What US constitutional amendments would you propose?

[Terralthra]

So what's your opinion on more elaborate cardinal voting systems? Like, say, rate each candidate on a scale from -10 to +10?

Mechanically, those are functionally identical to any other ranked-choice voting system. If they're easier to understand for the average voter, fine, go for it, but the system which reads the ballots and calculates the winning candidate is no more complicated, and operates functionally identically, minus whatever tie-breaking goes into candidates who have an equal number of the same rating. E.G.:

John Doe +2
Joe Blow +5
Darth Wong +3
My Ass -1
Douchenozzle -4
God -6
Zombie Carl Sagan +10

is functionally identical to

John Doe #4
Joe Blow #2
Darth Wong #3
My Ass #5
Douchenozzle #6
God #7
Zombie Carl Sagan #1

Only if a significant amount of voters rank multiple candidates at equally high preferences (with no winning candidate above them) does the tabulation code even change.

[Simon_Jester]

Hm. Putting an algorithm into the Constitution seems a little strange. But maybe it shouldn't be.

Any serious human activity is improved by the introduction of math.

If writing a constitution isn't serious, what is?

[Alyeska]

Universal fucking healthcare amendment.

[bilateralrope]

So what's your opinion on more elaborate cardinal voting systems? Like, say, rate each candidate on a scale from -10 to +10?

I'd expect most people to only give the maximum or minimum possible scores. First it will happen out of laziness, then other people will follow suit in an attempt to counteract it.

[Gaidin]

Hm. Putting an algorithm into the Constitution seems a little strange. But maybe it shouldn't be.

Any serious human activity is improved by the introduction of math.

If writing a constitution isn't serious, what is?

What happens when we figure out a better algorithm? I wouldn't be so wary if amending it weren't so hard. Unless the idea were to generalize it and say to use the 'better' algorithm in a defined fashion.

[Adam Reynolds]

What happens when we figure out a better algorithm? I wouldn't be so wary if amending it weren't so hard. Unless the idea were to generalize it and say to use the 'better' algorithm in a defined fashion.

The downside here would be that if it were easy to change it gives too much wiggle room and the debates about changing it would be too difficult.

Any serious human activity is improved by the introduction of math.

The largest issue with that is that the people that tend to be involved in idea like this aren't mathematicians and many dislike the idea of using mathematics in politics. Look at the resistance given to Nate Silver's predictions about the 2012 elections in which he statistically predicted which way every single state would swing. While most of that resistance was among ignorant Republicans because it was telling them something they didn't want to hear, they would likely resist this concept for the same reason. Having said that I do have to agree completely.

[blahface]

IRV's slightly better than plurality. (What's not?) But it's still got problems. I'm a bigger fan of approval voting, which is trivial to implement, easier to explain, and in simulations has vastly better results. It also has the advantage of breaking, not just the two-party system, but the entire concept of political parties as we understand them. If you can vote for two or six or twenty candidates equally, why bother having a primary?

I disagree on approval voting, strongly. It does better in simulations, as long as we assume that everyone has dichotomous preferences: Person X approves of these candidates, and does not approve of those. As soon as any significant portion of the electorate has more then "yes/no" opinions - say, wanting a Green Party candidate to win, but preferring a Democrat to a Republican, and anything is better than a Tea Party member - approval voting falls on its face. It's trivially easy to come up with scenarios in which people who have ordinal preferences, but use approval voting and elect a Condorcet loser (someone who would straight up lose a 1:1 election with another candidate), because their preferences can not be expressed. IRV (or some other system with preferences expressed beyond "yes/no") fix this problem at the cost of only a little complexity.

If we are starting over from scratch, I'd want to use the Condorcet method. The most pragmatic solution now though would be approval voting since it is easy to understand and it can be implemented with the current punch card machines. What I'd like to see is an open non-partisan primary with approval voting. The two candidates with the most approval would face off in the general election. This gives you a bit of a cushion and allows you to vote a little more honestly.

[Terralthra]

IRV's slightly better than plurality. (What's not?) But it's still got problems. I'm a bigger fan of approval voting, which is trivial to implement, easier to explain, and in simulations has vastly better results. It also has the advantage of breaking, not just the two-party system, but the entire concept of political parties as we understand them. If you can vote for two or six or twenty candidates equally, why bother having a primary?

I disagree on approval voting, strongly. It does better in simulations, as long as we assume that everyone has dichotomous preferences: Person X approves of these candidates, and does not approve of those. As soon as any significant portion of the electorate has more then "yes/no" opinions - say, wanting a Green Party candidate to win, but preferring a Democrat to a Republican, and anything is better than a Tea Party member - approval voting falls on its face. It's trivially easy to come up with scenarios in which people who have ordinal preferences, but use approval voting and elect a Condorcet loser (someone who would straight up lose a 1:1 election with another candidate), because their preferences can not be expressed. IRV (or some other system with preferences expressed beyond "yes/no") fix this problem at the cost of only a little complexity.

If we are starting over from scratch, I'd want to use the Condorcet method. The most pragmatic solution now though would be approval voting since it is easy to understand and it can be implemented with the current punch card machines. What I'd like to see is an open non-partisan primary with approval voting. The two candidates with the most approval would face off in the general election. This gives you a bit of a cushion and allows you to vote a little more honestly.

