:
…
If there’s no voted CW, due to a top-cycle, then:
…
Drop the weakest defeat in each cycle.
…
Elect the resulting undefeated candidate
…
(The strength of a defeat is measured by how many ballots rank the defeater over the defeated.)
…
2. But, if progressives really insist on “RCV”, then the task becomes the job of educating voters in Hare to not be lesser-evil favorite-burial giveaway suckers. Good luck.
…
The progress of “RCV” enactment around the country, in many cities & 2 states, could mean that we just have to accept it as the way voting-system reform is going to happen,
…
On the other hand, 2 states isn’t diddly-squat !!! In 35 years of heavily-funded pushing & promotion, FairVote only has 2 states to show for it??? What kind of progress is that?
…
That’s why Approval seems the best prospect, in spite of “RCV” ‘s alleged success. Approval is good enough, & easy enough to define, describe, propose, enact, inexpensively implement, administer, oversee & audit against error & count-fraud… to take off a lot faster, everywhere, than “RCV” has. That’s why it’s Approval that I propose. - Offer them a better rank-method. Yes, IRVists point out that there’s much to be said for not electing a middle CW compromise who is unliked. Yes, when there’s a Mutual-Majority, Hare is guaranteed to choose the favorite candidate of the largest faction of a Mutual-Majority. But the IRVists are forgetting something when they use an argument about an unfavorite compromise:
-
@sarawolk said in The dangers of analysis paralysis in voting reform:
@lime Among people who know about them or know about voting methods already, or among people who have never heard of them and get a simple explanation and form an intuitive opinion?
It looks like there's been a bunch of research on the topic, and the consensus seems to be that people prefer slightly more categories—up to 10, with more than 10 having smaller effects. When given the option to rate candidates on a 100-point scale, voters almost-always choose multiples of 10 (with no peaks at multiples of 20); asked to explicitly state which of several scales they liked best, they usually go with 10-point scales; and in psychometrics, the validity of self-ratings usually increases up to 9-11 categories.
https://www.rangevoting.org/RateScaleResearch.htmlOn the other hand, having to fit ten stars on a ballot would be a pain, so I think a +0.5 option is a better presentation. But it looks like 5 is uncomfortably few for most people.
-
Yes, & so the obvious solution is to propose the absolute MINIMAL multicandidate voting-system.
My point was NOT that we should only promote STAR or that we should eliminate Approval for consideration. I think they are both great options that are not redundant at all.
The point was that having TOO MANY options is harmful to the adoption of any of them. How many is too many? I would (jokes aside) argue for the elimination of all that are redundant and don't add to the conversation. One traditional ballot, one ranked, and one scored system is plenty. We can eliminate any that have serious issues with vote-splitting and accuracy, that have have serious issues with voided ballots, or voter error, that have problematic strategic incentives, or that have unnecessary complexity that doesn't add much to the question.
To me that leaves us with STAR, Approval, and Ranked Robin (aka Condorcet), and the multi-winner and PR versions of those.
