Alternate Voting Systems (Range, Condorcet etc)

For the serious discussion of weighty matters and worldly issues. No off-topic posts allowed.

Moderators: Azrael, Moderators General, Prelates

User avatar
BrainMagMo
Posts: 185
Joined: Tue Jul 22, 2008 6:22 am UTC
Location: Southern California
Contact:

Re: Range voting

Postby BrainMagMo » Fri Mar 13, 2009 5:59 am UTC

ManaUser wrote:
guenther wrote:Why do you think this? In terms of understanding, I think just the opposite. For ranged voting (assuming something more complicated than approval), just make sure you have a number in range next to each candidate (blank would mean zero). Ranked voting has more restrictions: rank the candidates, no duplicate numbers, etc. Ranked voting seems harder by definition. Perhaps you mean in terms of explaining the concept rather than the actual process, but I don't see why range would be any harder there. In fact, my perceived complexity of ranked voting is what drew me to range voting.

Yes, that's more what I meant. I think the concept would be alien and confusing to alot of people. I can just imagine thousands of little old ladies (sorry to stereotype but...) going "What? I have to rate them now? How do I know what to put?" Whereas the concept of ranking seems fairly easy to grasp.

guenther wrote:In terms of "gaming", I'd like to understand how it is done. I tried to follow the link to the paper on ruining approval elections, but all I got was the abstract. I'm fully open to the possibility that range voting has some serious flaws if people aren't voting honestly, but I'd like to see examples so I can understand it.

I didn't read the paper so I can't comment on that, but what I'm talking about is simple enough. Let me use a real example of what I mean. Bush/Gore/Nader (I'll just ignore the other minor parties).

Republicans would approve Bush only.
Democrats would approve Gore, and possibly Nader.
Greens would approve Nader, and probably Gore even if they only see him as the lesser of two evils.

So far so good, the spoiler effect is eliminated. But if Nader had had a realistic chance of winning this would change. Democrats would no longer give him an "approve" (even if they really did approve), fearing they would be helping him beat their preferred candidate.
you're supposed to vote for your preferred candidate.
Likewise, Greens would be disinclined to approve Gore. So it looks to me like there's still a barrier to minor parties ever becoming "major".
How is plurality or IRV better at eliminating this problem? At least under range/approval, you can vote for both.
guenther wrote:Actually if everyone either votes "high" or "low" it will just reduce to approval voting, not plurality.

Right. It's the other part (explained above) that makes it "almost the same" as plurality voting. I admit it's still somewhat better, though.

guenther wrote:In my head I imagine people will vote their favorite candidate (e.g. the Ronpaul) at max and perhaps a palpable but more popular candidate (e.g. John McCain) at max-1. And then vote everyone else a 0.

Okay, then. Why would they do that?
because it reflects how they feel.

guenther wrote:This is what sold me off of IRV. As I mentioned, there's more ways to spoil a ballot.

Well this is a whole 'nother can of worms, but it occurs to me that ranked voting would be a good match for a touch-screen "ballot".
It requires less rules to make range than ranked.
Just make it a set of mini-elections:
1. Kerry5, Kerry4, Kerry3, Kerry2 Or Kerry1?
2. Bush5, ...

Or removed the over-vote rule (that voting for >1 = rejected ballot)
How would you change the programming to do ranked?

Another possible solution would be to change the rules slightly so ranking two candidates the same is not incorrect. I haven't fully thought this through, but perhaps there could be kind of a hybred IRV/Approval system. Suppose I voted like so:

1. Ron Paul (Get out of my keyboard, Randall.)
2. Barack Obama
2. Ralph Nader
3. John McCain

That would mean that as long as Paul is in the running, I approve him. If he's eliminated, I approve Obama and Nader. If one of them is eliminated, I approve the other. If they're both eliminated approve McCain.
Why not just use range, and not make it u:ber complicated like that?

Double posts are bad. Quoting back the entirely of a long post and is convenient available in it's full form here in the thread is bad. Fix it.

-Az

Carnildo
Posts: 2023
Joined: Fri Jul 18, 2008 8:43 am UTC

Re: Range voting

Postby Carnildo » Fri Mar 13, 2009 6:10 am UTC

BrainMagMo wrote:"Approval voting

Steven Brams and Dudley R. Herschbach argued in a paper in Science magazine in 2001 that approval voting was the system least amenable to tactical perturbations. Authors of a 2004 paper in Public Choice disagreed, saying, "AV is one of the most susceptible systems to manipulation by small groups of voters (for example, small, maverick groups could determine the AV outcome.)"[2]"

Add a quorum requirement, such that at least 50% vote/rank someone, and RV gets the least tactical.


I don't have access to the paper cited in [2], but from the abstract, it sounds like their point is that they expect most approval-voting elections to be close, in which case a small number of people changing their vote can select the outcome -- but that's true of any election system in a close election. The point of "resistance to tactical voting" is that voting against your true wishes cannot result in a desirable outcome.

Soralin
Posts: 1347
Joined: Wed May 07, 2008 12:06 am UTC

Re: Range voting

Postby Soralin » Fri Mar 13, 2009 9:00 am UTC


Another possible solution would be to change the rules slightly so ranking two candidates the same is not incorrect. I haven't fully thought this through, but perhaps there could be kind of a hybred IRV/Approval system. Suppose I voted like so:

1. Ron Paul (Get out of my keyboard, Randall.)
2. Barack Obama
2. Ralph Nader
3. John McCain

That would mean that as long as Paul is in the running, I approve him. If he's eliminated, I approve Obama and Nader. If one of them is eliminated, I approve the other. If they're both eliminated approve McCain.
Why not just use range, and not make it u:ber complicated like that?

How is that complicated, it seems very simple to me, and the closest match to what people actually think about candidates. The above would mean, I'd rather have Ron Paul than Barack Obama, and I'd rather have Barack Obama than Ralph Nader, and I'd rather have Ralph Nader than John McCain.

The ranking system seems to match exactly how I'd think about candidates, I'd rather have A than B, and I'd rather have B than C. How do you express that in range voting? I could put A at 10, and C at 0, but where do I put B? The higher I rank B, the more chance there is that B will win over A, which would be bad, and the lower I rank B, the more chance there is that C would win over B, which is also bad. Which means any choice that I make will have to be strategically, rather then being what I actually think. And the whole range system has to be strategic in the fist place rather then what they actually think about the candidate. Because if you say put a 5 for one candidate, and a 0 for the rest of them, because that's how you would actually rank them on a 0-10 scale if asked separately, and not as part of a vote, then that means you only get half a vote. Two people ranking one candidate as a 5 and the rest as a 0, is the same as one person ranking a candidate as a 10 and the rest as 0. So even with just 2 candidates, you already have strategic voting, because people can't be put down how they would actually rank them individually without throwing out half their vote.

I think one the the highest criteria for a voting system should be the inability to vote strategically. Voting strategically tends to break people up into groups, and foster an us vs. them mentality. It can lead people into voting as a group, rather then voting individually, because voting as a group provides more power, even if what it votes for isn't exactly what they want. And that in turn leads to people not thinking about the effects of what they're actually voting for, and instead using a herd mentality. And overall I think produces worse effects on society then would be expected by just looking at the result of the vote itself and ignoring the psychology produced.

And when you think about it, all strategic voting is attempting to produce a more preferable result then would be produced by voting for what you actually think. And so therefore, any system that doesn't allow for strategic voting means that what people actually want, and what they vote for, will match as close as possible.

Philwelch
Posts: 2904
Joined: Tue Feb 19, 2008 5:33 am UTC
Location: RIGHT BEHIND YOU

Re: Range voting

Postby Philwelch » Fri Mar 13, 2009 6:33 pm UTC

BrainMagMo wrote:you're supposed to vote for your preferred candidate.


