I know it hasn't been long but my last post honestly got a lot more attention than I expected based on how active this community is with a bunch of people arguing both sides.

Honestly, love that. Love how active y'all were. I'm sorry if I offended anyone with my post or replies btw, I do tend to get a bit... passionate on a keyboard. I don't think any of y'all are dumb or anything, we're all voting enthusiasts that hate FPTP here, right?

Anyway, I sort of got lost in the weeds and I can't keep track of all the comments, I'm not used to posts of mine getting more than like 20 comments tops lol. But I did try to read and respond to as much as I could and sort of lost sleep and coherency in the process, oof.

I'd just like to use this post to address some general rebuttals to my last post that were sort of similar across various different comments as well as point out some interesting rebuttals I genuinely didn't consider before.

Tactical Voting

So, a lot of comments did focus on this again. And they were mostly of the same flavor of the fact that unless you're bullet voting, you have to inherently tactically vote in Approval Voting as it will lower the chances of your genuine favorite winning.

This is the later-no-harm criterion. The later-no-harm criterion states that adding an additional ranking or rating shouldn't lower the chances of a higher ranked/rated candidate to win.

The example used on Wiki for example is if Alice and Bob were going head to head. A subset of the voters place Alice as 2nd and Bob as 6th. Alice won. If another election happened, and only that subset of voters changed their ranks to Alice 2nd and Bob 3rd, Bob shouldn't suddenly beat Alice.

Honestly, I don't think this is a good voting criterion for many reasons. Just from distribution, not a lot of voting methods actually satisfy this criterion. All spoiler-proof cardinal methods don't, all Condorcet methods don't. The ones that do are ones that most people who know at least a little about voting already agree aren't really ideal voting systems, such as FPTP, IRV, and literally random ballot (you pick a random ballot, whoever is ranked highest on that one wins).

There are many varieties of tactical voting, obviously, and Gibbard's Theorem proves that if you have more than 2 candidates on a ballot then it requires tactical voting. Different voting methods fall prey to different kinds of tactical voting, and honestly later-no-harm is just... The one kind that I honestly believe is genuinely okay.

Like, the reason why all spoiler-proof cardinal voting methods fail this criterion is simple. If Alice just beats Bob by having a slightly higher average score (such as 5.3 vs 5.2), it doesn't matter whether you're using 0–10, sliding scale, 0–5, -1–+1, etc. If a subset of voters increase Bob's rating but it's still below Alice (from 10 vs 0 to 10 vs 9), Bob's average score will increase. And if that average is above Alice, he wins.

Like... Of course he should? He has the higher average rating? Why should one subset of voters be able to hold the election hostage? Just because they still prefer Alice more than Bob, this doesn't actually mean anything when it comes to the population's preference. Even if that subset still prefers Alice slightly, just by increasing Bob's preference then yeah sometimes the population's preference can change.

Anyway, it's harder to explain in less than a wall of text how this works for Condorcet and some other ranked voting methods. It's not like, contrived or anything. It's more of a "it's harder to explain how to play this card game than it is to just play this card game."

But the point is even Condorcet methods fail this criterion. They also inherently fail some others, but this specific one I think is a good thing. It's just not a very good criterion. Even the guy who conceptualized later-no-harm admitted it's a controversial one.

Honestly the best way I can explain later-no-harm is that it's just equivalent to tyranny of the minority. Blegh. Also every system I've seen that satisfies this criterion does not satisfy Condorcet/Smith criteria (depending on your goals this is either bad or meh), as well as reversal symmetry (this one's just bad to not fulfill yikes).

Anyway, it's not the entire list of voting criteria, but if you just look at Wikipedia's comparison table of different voting methods and what criterion they fulfill, you'll notice that Approval Voting basically fulfills everything that isn't later-no-harm or Condorcet related stuff (since pretty much any cardinal system is incompatible with Condorcet just like Condorcet is incompatible with IIA, sincere favorite, later-no-harm, etc).

And honestly, I think it's just unhelpful to get too hung up about tactical voting. Aside from spoilers and center squeeze, other forms of tactical voting hasn't been much of a problem or scandal in history. In places that have tried Approval Voting, there's no evidence that tactical voting is a problem.

There are a few times in history where people have tried to tactically vote en masse, but they often result in the strategic voters getting the opposite intended effect. On the other hand, if you look for things like decoy parties in mixed member electoral systems, there's a list. It's not like the biggest list or anything, but it's clear that electoral manipulation more often happens and is a problem from the candidate side, not the voter side.

Psychology

Honestly, I sort of... Just ignored this aspect in my last post. I shouldn't have, I'm sorry.