The "Condorcet method" isn't a method, it's a set of voting methods which will always elect the Condorcet winner. Hilariously, your preferred "pragmatic solution" of approval voting is not a Condorcet method.

[blahface]

The "Condorcet method" isn't a method, it's a set of voting methods which will always elect the Condorcet winner. Hilariously, your preferred "pragmatic solution" of approval voting is not a Condorcet method.

I never said it was. I said that I would prefer the Condorcet method (or a Condorcet method if you want to be picky about semantics) , but the most pragmatic solution NOW to first-past-the-post is approval.

[Terralthra]

Must I repeat myself? There is no "Condorcet method." Condorcet methods are any number of systems which satisfy the Condorcet criterion, that whoever wins in a multi-candidate election would win a 1:1 election with any candidate who does not win. Not all fields of candidates have Condorcet winners, no matter what voting method you use.

[blahface]

Must I repeat myself? There is no "Condorcet method." Condorcet methods are any number of systems which satisfy the Condorcet criterion, that whoever wins in a multi-candidate election would win a 1:1 election with any candidate who does not win. Not all fields of candidates have Condorcet winners, no matter what voting method you use.

It doesn't matter. You know what I meant. I don't care if it is the Schulze method, Ranked Pairs, or what. My point was that A Condorcet method should be used and I think that came across clearly. Nitpicking on semantics just clogs the thread.

[Gaidin]

What happens when we figure out a better algorithm? I wouldn't be so wary if amending it weren't so hard. Unless the idea were to generalize it and say to use the 'better' algorithm in a defined fashion.

The downside here would be that if it were easy to change it gives too much wiggle room and the debates about changing it would be too difficult.

If you're talking about changing the constitutuinally defined requirements, you'd be right, that should be difficult. If you're talking about merely changing the algorithm, that should be as easy as proving the proposed one is mathematically better according to the constitutionally defined requirements.

[Simon_Jester]

What happens when we figure out a better algorithm? I wouldn't be so wary if amending it weren't so hard. Unless the idea were to generalize it and say to use the 'better' algorithm in a defined fashion.

Honestly, we don't need a "better" algorithm, we just need a vaguely goodish one that can't be gamed to create distorted districts. Gerrymandering as we know it only really works because modern technology let people sit down and 'optimize' the political content of districts, one tiny chunk of land at a time. Defeat the ability to just make the district whatever shape you want, put even minor constraints on the process, and it's a lot easier to do things properly.

The largest issue with that is that the people that tend to be involved in idea like this aren't mathematicians and many dislike the idea of using mathematics in politics. Look at the resistance given to Nate Silver's predictions about the 2012 elections in which he statistically predicted which way every single state would swing. While most of that resistance was among ignorant Republicans because it was telling them something they didn't want to hear, they would likely resist this concept for the same reason. Having said that I do have to agree completely.

OK, so this one wouldn't pass.

The problem is, there's no way to exercise any control over gerrymandering without at least using basic geometry, even if you go "all districts must be a triangle or rectangle" or some such. So math is unavoidable here, and I think trying to pretend it can be avoided won't help the process.

[ZOmegaZ]

So what's your opinion on more elaborate cardinal voting systems? Like, say, rate each candidate on a scale from -10 to +10?

That's not exactly a cardinal voting system. Cardinal systems (I think by definition, 95% confidence) don't allow you to give candidates equal rank. That subjects them to Arrow's Impossibility Theorem. Score voting and approval voting don't have that problem.

[ZOmegaZ]

IRV's slightly better than plurality. (What's not?) But it's still got problems. I'm a bigger fan of approval voting, which is trivial to implement, easier to explain, and in simulations has vastly better results. It also has the advantage of breaking, not just the two-party system, but the entire concept of political parties as we understand them. If you can vote for two or six or twenty candidates equally, why bother having a primary?

I disagree on approval voting, strongly. It does better in simulations, as long as we assume that everyone has dichotomous preferences: Person X approves of these candidates, and does not approve of those. As soon as any significant portion of the electorate has more then "yes/no" opinions - say, wanting a Green Party candidate to win, but preferring a Democrat to a Republican, and anything is better than a Tea Party member - approval voting falls on its face. It's trivially easy to come up with scenarios in which people who have ordinal preferences, but use approval voting and elect a Condorcet loser (someone who would straight up lose a 1:1 election with another candidate), because their preferences can not be expressed. IRV (or some other system with preferences expressed beyond "yes/no") fix this problem at the cost of only a little complexity.

Well, sure you can come up with scenarios where approval voting doesn't select a Condorcet winner (assuming one exists). You can do the same for IRV. The real question is how often such things occur with both systems.

The simulation results I've seen show that approval voting has lower Bayesian regret than almost any other system, even if the voters pick the most tactical place to put their approve/disapprove line. Range voting (score-based) gives somewhat better results, but is far more difficult to implement. IRV is way down the list, second worst, just above plurality.