Yes, of course you're supposed to. But a flaw in many voting systems is that doing something different ("tactical voting") would end up being more effective, which means the voting system no longer accurately reflects the will of the people.
Fascism: If you're not with us you're against us.
Leftism: If you're not part of the solution you're part of the problem.

Perfection is an unattainable goal.

electopia
Posts: 1
Joined: Sat Mar 14, 2009 8:41 am UTC

score voting objectively superior

Postby electopia » Sat Mar 14, 2009 8:46 am UTC

So ... this used to be a quintuple post that made the same point, referencing several different people, over and over again. Now it's only one post that makes the same point, referencing several different people, over and over again. And what do we do with Giant Walls of Redundant Text?

Spoiler:
There is one "right measure" of voting method performance. It is Bayesian regret. It measures how well a voting method does precisely what it is supposed to do - make a satisfying choice. The rational voter seeks to maximize his utility. Score voting is superior to the other common methods (e.g. IRV) by a large margin.

http://scorevoting.net/UniqBest.html

IRV is one of the worst methods actually, almost as bad as plurality. This is true with any mixture of honest or strategic voters (in fact the improvement of score voting over the alternatives actually increases the more strategic voters you have), so the argument about score voting not being good if people are strategic is baloney. Here's more on that specific subject.

http://scorevoting.net/StratHonMix.html

Also a head-to-head comparison of score voting and IRV that is devastating to IRV is http://scorevoting.net/CFERlet.html

In a scientific sense, this argument is as open-and-shut as the creationism "debate".

Soralin wrote:The ranking system seems to match exactly how I'd think about candidates, I'd rather have A than B, and I'd rather have B than C. How do you express that in range voting?


On the contrary, rankings express LESS information than ratings, because you say nothing about intensity of preference, just order. And IRV in particular ignores exponentially 100% of your ballot information. => http://scorevoting.net/IgnoreExec.html

I could put A at 10, and C at 0, but where do I put B? The higher I rank B, the more chance there is that B will win over A, which would be bad, and the lower I rank B, the more chance there is that C would win over B, which is also bad. Which means any choice that I make will have to be strategically, rather then being what I actually think.


You don't have to vote strategically, it's your choice. The same is true of ALL deterministic voting methods. IRV is so bad, for instance, that it behaves about as well with 100% honest voters as score voting does with 100% strategic voters (hence it's preposterous to favor IRV because you think strategic voting is terrible). It is an objective fact (backed up by Bayesian regret calculations and some good theorems based on plausible voter strategy) that score voting actually reacts mildly to strategic voting compared to most other methods. Here's a good link about that issue: http://scorevoting.net/Honesty.html

And the whole range system has to be strategic in the fist place rather then what they actually think about the candidate. Because if you say put a 5 for one candidate, and a 0 for the rest of them, because that's how you would actually rank them on a 0-10 scale if asked separately, and not as part of a vote, then that means you only get half a vote. Two people ranking one candidate as a 5 and the rest as a 0, is the same as one person ranking a candidate as a 10 and the rest as 0. So even with just 2 candidates, you already have strategic voting, because people can't be put down how they would actually rank them individually without throwing out half their vote.


The idea of a sincere score voting ballot is that you normalize your sincere utilities. So if there are two candidates, you are expected to score one of them a 10 and the other a 0. But you are free to do whatever you want. If you want to be more expressive as a voter, then score voting is better. If you just care about your candidate-result utility, then you can vote strategically and score voting will be better. Score voting is better for both strategic and honest voters.

I think one the the highest criteria for a voting system should be the inability to vote strategically.


Well, that's wrong. The one and ONLY criterion for a voting method (other than practical ease of implementation itself) is that it has the best utility efficiency. The point of a choice is to maximize utility (which is why we evolved choosing machines, called "brains"). Score voting will give voters a better expected value with election results. To want something else is irrational. More to the point, the only known voting methods that are immune to strategic voting are HORRIBLE.

1) Pick a voter at random and let his plurality voting ballot be a dictator ballot.
2) Voters rank the candidates. Two candidates are drawn at random and pitted head-to-head based on those rankings.

These methods make honesty the optimum strategy - yet they both SUCK. So if you support these methods because strategy is so revolting to you, then you are just being irrational, and asking for society to be super unhappy with its election results, and hence for more poverty and war and disease than you could have gotten by using a non-idiotic voting method, selected based on actual science instead of your gut intuition that strategic voting is "bad".

Voting strategically tends to break people up into groups, and foster an us vs. them mentality.


On what expertise do you base this vague blanket statement? On what grounds do you think that this is even approximately as important as picking good winners? Well, it's a moot point, since all deterministic voting methods are susceptible to strategic voting.

And when you think about it, all strategic voting is attempting to produce a more preferable result then would be produced by voting for what you actually think. And so therefore, any system that doesn't allow for strategic voting means that what people actually want, and what they vote for, will match as close as possible.


Wrong. The problem is that the benefit a voter gets by voting strategically is much smaller than the collective detriment he feels from all the OTHER voters being strategic (the moral of the prisoners' dilemma). Hence just about EVERY voting method produces less satisfaction the more strategic the voters behave.

http://scorevoting.net/StratHonMix.html

xev wrote:I'd like something that allows for shades of gray in the results (very necessary!), but I am slightly confused about the benefits of Range Voting over proportional representation


Proportional representation is only achievable in multi-winner elections. Score voting (aka range voting) is a single-winner method for seats like mayor, governor, president, etc. There are two multi-winner methods that are related to score voting. Asset voting and reweighted score voting.

http://scorevoting.net/Asset.html
http://scorevoting.net/RRVj.html

Enjoy!

Iv wrote:The perfect method is known since the eighteenth century. Never implemented that I am aware of.


Condorcet methods are certainly not "perfect". They are worse than score voting (more Bayesian regret). Most of them are extremely susceptible to strategic voting. => http://scorevoting.net/DH3.html

scikidus wrote:I'm a fan of IRV (rank voting), but I'm also interested by subconscious systems of voting/polling. For instance, if murals of the two candidates were drawn on a sidewalk, and a hidden camera was installed. One might be able to judge public opinion of the candidates based on how people react to the murals (walk over them, walk around them, etc.)


Well, IRV is horrible. Score voting is vastly superior. http://scorevoting.net/CFERlet.html


That.

-Az

guenther
Posts: 1840
Joined: Sat May 17, 2008 6:15 am UTC

Re: Range voting

Postby guenther » Sun Mar 15, 2009 10:30 pm UTC

Thanks everyone for the responses! I hadn't seen the score voting site before, so thanks for the link. I was going to defend range voting more, but that site seems to take care of it. If anyone's argument is not covered there, please post why you think so.

Here's some further thoughts:

1. For any system where you can vote for multiple candidates, it seems there must be a way to limit the number of candidates. We can't reasonably expect a voter to understand 30 different people's platforms. Are there any proposed ways to limit this? Like a giant primary where only the top six candidates go to the final round? (I don't like the idea of party primaries because perhaps two Republicans are both great candidates but only one could make it out of a party primary.)

2. Assuming there's a limit to the number of candidates (see above), here's a proposal that might make range voting seem easier. Say there's six candidates, then the scoring range is 1 to 6. So the most obvious thing to do is rank them by preference (with most preferable scoring 6). But it would be explained that you can have two people scoring a 6 if you want, or two people with a 1, or any other combination.

With this idea, you could mentally rank them (which I think is what our brain is best at), and then you could move them around if you feel two candidates are equally good or bad. It's a purely range voting system, but it has a veneer of ranked voting that people might prefer.
A gentle answer turns away wrath, but a harsh word stirs up anger.