Okay, so I'll admit that Approval Voting does have a bit of a weakness here. Whether you're a voting enthusiast or the average voter, not many people want to grapple with "What's my threshold, what do I do with candidates I neither like nor dislike?!"

Like yeah it's easy to pick who you like and who you dislike, but what about everyone in between? How are you supposed to choose?? Because yeah. Voting for mid candidates helps them while hurting both your likes (doesn't fulfill later-no-harm) and dislikes.

This part is honestly sort of... Also purposefully baked into the method? Basically, there's a phenomenon called the law of large numbers.

Basically, imagine a jar of jelly beans. A bunch of people are asked how many jelly beans do they think is in the jar. Almost everyone will be wrong. BUT. When you average everyone's guess... It's suddenly very close to the correct answer? Every time? How's that possible?

Basically. Brains are beautiful, chaotic, mysterious things that are very good at heuristics. It's not really good at arriving at the exact right answer, but if you average out a bunch of heuristics, every one of which has certain biases, those biases tend to cancel each other out bringing you closer to the true answer than you might assume.

However, there are caveats. The law of large numbers doesn't work if people can mess with the jar. It doesn't work if they see others' guesses before giving their own guess. It doesn't work if people discuss with each other to decide what they all will guess.

This is why private voting booths are very important. You can't see other people's votes. The polls are just a jar of jelly beans that don't give you precise, concise measurements. And even if you do discuss how to vote beforehand, that group pressure disappears inside a private voting booth. You can just vote how you would have alone, and just... Lie when you get out.

So, this is why those of us who support Approval Voting keep telling you it doesn't matter how you or another voter grapples with what a threshold even means. Due to the law of large numbers, Approval Voting will still automatically approach the true proportionate approval of each candidate despite just being a series of binary choices by cancelling out inherent biases we all have in our brains.

HOWEVER. If this still seems really, really sucky to you. I mean you could just try a different spoilerproof cardinal method? There's combined Approval Voting, which is -1, 0, and +1 essentially. Basically, let us know who you like, dislike, and don't care about.

Now before people bash me about "Hey! You claimed Approval Voting is ideal! Are you going back on your word? Gotcha!" Nah dawg. Spoilerproof cardinal methods that just add up the scores all behave the same. The results are the same. They're all just Approval Voting wearing a trench coat.

That's crazy right? There must be a difference, right? In terms of final rankings, no not really. Based on some data like France 2012, the specific proportions can change due to purely psychological effects from the voters, but when determining who wins, who's second, whatever, there's not really a difference. But hey, if it makes you feel better, then I'm for it.

Voter Satisfaction Efficiency

Imma shout out u/subheight640. You probably don't like me based on my replies, but I really have no hard feelings against you. I was very sleepy, so I'm sorry if I offended you lol. Anyway, they actually taught me something new. Voter Satisfaction Efficiency. It's always fun learning something new.

Basically, it's a metric for rating voting methods based on how well they elect the candidate with the highest average utility. It's normalized so that an impossibly perfect method would have a VSE of 1, while a VSE of 0 is basically no better than random election, and a VSE lower than 0 means you're literally just better off electing randomly. You can also measure them assuming various levels and kinds of strategic voting from the voters.

BUT after a good sleep I still have issues with your simulation. As a brief overview, his results are basically: STAR and Condorcet good; IRV, score, and approval mid; plurality bad (boo FPTP).

I researched more about VSE afterwards and have looked at other people's simulations and other charts from various orgs. I already said I have problems with the way they did the simulation, but let's just look at other charts.

Here's the sim they link explaining VSE

Here's more sims from a blog

What you'll notice is that Condorcet methods actually vary in their performance based on the sims. Same with Approval. They sort of flip flop between being good or mid. Plurality is always bad (Boo FPTP). INTERESTINGLY, STAR is always high.

So let's dissect it. VSE is not a perfect metric. In fact, the variability of the scores of various voting methods based on different people's models to measure VSE shows that it's a flawed metric in the first place, perhaps purely due to the difficulty of accurately simulating voter behavior.

Even subheight admits they made various assumptions about voter behavior for score voting methods, but their claim is basically approval must be mid based on those assumptions.

Not only that, the various models themselves already assume how each voter genuinely feels about various candidates. That's fine for determining how they would honestly vote. But when modeling strategic behavior, it's based on how other voters genuinely feel about candidates.

Irl, even with polling, you can't possibly know how fellow constituents actually feel or whether they'll vote honestly or with a specific strategy during an election, so you can't possibly know at all the best way to strategically vote to give your own personal favorite candidates the best possible chances.

This touches back onto the fact that aside from obvious spoiler effects, voters generally don't even bother tactically voting, or if they do, it's not very effective. And these models also assume various levels of strategic voting but like it's kind of ridiculous to say "oh 100% of voters will use a one-sided strategy in this simulation."