For me, the worst part is that IRV isn't even monotonic; ranking a candidate higher can actually hurt their chances of winning! That makes IRV very susceptible to tactical voting, which is the problem we're trying to fix in the first place. The only tactical aspect to approval voting is where you draw your approve/disapprove line, and the effect of that change (in terms of Bayesian regret) is relatively small.

IRV is also more expensive to count, and counting can't be decentralized.

[Formless]

Mechanically, those are functionally identical to any other ranked-choice voting system. If they're easier to understand for the average voter, fine, go for it, but the system which reads the ballots and calculates the winning candidate is no more complicated, and operates functionally identically, minus whatever tie-breaking goes into candidates who have an equal number of the same rating. E.G.:

John Doe +2
Joe Blow +5
Darth Wong +3
My Ass -1
Douchenozzle -4
God -6
Zombie Carl Sagan +10

is functionally identical to

John Doe #4
Joe Blow #2
Darth Wong #3
My Ass #5
Douchenozzle #6
God #7
Zombie Carl Sagan #1

Only if a significant amount of voters rank multiple candidates at equally high preferences (with no winning candidate above them) does the tabulation code even change.

My understanding was that Range Voting systems are scored as cardinal numbers, not ranks: it seems like the results shouldn't be as similar as you suggest. Every candidate gets an average, much like a grade in school (in fact one method I've seen uses A-B-C-D-F rather than numbers), and the one with the highest GPA wins.

Note that I don't necessarily advocate Range Voting, rather its a system I don't see often talked about. Personally, I think the most important thing is to get off of our current First Past The Post Electoral College system. Approval voting, IRV, Random Ballot, whatever. Surely any one of them is an improvement over what we have. If IRV is easier to justify legally, go for it. If Approval vote is easier to implement with current ballot methods, go for it. I have my ideal Random Ballot method, but I can settle for "better".

Incidentally, Range Voting also gives you an automatic approval rating for the President independent of opinion polling, which can be used elsewhere in the political process; at least assuming that we're starting from scratch and writing a Constitution right now devoid of the baggage of our existing document. Not all advantages of a voting system stem from its ability to represent the population fairly. For instance, plural voting (that one where "one person, one vote" is discarded as the measure of voter equality) can let you empower people with differential circumstances; you could for instance give households with children extra votes so that part of the population normally unrepresented due to their age are given power by proxy of their parents; or you could give immigrants a lesser but existent portion of votes so that they have some say in how the country they nevertheless live in is run.

Just something to think about. There is more than one way to look at these voting processes.

I'd expect most people to only give the maximum or minimum possible scores. First it will happen out of laziness, then other people will follow suit in an attempt to counteract it.

Admittedly, that is a pretty obvious tactic in range voting.

That's not exactly a cardinal voting system. Cardinal systems (I think by definition, 95% confidence) don't allow you to give candidates equal rank. That subjects them to Arrow's Impossibility Theorem. Score voting and approval voting don't have that problem.

It is, actually. By definition, the most basic cardinal system is Approval voting, which is basically a binary version of Range Voting. You either approve of someone or you don't; whereas in Range voting you can approve, strongly approve, don't care, or mark them as an asshole. See wikipedia, I left a link on the last page. It just means that the numbers are counted as cardinal numbers (i.e. values) and not ranks (i.e. preferences).

You might be thinking of cumulative voting (which can be combined with plural voting).

For me, the worst part is that IRV isn't even monotonic; ranking a candidate higher can actually hurt their chances of winning! That makes IRV very susceptible to tactical voting, which is the problem we're trying to fix in the first place. The only tactical aspect to approval voting is where you draw your approve/disapprove line, and the effect of that change (in terms of Bayesian regret) is relatively small.

Of course, you realize that Random Ballot methods are the only way of eliminating tactical voting advantage entirely, right?

[Formless]

An idea Sorchus and I had in a conversation this morning, an addition to the Bill of Rights: Protection of Private Speech added under the First Amendment. This way we not only have privacy law in the Constitution, but it is unambiguous why privacy rights are important to a free society. Obviously there would be (strict) provisions for law enforcement to get warrants to read your mail and such, but it would protect people from abuses a la the NSA.

[Highlord Laan]

Make voting compulsory for all adults, with large fines for not doing so. Also require more voting sites in every state, with expanded hours.

[Zaune]

Any serious human activity is improved by the introduction of math.

That... really depends on your definition of a serious human activity, to say the least.

[Batman]

Make voting compulsory for all adults, with large fines for not doing so. Also require more voting sites in every state, with expanded hours.

I'm not sure that's going to do much good (the compulsory+fine, more voting sites is good). Some of the people who didn't vote in the past because they thought the candidate/party of their choice had no chance anyway might show up to turn in their 'already lost' vote, but I'd expect a lot if not most of the people who don't care enough (or are sufficiently convinced their candidate/party can't win anyway) to vote pretty much at random to avoid the fine, not to mention a lot of 'spite' voting for fringe groups (assuming they're available) because people in general and apparently US americans in particular really don't like to be told they have to do something by the government, or else.

[ZOmegaZ]

I was getting my words confused. Approval is not an ordinal voting system, so Arrow doesn't apply as it does to cardinal systems.