Carnildo
Posts: 2023
Joined: Fri Jul 18, 2008 8:43 am UTC

Re: Range voting

Postby Carnildo » Sun Mar 15, 2009 11:20 pm UTC

guenther wrote:1. For any system where you can vote for multiple candidates, it seems there must be a way to limit the number of candidates. We can't reasonably expect a voter to understand 30 different people's platforms. Are there any proposed ways to limit this? Like a giant primary where only the top six candidates go to the final round? (I don't like the idea of party primaries because perhaps two Republicans are both great candidates but only one could make it out of a party primary.)

Washington State has a top-two primary. It occasionally results in two Republicans or two Democrats running for the same seat, but more often, it simply keeps third parties from running. That's what the general effect of a partisan top-N primary will be: to reduce the visibility of minor parties.

User avatar
ManaUser
Posts: 284
Joined: Mon Jun 09, 2008 9:28 pm UTC

Re: Range voting

Postby ManaUser » Mon Mar 16, 2009 8:14 pm UTC

Soralin wrote:

Another possible solution would be to change the rules slightly so ranking two candidates the same is not incorrect. I haven't fully thought this through, but perhaps there could be kind of a hybred IRV/Approval system. Suppose I voted like so:

1. Ron Paul (Get out of my keyboard, Randall.)
2. Barack Obama
2. Ralph Nader
3. John McCain

That would mean that as long as Paul is in the running, I approve him. If he's eliminated, I approve Obama and Nader. If one of them is eliminated, I approve the other. If they're both eliminated approve McCain.
Why not just use range, and not make it u:ber complicated like that?

How is that complicated, it seems very simple to me, and the closest match to what people actually think about candidates. The above would mean, I'd rather have Ron Paul than Barack Obama, and I'd rather have Barack Obama than Ralph Nader, and I'd rather have Ralph Nader than John McCain.

Actually that's not what I said. Notice that those are numbered 1, 2, 2 & 3, meaning I have no preference between Obama and Nader. It was an example of my proposed hybrid IRV /Approval method. That was mainly to address the point that a ranked ballot was easier to spoil than an approval ballot. I admit it was a little complicated, but not nearly as much as some voting systems, for instance the Condorcet methods. And in any case it's not the kind of complexity that the individual voter needs to worry about. The whole process should to be open of course, but if a voter isn't interested in the details, all they need to know is that they should rank candidates in order of preference, and if they wish they can rank some of them equally.

guenther
Posts: 1840
Joined: Sat May 17, 2008 6:15 am UTC

Re: Range voting

Postby guenther » Mon Mar 16, 2009 8:47 pm UTC

ManaUser wrote:The whole process should to be open of course, but if a voter isn't interested in the details, all they need to know is that they should rank candidates in order of preference, and if they wish they can rank some of them equally.

You're hybrid system sounds similar to my suggestion above: if you have N candidates, have range scoring of 1 to N. You can rank them if you want, and you can rank some equally. The only difference I see is that you have "1" be most preferential, and I chose "N" for most preferential. Do you see any other differences?
Carnildo wrote:Washington State has a top-two primary. It occasionally results in two Republicans or two Democrats running for the same seat, but more often, it simply keeps third parties from running. That's what the general effect of a partisan top-N primary will be: to reduce the visibility of minor parties.

This is why I'm not a fan of the party primary system. But one giant primary might not be a great solution either since it might require voters to follow too many candidates. So I was hoping someone had already thought of an alternative. It seems like any voting system that opens the door to third parties will have this problem.

We have an edit button for a reason. -Az
A gentle answer turns away wrath, but a harsh word stirs up anger.

User avatar
ManaUser
Posts: 284
Joined: Mon Jun 09, 2008 9:28 pm UTC

Re: Range voting

Postby ManaUser » Mon Mar 16, 2009 10:18 pm UTC

guenther wrote:
ManaUser wrote:The whole process should to be open of course, but if a voter isn't interested in the details, all they need to know is that they should rank candidates in order of preference, and if they wish they can rank some of them equally.

You're hybrid system sounds similar to my suggestion above: if you have N candidates, have range scoring of 1 to N. You can rank them if you want, and you can rank some equally. The only difference I see is that you have "1" be most preferential, and I chose "N" for most preferential. Do you see any other differences?

The mechanics of voting would be the same, but the winner selection method would be different. From the voter's perspective, the difference is that if they thought about it, they would quickly realize that putting candidates in their actual order of preference is probably not the best use of their vote under your method.

guenther
Posts: 1840
Joined: Sat May 17, 2008 6:15 am UTC

Re: Range voting

Postby guenther » Wed Mar 18, 2009 6:08 am UTC

ManaUser wrote:The mechanics of voting would be the same, but the winner selection method would be different. From the voter's perspective, the difference is that if they thought about it, they would quickly realize that putting candidates in their actual order of preference is probably not the best use of their vote under your method.

That's a good point that I didn't consider. :) Assuming 3 candidates, if you score them in order 113 (A=1, B=1, C=3), it's the same as 112. But in my system, those are different and could affect the outcome. I still like range voting though. :)
A gentle answer turns away wrath, but a harsh word stirs up anger.

User avatar
BrainMagMo
Posts: 185
Joined: Tue Jul 22, 2008 6:22 am UTC
Location: Southern California
Contact:

Re: Range voting

Postby BrainMagMo » Sat Mar 21, 2009 7:43 am UTC

Carnildo wrote:
BrainMagMo wrote:"Approval voting

Steven Brams and Dudley R. Herschbach argued in a paper in Science magazine in 2001 that approval voting was the system least amenable to tactical perturbations. Authors of a 2004 paper in Public Choice disagreed, saying, "AV is one of the most susceptible systems to manipulation by small groups of voters (for example, small, maverick groups could determine the AV outcome.)"[2]"

Add a quorum requirement, such that at least 50% vote/rank someone, and RV gets the least tactical.


I don't have access to the paper cited in [2], but from the abstract, it sounds like their point is that they expect most approval-voting elections to be close, in which case a small number of people changing their vote can select the outcome -- but that's true of any election system in a close election. The point of "resistance to tactical voting" is that voting against your true wishes cannot result in a desirable outcome.


I was mainly concerned with the first part of the quote, I just happen to add a reply to the second part since I neither wanted to quote-snipe out the opposition to the idea that AV is the least tactical voting system nor not address the opposition I had quoted.
Unfortunately, Wiki has failed to provide a source for the first part, I may go looking for the article soon.




Soralin wrote:

Another possible solution would be to change the rules slightly so ranking two candidates the same is not incorrect. I haven't fully thought this through, but perhaps there could be kind of a hybred IRV/Approval system. Suppose I voted like so:

1. Ron Paul (Get out of my keyboard, Randall.)
2. Barack Obama
2. Ralph Nader
3. John McCain

That would mean that as long as Paul is in the running, I approve him. If he's eliminated, I approve Obama and Nader. If one of them is eliminated, I approve the other. If they're both eliminated approve McCain.
Why not just use range, and not make it u:ber complicated like that?

How is that complicated, it seems very simple to me, and the closest match to what people actually think about candidates. The above would mean, I'd rather have Ron Paul than Barack Obama, and I'd rather have Barack Obama than Ralph Nader, and I'd rather have Ralph Nader than John McCain.