That just never happens and isn't realistic or even worth simulating. Especially since, when looking at absolute average VSE's, the scores that different voting systems get under various assumptions of strategic voting don't have largely different VSE's, except like one case with Borda going all the way to the negatives for some reason.

So, is that it then? Either way, it seems STAR voting is the best if we just average out different people's VSE models? Lol God no, absolutely not.

What VSE doesn't tell you and why simplicity matters

VSE's only tell you how good a voting method is at electing the highest average utility winner. That's it. As the Medium blog states, "It doesn’t account for how difficult voting methods are to tabulate, the incentives they give to candidates, whether they promote the growth of third parties when their chosen candidates lose, etc."

This is pretty clear to see from the fact that although plurality does perform the worst often (boo FPTP), the objective VSE score isn't that low. But we know that FPTP is massively flawed, not just slightly behind other voting methods. It incentivizes two-party systems with massive center squeeze and potential for political slide. It's terrible, not just "slightly worse but really 70-80% of the way there."

In fact, let's just obverse STAR voting. Averaging out various VSE models would have you believe that STAR voting is in fact the optimal voting method. I mean it consistently scores as one of the highest, so it must be great, right?

I'm sorry to STAR voting supporters, but no. Let's look at both the theoretical and the practical. Theoretically, it fulfills the Majority/Condorcet Loser criteria. Not very important criteria as far as election criteria go.

It fails to meet Majority Winner, a very basic condition (if majority of voters rank 1 candidate first, they must win), it's affected by spoilers, it's heavily affected by clone candidates, it's not consistent (if two sets of voters prefer A over B, then the combination of the sets should prefer A over B, AKA immunity to gerrymandering. Condorcet methods also fail this, which is bad), it fails later-no-help (different from later-no-harm, this means vulnerability to mischief voting which reduces chances of sincere Condorcet winners being elected, which is actually something all Condorcet methods also fail btw), and it doesn't even fulfill the Sincere Favorite criterion, which is a pretty basic criterion for cardinal systems (you have incentive to vote/rate highest your genuine favorite candidate).

So a lot of theoretical problems with STAR voting. What about practical data? Well, there's a case where ~30% of ballots had to be tossed because voters didn't fill them in properly.

That last bit is the real kicker. As voting enthusiasts (or really any enthusiasts of procedural methods of all kinds), it's often easy to be sucked into the idea of the "best" method. And often, this enthusiasm makes simple systems seem... Unworth the enthusiasm. "There must be something it's missing. It's too simple, it can't possibly work. It must fail to consider certain things!"

When trying to find the best procedure for... Anything... You have to look to the average user.

The average voter... Does not give one flying fuck about voting systems outside of an election season. They don't understand various theoretical terms. They can't figure out the consequences of a given system if it's too complex to bother worrying about. Even in a simple system, the average voter is not knowledgeable enough to successfully implement tactical strategies.

This is why complicated systems like STAR Voting, Condorcet Methods, etc won't work. Even if I'M wrong that Approval Voting is theoretically the best, complicated systems will fail in a practical sense more often than it.

Even if we made STAR Voting permanent, you'd assume voter performance will improve over time and with better education, right? No, not really. Learning and doing is also different. At the end of the day, some ballots will still be thrown out due to improper application. And that will be skewed towards newer and younger voters.

This is inherently, absolutely undemocratic. Which is the whole reason we're trying to end FPTP, as a reminder. Complicated voting methods will always favor more experienced voters, not the general populace. We need a system that is as accessible as possible. That doesn't mean using it is easy or not frustrating or feels emotionally like it must be correct, it just means that the system's general mechanics are easy to understand at a glance. Approval Voting is easy to understand at a glance, whether y'all enthusiasts trust it or not.

On the other hand, aside from voting enthusiasts, actual politicians and political advisors are absolutely aware of how voting systems work. If they have an opportunity to manipulate elections to help themselves, you can damn be sure they'll do that.

This is why I don't like Condorcet methods. Not the contrived scenarios where tactical voting is a problem. No. They are susceptible to gerrymandering. And you can't just ban gerrymandering. The politicians would be in charge of that. Why would they ban it as long as they can benefit from it? We've all known politicians who promise electoral reform and just... Don't. Cough Trudeau cough

And this is also why I don't like mixed member electoral systems. They exist to try to fix the flaws of FPTP rather than replace it, but they inherently have to balance democracy (extra seats are inherently undemocratic as they don't represent any constituents and have more loyalty to party than people. More hanging seats, less democratic), national representation (something that is only a problem because of spoilers and center squeeze, and requires those hanging seats to achieve), and other tactical forms of manipulation like decoy parties. Fuck. That.