The ranking system seems to match exactly how I'd think about candidates, I'd rather have A than B, and I'd rather have B than C. How do you express that in range voting? I could put A at 10, and C at 0, but where do I put B? The higher I rank B, the more chance there is that B will win over A, which would be bad, and the lower I rank B, the more chance there is that C would win over B, which is also bad. [/quote]Why would it be bad if it's how you feel?
You'd rather give up control over the relative ability for them to win, and assign them each a pre-set difference in channce to win?
Under RV, you could express by how much you prefer them. You can express that "I prefer b to c, but not by much"
However, all these "If"s make it complicated.
RV doesn't have many "If"s:
1. Rate each candidate.
2. The candidate with the greatest ratio wins.
3. No candidate may win if they are not ranked by 50% of the people (rangevoting.org uses a slightly different quorum rule)

Which means any choice that I make will have to be strategically, rather then being what I actually think.
I think you have it backwards...
And the whole range system has to be strategic in the fist place rather then what they actually think about the candidate. Because if you say put a 5 for one candidate, and a 0 for the rest of them, because that's how you would actually rank them on a 0-10 scale if asked separately, and not as part of a vote, then that means you only get half a vote. Two people ranking one candidate as a 5 and the rest as a 0, is the same as one person ranking a candidate as a 10 and the rest as 0. So even with just 2 candidates, you already have strategic voting, because people can't be put down how they would actually rank them individually without throwing out half their vote.
No. You vote is weighed equally no matter what you vote. Voting like that means, "They all suck but this guy ain't so bad" And the ten, "this guy's really great!"
Also, I'm not sure how this is strategic.
Let's try it:
John: Bush = 5, Kerry = 0, Nader = 0
Jane: Bush = 5, Kerry = 0, Nader = 0
Joan: Bush = 0, Kerry = 10, Nader = 0
Results: Bush:3.3, Kerry:3.3, Nader:0
...
How is this strategic voting?
I don't get it.
It accurately represents how the people feel.

To prove that strategic voting would occur, you have to give two elections. One in which people vote honestly, one where they don't and show that the second would cause how the people want to be how the election turns out.
Not a single election in which some people dislike all the candidates but don't dislike the one favored by a third guy enough to think expressing so would be relevant.

I think one the the highest criteria for a voting system should be the inability to vote strategically. Voting strategically tends to break people up into groups, and foster an us vs. them mentality. It can lead people into voting as a group, rather then voting individually, because voting as a group provides more power, even if what it votes for isn't exactly what they want. And that in turn leads to people not thinking about the effects of what they're actually voting for, and instead using a herd mentality. And overall I think produces worse effects on society then would be expected by just looking at the result of the vote itself and ignoring the psychology produced.

And when you think about it, all strategic voting is attempting to produce a more preferable result then would be produced by voting for what you actually think. And so therefore, any system that doesn't allow for strategic voting means that what people actually want, and what they vote for, will match as close as possible.


Good thing RV has experimentally been found to be the least strategic and most likely to foster 3rd parties:
http://rangevoting.org/NurserySumm.html
Last edited by BrainMagMo on Sat Mar 21, 2009 8:02 am UTC, edited 2 times in total.

Silas
Posts: 1091
Joined: Sat Feb 02, 2008 9:08 pm UTC

Re: Range voting

Postby Silas » Sat Mar 21, 2009 2:15 pm UTC

BrainMagMo wrote:
Which means any choice that I make will have to be strategically, rather then being what I actually think.
I think you have it backwards...
And the whole range system has to be strategic in the fist place rather then what they actually think about the candidate. Because if you say put a 5 for one candidate, and a 0 for the rest of them, because that's how you would actually rank them on a 0-10 scale if asked separately, and not as part of a vote, then that means you only get half a vote. Two people ranking one candidate as a 5 and the rest as a 0, is the same as one person ranking a candidate as a 10 and the rest as 0. So even with just 2 candidates, you already have strategic voting, because people can't be put down how they would actually rank them individually without throwing out half their vote.
No. You vote is weighed equally no matter what you vote. Voting like that means, "They all suck but this guy ain't so bad" And the ten, "this guy's really great!"
Also, I'm not sure how this is strategic.
Let's try it:
John: Bush = 5, Kerry = 0, Nader = 0
Jane: Bush = 5, Kerry = 0, Nader = 0
Joan: Bush = 0, Kerry = 10, Nader = 0
Results: Bush:10, Kerry:10, Nader:0
...
How is this strategic voting?
I don't get it.
It accurately represents how the people feel.

That's not strategic voting. Strategic voting is when John and Jane say to themselves, "I prefer Bush to Kerry and Nader both, so why am I only giving him half the advantage I can?" So what really happens is:
John: Bush = 10, Kerry = 0, Nader = 0
Jane: Bush = 10, Kerry = 0, Nader = 0
Joan: Bush = 0, Kerry = 10, Nader = 0
Results: Bush:20, Kerry:10, Nader:0

The optimal strategy for range voting, when there are two clear front-runners, is well-known. Rank your favorite front-runner, and everyone you like better than him, at 10; rank the other front-runner, and everyone you hate more, at zero; the candidates in between, you space out proportionally to your actual preferences (there may be complications, if you care how non-winning candidates are ranked).

The point is that if you vote how you actually feel, you won't apply as much political leverage as you could if you fiddled with your ballot.
Felstaff wrote:Serves you goddamned right. I hope you're happy, Cake Ruiner

User avatar
ManaUser
Posts: 284
Joined: Mon Jun 09, 2008 9:28 pm UTC

Re: Range voting

Postby ManaUser » Sun Mar 22, 2009 3:53 am UTC

BrainMagMo wrote:I don't get it.
It accurately represents how the people feel.

You're right. You don't get it.

Here's the problem. Elections are not opinion polls. The goal isn't to let anybody know how you feel, it's to put somebody you consider suitable in office. And here's the important part, you are competing with the other voters. By giving your most preferred candidate a 5 instead of a 10 you make your vote half as strong compared to someone who gives a 10 to another candidate. And as a result, the guy you liked better (or hated less) loses. Why would you do that?

BrainMagMo wrote:Good thing RV has experimentally been found to be the least strategic and most likely to foster 3rd parties:
http://rangevoting.org/NurserySumm.html

I have several problems with that analysis. Most importantly, ranked voting of any kind was not part of the experiment. They compared it to results from of IRV in Australia, specifically the top vote from each ballot but that's comparing apples and oranges.

It measured "third party" votes, not winners. They admit in their analysis that once minor parties are "out of the nursery" (i.e. have a real chance of winning) people would no longer vote honestly for them.

It wasn't a real election, just an exit poll, so "voters" had no reason to be dishonest.

guenther
Posts: 1840
Joined: Sat May 17, 2008 6:15 am UTC

Re: Range voting

Postby guenther » Sun Mar 22, 2009 4:44 am UTC

In the end, I have to say I like range voting still. I like it's simplicity in explanation, tallying, and technical implementation. In terms of quality versus complexity, it's seems like a great pick.

However, if there were a groundswell for some form of ranked voting, I'd hop on board to move us away from plurality. I still feel plurality fosters so much of the polarization in our country.

Thanks for the great discussion and information!
A gentle answer turns away wrath, but a harsh word stirs up anger.

User avatar
KrazyerKate
Posts: 431
Joined: Sat Oct 17, 2009 2:04 pm UTC

Different Style of Voting...Multiple Votes

Postby KrazyerKate » Fri Dec 10, 2010 12:41 pm UTC

I live in the US, and I hear that some european countries don't have this problem, but the States only have two parties to vote for, with the occasional third (that one never wins though). Say you're allowed to choose your two favorite candidates instead of one. Would this encourage more diverse parties? Would this help alleviate the problem of smaller parties "stealing" votes from larger parties?

User avatar
Silknor
Posts: 842
Joined: Fri Jun 20, 2008 1:21 am UTC

Re: Different Style of Voting...Multiple Votes

Postby Silknor » Fri Dec 10, 2010 1:13 pm UTC

This sounds likeapproval voting, so I'll just go with that, since there is no difference with 3 parties, and only minimal ones with more from allowing approval of as many as you like to approval of just two.

While you are correct that it would boost the share of third parties, the effect of this is superficial. It's safe to vote for all parties you like as long as only two parties have a shot at winning and the rest are tiny, as long as you like one big party and dislike the other (eg. everyone votes either for the Democrat or the Republican, and then for whatever small third parties they like). But as soon as a third party becomes a legitimate contender, strategic voting comes back into the picture, and you have your vote "stealing" problem in a new form.

So it has no real effect as long as third parties can't win (except to give third parties a large, but meaningless, share of the vote), and breaks down from its intent once third parties can win.
Nikc wrote:Silknor is the JJ Abrams of mafia modding

Роберт
Posts: 4285
Joined: Wed May 14, 2008 1:56 am UTC

Re: Different Style of Voting...Multiple Votes

Postby Роберт » Fri Dec 10, 2010 3:59 pm UTC

when the threads merged, it obsoleted this post

Condorcet > range
Last edited by Роберт on Mon Dec 13, 2010 6:23 pm UTC, edited 1 time in total.
The Great Hippo wrote:[T]he way we treat suspected terrorists genuinely terrifies me.

broken lader
Posts: 63
Joined: Mon Dec 13, 2010 4:11 am UTC

Re: Different Style of Voting...Multiple Votes

Postby broken lader » Mon Dec 13, 2010 4:30 am UTC

@silknor

Approval Voting has an enormous pro-third-party impact, not a superficial one.

For instance, say the Orange Party starts up with a moderate platform which encompasses a mixture of conservative and liberal ideas. They aren't your typical far-from-center parties like the Greens or Libertarians or what have you.

Say they are the favorite of 25% of the voters, but those voters tactically decide to support their favorite between the Democrat and the Republican, for the most part, in addition to casting a sincere vote for Orange. So then Orange gets roughly half of the votes of the two major parties.

But that's still pretty good, way beyond a paltry 3% or so like third parties usually get. So they aren't seen as unelectable nut jobs, but instead as a sane and viable alternative to the two major parties. Their notoriety grows, and say in some future election, voter preferences look more like this.

24% R>O>D
26% O>R>D
26% O>D>R
24% D>O>R

E.g. the first row prefers Republican over Orange over Democrat.

Now since the R's and D's have historically been successful, a sane voter will generally vote for his favorite between those two, plus any candidates he likes even more. So this would translate into approvals like this:

24% R
26% O R
26% O D
24% D

Now Orange wins, despite not being seen as electable, and despite strategic voting based on that appearance of unelectability.

And also consider Duverger's Law, which observes that most countries that use Plurality Voting, followed by a runoff, have escaped two-party duopoly. If even Plurality Voting can produce a multi-party democracy, when merely combined with a runoff process, then it appears Approval Voting can certainly do it.

To say that Approval Voting breaks down so that third parties can't win, is just inconsistent with the evidence.

@Роберт

Condorcet voting is not the ideal actually. Score Voting and Approval Voting have lower Bayesian regret (i.e. they are objectively "better").
ScoreVoting.net/BayRegsFig.html

And when you consider the implications of naive voter exaggeration strategy, Approval Voting actually may be a better Condorcet method than real Condorcet methods.
ScoreVoting.net/AppCW.html

Also, Condorcet can severely fail when voters are tactical.
ScoreVoting.net/DH3.html

Clay Shentrup
Secretary, Director - The Center for Election Science
San Francisco, CA

User avatar
Azrael
CATS. CATS ARE NICE.
Posts: 6491
Joined: Thu Apr 26, 2007 1:16 am UTC
Location: Boston

Re: Range voting

Postby Azrael » Mon Dec 13, 2010 5:44 pm UTC

Newest thread merged and re-titled.

Роберт
Posts: 4285
Joined: Wed May 14, 2008 1:56 am UTC

Re: Different Style of Voting...Multiple Votes

Postby Роберт » Mon Dec 13, 2010 6:43 pm UTC

broken lader wrote:@Роберт

Condorcet voting is not the ideal actually. Score Voting and Approval Voting have lower Bayesian regret (i.e. they are objectively "better").
ScoreVoting.net/BayRegsFig.html

And when you consider the implications of naive voter exaggeration strategy, Approval Voting actually may be a better Condorcet method than real Condorcet methods.
ScoreVoting.net/AppCW.html

Also, Condorcet can severely fail when voters are tactical.
ScoreVoting.net/DH3.html


Clay Shentrup
Secretary, Director - The Center for Election Science
San Francisco, CA

Scorevoting.net, aka rangevoting.org, hardly seems like an unbiased source. The Bayesian regret is actually very similar vs range voting with honest votes. However, in Range Voting, tactical, non-honest votes are a good idea. In Condorcet voting, there is no reason to not give an honest vote. Therefore, looking at the graph, it looks like the realistic Bayesian regret is better for Condorcet.

The DH3 scenario is completely bogus crap. There is absolutely no benefit from the "exaggeration" that they are suggesting voters would do in condorcet... and there's no reason that retarded voters couldn't cause the exact same problem with range voting.
Now, what happens? The A-supporters say to themselves: "We are in trouble. Polls suggest A is going to lose if we just vote A>B>C>D as is our honest opinion. But if we exaggeratedly vote A>D>B>C downgrading A's main rivals as far as we can, then maybe A will have a chance." The B-supporters say "those rotten A-supporters for sure are going to exaggerate and effectively get twice the A-versus-B discriminating power as if they were honest. We cannot sit still and just take that. We have to fight back by also exaggerating: B>D>C>A." And similarly the C-supporters say "we will not just sit back and be robbed of our deserved victory by those dishonest exaggerating scum. We will also exaggerate: C>D>A>B." (And by the way, they are completely right. C would definitely lose to A or B if they just sat there.)

Incidentally, some purists may quibble: why did the A-fan voters decide to exaggerate? Well the C-fans felt forced to do so because, given that the A- and B-fans already chose to exaggerate, the C-voters knew that C could not win without exaggeration. But all three kinds of voters do not know what the others are going to do and how many of them are going to do it (nor even how many of them there are), and hence have to guess, and their guess is "most of those rotters are probably going to exaggerate"! So based on this guess, they feel they too must exaggerate to get any chance of victory. (And that feeling is always accurate in the sense that, if some appropriate fraction of the opposing voters exaggerated, then [1] our candidate would be sure to lose to a rival, but [2] by such exaggeration we could regain the victory.)
Incorrect. C supporters do not increase the chances of C winning by moving D farther towards the front. The chances of a candidate winning are not affected by changes in ordering that do not cross the candidate. Also, look:
Now, what happens? The A-supporters say to themselves: "We are in trouble. Polls suggest A is going to lose if we just vote A=10 B=6 C=3 D=0 as is our honest opinion. But if we exaggeratedly vote A=10 D=6 B=3 C=0 downgrading A's main rivals as far as we can, then maybe A will have a chance."
And, of course, here the exaggeration actually could be useful in that it would directly increase A's chances of beating B.

EDITS FOLLOW
Clearly they are thinking of Borda voting, not Condorcet
Note: they say the exaggeration of range voting in this case would be A=10, everyone else =0, fair enough, people could exaggerate that way if they wanted and avoid DH3, but similarly for Condorcet, you could just say A > everyone else. The difference is with range voting, if you want to put your relative preference for the other guys, you weaken your A vote, but for Condorcet, you're free to order the others without affecting the strength of your A vote.
The Great Hippo wrote:[T]he way we treat suspected terrorists genuinely terrifies me.

broken lader
Posts: 63
Joined: Mon Dec 13, 2010 4:11 am UTC

Re: Alternate Voting Systems (Range, Condorcet etc)

Postby broken lader » Mon Dec 13, 2010 9:11 pm UTC

This definitely applies to Condorcet as well as Borda. Here are concrete examples of Condorcet elections where the burial strategy discused on that DH3 page works:
ScoreVoting.net/CondStratProb.html

Here are the results of some computer simulations that show how often the DH3 tactic works, with various Condorcet methods.
ScoreVoting.net/CondBurial.html

If you don't trust the site because you think it's "biased", you can verify the computations yourself.