Final Comparison

Majority Winner: Can it elect someone that most people rank first?

Yes: Most

No: Borda, anti-plurality, quadratic, STAR (!!), random, sortition

Condorcet Winner: Will it always elect the Condorcet winner, if it exists?

Yes: Condorcet methods

No: Non-Condorcet ranked methods, cardinal methods, everything else

Pareto: If every voter prefers A to B, does A win?

Yes: Most

No: Anti-plurality and sortition

Independence of Irrelevant Alternatives: Is the system not affected by spoilers?

Yes: Approval Voting, other forms of spoilerproof cardinal methods, random ballot, sortition

No: All ranked methods including Condorcet, STAR and other spoiled cardinal methods

Cloneproof: Adding non-winning candidates similar to those already on the ballot shouldn't affect the outcome.

Yes: IRV, spoilerproof cardinal methods, some Condorcet methods, random ballots

No: Plurality, anti-plurality, some Condorcet methods, STAR (!!), quadratic, sortition

Monotone: Increasing a candidate's ranking or rating shouldn't cause them to lose.

Yes: Most

No: IRV, some Condorcet methods

Consistency: If two sets of voters prefer A over B, their combined set should prefer A over B (immunity to gerrymandering).

Yes: Spoilerproof cardinal methods, plurality (unfortunately the electoral college is vulnerable to gerrymandering, this has nothing to do with FPTP), anti-plurality, Borda, quadratic, maximal lotteries

No: Everything else

Reversal Symmetry: Reversing rankings and ratings should flip the results

Yes: Some Condorcet methods, spoilerproof cardinal methods

No: Everything else (for some it's irrelevant like plurality: you can't vote for all but one lol)

Homogeneity: Does replicating every ballot keep The results the same?

Yes: most methods, at least some Condorcet methods

No: Dodgson

Unknown: some Condorcet methods, STAR, quadratic, maximal lotteries

Later-no-harm: Adding an additional ranking/rating doesn't cause a higher ranked/rated candidate from losing

No: Most

Yes: Plurality, IRV, two-round, random ballot, sortition (ew, all gross)

Later-no-help: Adding an additional ranking/rating doesn't cause a higher ranked/rated candidate from winning (mischief voting)

Yes: Plurality, two-round, IRV, random ballot, sortition, Borda, spoilerproof cardinal methods

No: anti-plurality, STAR, Condorcet methods

Sincere Favorite: A voter is incentivized to give the highest ranking/rating to their sincere favorite candidate.

Yes: random ballot, sortition, spoilerproof cardinal methods, anti-plurality, Coombs

No: Everything else

Simplicity: Will the average voter be able to understand how it works at glance/is there a low likelihood of ballots being thrown out for being filled improperly?

Yes: plurality, anti-plurality, spoilerproof cardinal methods, random ballot

No: Everything else

VSE: STAR ≥ Condorcet ≥ Spoilerproof cardinal ≥ IRV > Plurality (boo FPTP).

ANYWAY. If you really can't stomach thinking during a vote while being able to stomach thinking about how voting systems work, a combined approval vote (-1, 0, +1) should feel better. Still approval voting in disguise though.

Point is. Fuck FPTP. A lot of systems are still subpar. STAR ends up with votes being thrown out. Condorcet is great except it's susceptible to gerrymandering (which you can't ban under Condorcet).

Approval Voting... What, feels wishy washy with defining thresholds as a feature not a bug? It's simple, and the average voter will understand how to use it, whether they bullet vote or not.

It's Pareto efficient. It's not affected by spoilers. It's not affected by clones so parties can't just fill the ballot with decoys. It's monotone, which is part of its simplicity. It's consistent, so politicians can't gerrymander it. Voting for A doesn't cause B to win, which again is part of its simplicity. And you're always incentivized to give a vote to your genuine favorite.

Additional votes are tactical, yes. The tactic is do you help someone who's mid to harm someone you don't like at the expense of also maybe harming someone you like. But this is a feature of most voting systems. And in fact the ones that don't do this... Are just generally agreed upon to be terrible even if you agree with this criterion.

And it's basically the only thing approval voting is susceptible for. Saying that makes it inherently tactical is misguided. Everything is tactical. Or it's a dictatorship or a 1v1. At least this is the one thing you need to look out for, instead of juggling a bunch of shit with other systems. The latter of which the professional politicians will definitely be better at than the average voter, leading to potential for electoral manipulation.

Because again. Historically, voters purposefully trying to vote strategically en masse haven't exactly had the results work out for them. But politicians globally certainly have manipulated elections before to their benefit, even if it's not a yearly occurrence. Peace I need more sleep actually.