Now in reality, most voters have no idea how their ranked voting systems work. For instance, we use Instant Runoff Voting here in San Francisco, but even my fellow software engineers here at my company can't describe its relatively simple tabulation mechanism to me off the top of their heads. We actually see a great deal of empirical evidence however (e.g. from decades of use of IRV in Australia), that users of ranked voting methods use what we call "naive exaggeration", in which they essentially polarize the frontrunners, with the goal of having a stronger effect on the election. And to an extent, that strategy actually is pretty effective with a variety of voting methods. But that doesn't matter - what matters is that they DO it. We have to choose voting methods based on how they'll be have in reality, not in theory.

You point out that voters could technically do the same thing with Score Voting, but there is no real analog in Score Voting, since it satisfies Independence of Irrelevant Alternatives. For instance, say I sincerely believe that X=10, Y=9, and Z=0 -- but X and Y are the frontrunners. I might indeed push Y down to a 0. But it does me no tactical good to bring Z higher, the way it does with Condorcet methods. And naive voter intuition also doesn't seem to cause them to do that either, as it does with ranked methods.

Роберт
Posts: 4285
Joined: Wed May 14, 2008 1:56 am UTC

Re: Alternate Voting Systems (Range, Condorcet etc)

Postby Роберт » Mon Dec 13, 2010 9:57 pm UTC

Hmmm... interesting.

So basically, with Condorcet voting, it's possible for people to lie about their preferences in a way that makes their be no Condorcet winner. When you get in that situation (where there is no condorcet winner), it's hard to decide who should win.

I was unaware of this possibility... I must ponder voting some more.

EDIT: what if you did Condorcet voting but also gave values to the candidates and, if there is no Condorcet winner, fell back on range voting to eliminate the lowest-scoring candidate, and then re-checked for a Condercet winner, repeating as necessary?
The Great Hippo wrote:[T]he way we treat suspected terrorists genuinely terrifies me.

broken lader
Posts: 63
Joined: Mon Dec 13, 2010 4:11 am UTC

Re: Alternate Voting Systems (Range, Condorcet etc)

Postby broken lader » Tue Dec 14, 2010 3:35 am UTC

So basically, with Condorcet voting, it's possible for people to lie about their preferences in a way that makes their be no Condorcet winner.


And vice versa.

When you get in that situation (where there is no condorcet winner), it's hard to decide who should win.


It's always hard to decide who should win, since you can't read the voters' minds. It is mathematically proven that the Condorcet winner, even when one exists, is not necessarily the best candidate, no matter how you define "best".
See www.electology.org/majority-criterion

what if you did Condorcet voting but also gave values to the candidates and, if there is no Condorcet winner, fell back on range voting to eliminate the lowest-scoring candidate, and then re-checked for a Condercet winner, repeating as necessary?


Welcome to my world back in 2006:
groups.yahoo.com/group/RangeVoting/message/1207

The thing is, yes you can try all kinds of variants like this, aimed at making strategy so computationally difficult, that it is not feasible to do it effectively. But...

1) Most voters just use naive exaggeration anyway, so they'll still exaggerate even when it hurts them. We have tons of evidence for this. E.g.
ScoreVoting.net/AusAboveTheLine07.html

2) Such a Frankenstein voting method will be so complex as to increase costs, and decrease transparency. Even if it's for some little internal election for an organization, you'd have to code the tabulation algorithm. And you'd have to explain the process to the voters, many of whom might reject it for being so hard to understand. Whereas Score Voting and especially Approval Voting are just super simple and familiar, from sites like IMDB.com and Youtube.

3) Even if you want to do all that work, your Condorcet variant may have worse representativeness (i.e. higher Bayesian regret), even when voters are sincere. You'd have to perform some pretty complex computer modeling to get a good estimation of that. Is it really worth it? Maybe from an academic perspective, but if you look at that Bayesian regret graph, SV and AV are already so good it seems like diminishing returns to me.

Роберт
Posts: 4285
Joined: Wed May 14, 2008 1:56 am UTC

Re: Alternate Voting Systems (Range, Condorcet etc)

Postby Роберт » Tue Dec 14, 2010 5:12 pm UTC

broken lader wrote:
When you get in that situation (where there is no condorcet winner), it's hard to decide who should win.


It's always hard to decide who should win, since you can't read the voters' minds. It is mathematically proven that the Condorcet winner, even when one exists, is not necessarily the best candidate, no matter how you define "best".
See www.electology.org/majority-criterion

No, that's not true. Clearly, if there is a Condorcet winner, they should be elected.

To put it simply, the simplest election is between two people. The Condorcet winner is a candidate who would win the election if he was pitched head-to-head against every other candidate. It makes no sense to say someone other than the Condorcet winner should be elected.

The majority criterion link's example was showing the case where there is no Condorcet winner.

Anyway, I see no good reason to do range voting over approval voting... so if we're going to do voting reform, we could switch to approval voting.
The Great Hippo wrote:[T]he way we treat suspected terrorists genuinely terrifies me.

distractedSofty
Posts: 253
Joined: Tue Jul 06, 2010 5:29 pm UTC

Re: Alternate Voting Systems (Range, Condorcet etc)

Postby distractedSofty » Tue Dec 14, 2010 6:15 pm UTC

Роберт wrote:Anyway, I see no good reason to do range voting over approval voting... so if we're going to do voting reform, we could switch to approval voting.

And a very good reason not to do it: It's horrible UI.

Plurality voting meshes well with the question "which of these candidates would you prefer to have?", and everyone has the concept of favourites. Approval voting meshes with the question "which of these would you be ok with having?", a question I answer all the time with my wife and friends when deciding what to eat. Preferential voting systems (at least those that allow partial preferences) mesh with that question even better: "I want Chinese, and if not Chinese then sushi, if not sushi then Thai, and if not one of those three then I don't really care".

What question is range voting trying to get my answer to?

Роберт
Posts: 4285
Joined: Wed May 14, 2008 1:56 am UTC

Re: Alternate Voting Systems (Range, Condorcet etc)

Postby Роберт » Tue Dec 14, 2010 6:23 pm UTC

distractedSofty wrote:What question is range voting trying to get my answer to?

How good is this [amazon product, youtube video, presidential candidate]?
Click on the number of stars!

It doesn't have to have a horrible UI.
The Great Hippo wrote:[T]he way we treat suspected terrorists genuinely terrifies me.

distractedSofty
Posts: 253
Joined: Tue Jul 06, 2010 5:29 pm UTC

Re: Alternate Voting Systems (Range, Condorcet etc)

Postby distractedSofty » Tue Dec 14, 2010 6:54 pm UTC

Роберт wrote:
distractedSofty wrote:What question is range voting trying to get my answer to?

How good is this [amazon product, youtube video, presidential candidate]?
Click on the number of stars!

It doesn't have to have a horrible UI.

Those are a horible UI as well (in that they interface with the user (their brain, at least) badly). Before surveys and ratings started using numerical scales, I don't think that anyone used them. Even now: "Hey Bob, how do you like your new TV?" "7".

If you read Amazon reviews, you'll find many different interpretations of what the scale means, and if you look at the little histogram at the top, you'll find that most ratings are 5 or 1. Many people seem to use it as an enumeration of faults in the product (I had to knock a star of for it being boring, and a star for the cover being hot pink. So 3 stars)

Роберт
Posts: 4285
Joined: Wed May 14, 2008 1:56 am UTC

Re: Alternate Voting Systems (Range, Condorcet etc)

Postby Роберт » Tue Dec 14, 2010 7:09 pm UTC

distractedSofty wrote:Those are a horible UI as well (in that they interface with the user (their brain, at least) badly). Before surveys and ratings started using numerical scales, I don't think that anyone used them. Even now: "Hey Bob, how do you like your new TV?" "7".

People use numerical scales often... it's not great for casual conversation, though, since you have to have some definition of what it means.
The Great Hippo wrote:[T]he way we treat suspected terrorists genuinely terrifies me.

distractedSofty
Posts: 253
Joined: Tue Jul 06, 2010 5:29 pm UTC

Re: Alternate Voting Systems (Range, Condorcet etc)

Postby distractedSofty » Tue Dec 14, 2010 7:24 pm UTC

Роберт wrote:People use numerical scales often... it's not great for casual conversation, though, since you have to have some definition of what it means.

My point was that it's not how people think. Do you have any examples of people spontaneously using star ratings? (Do you comparison shop by giving each option a numerical score?)

Роберт
Posts: 4285
Joined: Wed May 14, 2008 1:56 am UTC

Re: Alternate Voting Systems (Range, Condorcet etc)

Postby Роберт » Tue Dec 14, 2010 8:15 pm UTC

People do ask "on a scale of 1-10, how do you feel about your marriage?" type questions is conversations.
The Great Hippo wrote:[T]he way we treat suspected terrorists genuinely terrifies me.

distractedSofty
Posts: 253
Joined: Tue Jul 06, 2010 5:29 pm UTC

Re: Alternate Voting Systems (Range, Condorcet etc)

Postby distractedSofty » Tue Dec 14, 2010 9:40 pm UTC

Роберт wrote:People do ask "on a scale of 1-10, how do you feel about your marriage?" type questions is conversations.

That's a good example, but in my experience, the answer is always "100" or "0". (Much like I was saying about Amazon ratings)

The principal advantage of range voting appears do be that it is more expressive: it allows the expression of preferences like "I like these two candidates the same amount", "I prefer X to Y, but not by much", and "I like A exactly twice as much as I like B". Has there ever actually been any psychological studies that show that people have these kinds of preferences? (Obviously not the first one: that's a given)

EDIT: In addition, I realise that it only allows you to express such preferences with some preparation. It would be quite hard to hold all those kinds of thoughts in your head and then put down the correct scores when you were in the voting booth. Range voting is the Linux of voting systems.

broken lader
Posts: 63
Joined: Mon Dec 13, 2010 4:11 am UTC

Re: Alternate Voting Systems (Range, Condorcet etc)

Postby broken lader » Wed Dec 15, 2010 3:49 am UTC

Роберт wrote:
broken lader wrote:It is mathematically proven that the Condorcet winner, even when one exists, is not necessarily the best candidate, no matter how you define "best".
See http://www.electology.org/majority-criterion

No, that's not true. Clearly, if there is a Condorcet winner, they should be elected.


Let me repeat, it's mathematically proven that the Condorcet winner is not necessarily the "best" or "most representative" candidate. Please see that link.

Роберт wrote:To put it simply, the simplest election is between two people. The Condorcet winner is a candidate who would win the election if he was pitched head-to-head against every other candidate. It makes no sense to say someone other than the Condorcet winner should be elected.


That is precisely what I thought too. It just seemed so logical and straightforward. And then I found out that there was a mathematical proof that I was wrong.

It is mathematically proven that the social utility function is just the sum of the voters' utilities, and "Condorcet winner" has zero bearing on the matter of "who is the best candidate".
See scorevoting.net/UtilFoundns.html

Роберт wrote:The majority criterion link's example was showing the case where there is no Condorcet winner.


It shows several cases. The first one has no Condorcet winner. The subsequent ones all have a Condorcet winner, and we prove that in at least one of them, the "best candidate" is not the Condorcet winner. That disproves the axiom that "the Condorcet winner is necessarily the right winner".

Роберт wrote:Anyway, I see no good reason to do range voting over approval voting... so if we're going to do voting reform, we could switch to approval voting.


There is a really good reason actually.

For the 10% or so of voters who prefer expressing themselves over maximizing their expected value, using the intermediate scores of Score Voting is preferable to being constrained to two options.

For the 90% or so who are tactical, they get a better expected value as a result of those expressive voters voluntarily giving up some power in exchange for expressiveness.

So essentially 100% of voters are better off with Score Voting. Approval Voting is a nice intermediate step between Plurality Voting and Score Voting, that despite not being as good as Score Voting, comes close and is incredibly simple.

broken lader
Posts: 63
Joined: Mon Dec 13, 2010 4:11 am UTC

Re: Alternate Voting Systems (Range, Condorcet etc)

Postby broken lader » Wed Dec 15, 2010 4:00 am UTC

distractedSofty wrote:The principal advantage of range voting appears do be that it is more expressive: it allows the expression of preferences like "I like these two candidates the same amount", "I prefer X to Y, but not by much", and "I like A exactly twice as much as I like B". Has there ever actually been any psychological studies that show that people have these kinds of preferences?


You don't need a psychological study to show that. It's a quite trivial aspect of economics, as seen in expected value comparisons.

For instance, say you have the option of getting a guarantee of 1 dollar, vs. a 50/50 lottery of getting a million dollars or zero dollars. Most any person would take the 50/50 guarantee. But say you have a guarantee of getting 900k, vs. the same 50/50 zero-million lottery. Then it's more likely you'd take the 900k guarantee. It depends on how rich you are, since there's a logarithmic relationship between money and utility.

Point being, there are lots of revealed preference exercises like this that one can use to demonstrate the universally cardinal nature of human preference.

distractedSofty wrote:In addition, I realise that it only allows you to express such preferences with some preparation. It would be quite hard to hold all those kinds of thoughts in your head and then put down the correct scores when you were in the voting booth. Range voting is the Linux of voting systems.


Back in 2006 I did a small Score Voting exit poll in Beaumont, Texas. People seemed to have absolutely no problem quickly marking down scores.
ScoreVoting.net/Beaumont.html

I would argue that Condorcet and/or IRV are the Linuxen of voting systems, as Score Voting is much more human-friendly and simple, both in terms of Kolmogorov complexity and empirical ballot spoilage rates. And even things like precinct summability, which adds transparency and fraud resistance.

Interestingly enough, Fedora Linux actually uses Score Voting for their steering committee stuff and release names.

++$_
Mo' Money
Posts: 2370
Joined: Thu Nov 01, 2007 4:06 am UTC

Re: Alternate Voting Systems (Range, Condorcet etc)

Postby ++$_ » Wed Dec 15, 2010 4:43 am UTC

broken lader wrote:
distractedSofty wrote:The principal advantage of range voting appears do be that it is more expressive: it allows the expression of preferences like "I like these two candidates the same amount", "I prefer X to Y, but not by much", and "I like A exactly twice as much as I like B". Has there ever actually been any psychological studies that show that people have these kinds of preferences?
You don't need a psychological study to show that. It's a quite trivial aspect of economics, as seen in expected value comparisons.
Economics is worthless unless it actually describes human behavior. You need psychological studies to find out whether it does or not.
For instance, say you have the option of getting a guarantee of 1 dollar, vs. a 50/50 lottery of getting a million dollars or zero dollars. Most any person would take the 50/50 guarantee. But say you have a guarantee of getting 900k, vs. the same 50/50 zero-million lottery. Then it's more likely you'd take the 900k guarantee. It depends on how rich you are, since there's a logarithmic relationship between money and utility.
This experiment has nothing to do with presidential candidates. At best, such an experiment shows that preferences are cardinal in the domain of monetary lotteries. There is no reason (outside of economic dogma) to believe that this is also true when it comes to presidential candidates.

(Also, there is no reason to believe that the utility of money is logarithmic, outside of doing an experiment, and in many situations it is obviously NOT logarithmic (because people do play the lottery).)
Point being, there are lots of revealed preference exercises like this that one can use to demonstrate the universally cardinal nature of human preference.
Yes -- those are psychological studies. And they have emphatically not shown that human preference is cardinal. In fact, there is reason to believe that at least in some cases, human preference may not even be transitive.

distractedSofty
Posts: 253
Joined: Tue Jul 06, 2010 5:29 pm UTC

Re: Alternate Voting Systems (Range, Condorcet etc)

Postby distractedSofty » Wed Dec 15, 2010 8:31 am UTC

broken lader wrote:For instance, say you have the option of getting a guarantee of 1 dollar, vs. a 50/50 lottery of getting a million dollars or zero dollars. Most any person would take the 50/50 guarantee. But say you have a guarantee of getting 900k, vs. the same 50/50 zero-million lottery. Then it's more likely you'd take the 900k guarantee. It depends on how rich you are, since there's a logarithmic relationship between money and utility.

I'm not quite sure what your trying to illustrate with this example: most people would take a smaller guarantee (say 200k) too. How does it demonstrate these quantitative preferences? You'll still choose one or the other.
broken lader wrote:Back in 2006 I did a small Score Voting exit poll in Beaumont, Texas. People seemed to have absolutely no problem quickly marking down scores.
ScoreVoting.net/Beaumont.html

That's an interesting datapoint. Although, I have 3 problems with it as evidence for the "It's easy to fill out a range ballot" conjecture: 1) there were only 5 candidates in the poll, 2) your scores were only 0-10 (Which does not give it the kinds of expressiveness that I was mentioning above), and 3) it was an exit poll, so the penalty for not expressing your views correctly was low.
broken lader wrote:I would argue that Condorcet and/or IRV are the Linuxen of voting systems, as Score Voting is much more human-friendly and simple, both in terms of Kolmogorov complexity and empirical ballot spoilage rates. And even things like precinct summability, which adds transparency and fraud resistance.

I assume when you say it has better Kolmogorov complexity you're referring to the voting method, and not to the ballot (which, as any range voting advocate will gladly tell you, contains more information). This is only true in the absolute sense; if we're discussing "human friendliness", we also need to know how humans maintain their preferences internally. If it's already an ordered list, then ordered lists are the best interface for a voting system.

I'm also curious as to the evidence you have for range voting having better empirical ballot spoilage rates. Range voting has never been put in place for any major elections that I am aware of, so the general public's proficiency with scoring candidates has yet to be measured.

Роберт
Posts: 4285
Joined: Wed May 14, 2008 1:56 am UTC

Re: Alternate Voting Systems (Range, Condorcet etc)

Postby Роберт » Wed Dec 15, 2010 3:33 pm UTC

broken lader wrote:
Роберт wrote:
broken lader wrote:It is mathematically proven that the Condorcet winner, even when one exists, is not necessarily the best candidate, no matter how you define "best".
See http://www.electology.org/majority-criterion

No, that's not true. Clearly, if there is a Condorcet winner, they should be elected.


Let me repeat, it's mathematically proven that the Condorcet winner is not necessarily the "best" or "most representative" candidate. Please see that link.

Роберт wrote:To put it simply, the simplest election is between two people. The Condorcet winner is a candidate who would win the election if he was pitched head-to-head against every other candidate. It makes no sense to say someone other than the Condorcet winner should be elected.


That is precisely what I thought too. It just seemed so logical and straightforward. And then I found out that there was a mathematical proof that I was wrong.

It is mathematically proven that the social utility function is just the sum of the voters' utilities, and "Condorcet winner" has zero bearing on the matter of "who is the best candidate".
See scorevoting.net/UtilFoundns.html

Okay, well I guess it depends on what you think is important/makes someone the best candidate. Good luck convincing a democratic country that the winner of an election should be someone that the majority of the people would vote against.
The Great Hippo wrote:[T]he way we treat suspected terrorists genuinely terrifies me.

User avatar
Zamfir
I built a novelty castle, the irony was lost on some.
Posts: 7604
Joined: Wed Aug 27, 2008 2:43 pm UTC
Location: Nederland

Re: Alternate Voting Systems (Range, Condorcet etc)

Postby Zamfir » Wed Dec 15, 2010 3:52 pm UTC

broken lader wrote:It is mathematically proven that the social utility function is just the sum of the voters' utilities, and "Condorcet winner" has zero bearing on the matter of "who is the best candidate".
See scorevoting.net/UtilFoundns.html

How is that mathematically "proven"? The main assumption there is the veil of ignorance:
J.C.Harsanyi, in a 2-page article involving no mathematics whatever [J.Political Economy 61,5 (1953) 434-435], came up with the following nice idea: "Optimizing social welfare" means "picking the state of the world all individuals would prefer if they were in a state of uncertainty about their identity."

Funny enough, Rawls more famous treatment of that same idea assumes that people would use a maximin rule, not a maximization of the average. That is, people would choose to maximize the lowest utility, not to maximize their expected utility. The veil itself only gives you Pareto optima, picking one among them requires extra assumptions.

But more than that, it is far from clear what it means that people "are in a state of uncertainty about their identity". Suppose "I" woke up tomorrow as "you", with all your memories and personality and your place in life and none of my current ones, and you woke up as me. Has the world then changed in any sense? How would we know? How can we know this isn't already happening all the time?

Suppose we were in a state of uncertainty about this. Let's say there was 50% chance that the switch happens tonight and 50% it doesn't. And let's say I was given a choice: if I pay $100 to a fund, you an I both get $70 tomorrow. Should I do it?

After all, in both cases the person who is in my place tomorrow will remember paying $100 in a fund and getting $70 back.

LtNOWIS
Posts: 371
Joined: Sun Dec 12, 2010 4:21 pm UTC
Location: Fairfax County

Re: Alternate Voting Systems (Range, Condorcet etc)

Postby LtNOWIS » Sat Mar 12, 2011 3:56 am UTC

Some quick updates on IRV. Apparently, they're looking to adopt in Hawaii for county and special congressional elections. They want to avoid a situation like in mid-2010, where two Democrats split the ticket and a Republican won with less than 40% of the vote.

However, they adopted IRV in San Francisco, and apparently it's getting less popular, because people don't understand how it works. Personally, I think a little confusion is probably better than electing someone with 40% or less of the vote. And it might be better than the two-round systems they have in the South, where lots of pointless runoff elections. In the South Carolina GOP primary this year, the first round results were 49-22-17-12. So the runoff was a foregone conclusion, and a waste of everyone's time.

Overall, I wouldn't call myself an IRV advocate though. All three systems have their pluses and minuses. It should be up to states and localities to decide which system to use.

User avatar
webby
Posts: 139
Joined: Sat Jul 03, 2010 11:02 am UTC
Location: Sydney, Australia
Contact:

Re: Alternate Voting Systems (Range, Condorcet etc)

Postby webby » Sat Mar 12, 2011 4:44 am UTC

There's a referendum in May in the UK about adopting the alternative vote (also called preferential voting in Australia and instant run-off voting in the US). Anyone British have an idea of what the likely result will be?

It's the system we have in Australia (actually for federal elections it's compulsory to put preferences for all the candidates - the proposed change in the UK just gives the option to preference additional candidates). Antony Green has a good series of posts about the system in the context of the British referendum. http://blogs.abc.net.au/antonygreen/uk- ... eferendum/


Return to “Serious Business”

Who is online

Users browsing this forum: No registered users and 15 guests