election-methods@mailman.electorama.com

Technical discussion of election methods

View all threads

Cooperative voting

AJ
Andy Jennings
Fri, Oct 16, 2015 6:28 PM

Cooperative voting

Voters indicate one favorite candidate and separately indicate all other
candidates they approve of.

Approvals are an agreement to "conditionally cooperate".  If A is my
favorite but I also approved B, the approval for B will not count unless
there is a corresponding B-voter who approved A.

Counting

  1. Candidates start with their plurality totals.

  2. Each pair of candidates (A,B) is examined.  Take the smaller of the
    number of A-voters who approved B and the number of B-voters who approved A
    and add it to both A's and B's totals.

Motivation

Having run the straw poll for the Repulican Liberty Caucus last weekend,
I've been thinking about approval voting, the chicken (Burr) dilemma, and
the prisoners' dilemma.

One solution to a prisoners' dilemma situation is to offer "conditional
cooperation".

Alternatives

  1. Use percentages instead.  In step 2, let p equal the smaller of the
    fraction of A-voters who approved B and the fraction of B-voters who
    approved A.  p times the number of A-voters is added to B's total.  p times
    the number of B-voters is added to A's total.

  2. Use non-approval instead.  In step 2, let x equal the larger of the
    number of A-voters who did NOT approve B and the number of B-voters who did
    NOT approve A.  The number of A-voters minus x (if positive) is added to
    B's total.  The number of B-voters minus x (if positive) is added to A's
    total.

Questions

Has this system been proposed before by another name?

How can it be abused?

Cooperative voting ------------------ Voters indicate one favorite candidate and separately indicate all other candidates they approve of. Approvals are an agreement to "conditionally cooperate". If A is my favorite but I also approved B, the approval for B will not count unless there is a corresponding B-voter who approved A. Counting -------- 1. Candidates start with their plurality totals. 2. Each pair of candidates (A,B) is examined. Take the smaller of the number of A-voters who approved B and the number of B-voters who approved A and add it to both A's and B's totals. Motivation ---------- Having run the straw poll for the Repulican Liberty Caucus last weekend, I've been thinking about approval voting, the chicken (Burr) dilemma, and the prisoners' dilemma. One solution to a prisoners' dilemma situation is to offer "conditional cooperation". Alternatives ------------ 1) Use percentages instead. In step 2, let p equal the smaller of the fraction of A-voters who approved B and the fraction of B-voters who approved A. p times the number of A-voters is added to B's total. p times the number of B-voters is added to A's total. 2) Use non-approval instead. In step 2, let x equal the larger of the number of A-voters who did NOT approve B and the number of B-voters who did NOT approve A. The number of A-voters minus x (if positive) is added to B's total. The number of B-voters minus x (if positive) is added to A's total. Questions --------- Has this system been proposed before by another name? How can it be abused?
KV
Kevin Venzke
Fri, Oct 16, 2015 10:52 PM

Hi Andy,
If I take this example, looking at only 51 out of 100 voters (imagine the other 49 vote for other candidates entirely):10 A>B41 B>A
Under the original rule, all A voters will approve B, and 10 B voters will approve A. B probably wins. That seems fair. The B voters get an advantage from having more plurality votes.
Under alternative 1, the smaller fraction is 100%. That would cause the two candidates to be tied at 51. (Or nearly tied, with a stray approval somewhere making the difference.) I think alternative 2 also gives this outcome. This seems a lot worse to me, because there's not really a sense that approvals constitute "lower" preferences.

From the perspective of a sincere voter, the "approval" option signifies conditioning their opinion of a candidate on how that candidate's supporters vote. It's not such a crazy idea if one doesn't know much about a given candidate. I wonder if more can be done with the concept.

My main concern with this type of method is the situation that a bloc of voters (supporting candidate A let's say) "defects" because they don't actually like candidate B. Even if B supporters know this is the situation, and even if B supporters would be willing to support A anyway, they don't have a (sincere) way of doing this.
Kevin

  De : Andy Jennings <elections@jenningsstory.com>

À : Election Methods election-methods@electorama.com
Envoyé le : Vendredi 16 octobre 2015 13h28
Objet : [EM] Cooperative voting

Cooperative voting

Voters indicate one favorite candidate and separately indicate all other candidates they approve of.

Approvals are an agreement to "conditionally cooperate".  If A is my favorite but I also approved B, the approval for B will not count unless there is a corresponding B-voter who approved A.

Counting

  1. Candidates start with their plurality totals.

  2. Each pair of candidates (A,B) is examined.  Take the smaller of the number of A-voters who approved B and the number of B-voters who approved A and add it to both A's and B's totals.

Motivation

Having run the straw poll for the Repulican Liberty Caucus last weekend, I've been thinking about approval voting, the chicken (Burr) dilemma, and the prisoners' dilemma.

One solution to a prisoners' dilemma situation is to offer "conditional cooperation".

Alternatives

  1. Use percentages instead.  In step 2, let p equal the smaller of the fraction of A-voters who approved B and the fraction of B-voters who approved A.  p times the number of A-voters is added to B's total.  p times the number of B-voters is added to A's total.

  2. Use non-approval instead.  In step 2, let x equal the larger of the number of A-voters who did NOT approve B and the number of B-voters who did NOT approve A.  The number of A-voters minus x (if positive) is added to B's total.  The number of B-voters minus x (if positive) is added to A's total.

Questions

Has this system been proposed before by another name?

How can it be abused?


Election-Methods mailing list - see http://electorama.com/em for list info

Hi Andy, If I take this example, looking at only 51 out of 100 voters (imagine the other 49 vote for other candidates entirely):10 A>B41 B>A Under the original rule, all A voters will approve B, and 10 B voters will approve A. B probably wins. That seems fair. The B voters get an advantage from having more plurality votes. Under alternative 1, the smaller fraction is 100%. That would cause the two candidates to be tied at 51. (Or nearly tied, with a stray approval somewhere making the difference.) I think alternative 2 also gives this outcome. This seems a lot worse to me, because there's not really a sense that approvals constitute "lower" preferences. >From the perspective of a sincere voter, the "approval" option signifies conditioning their opinion of a candidate on how that candidate's supporters vote. It's not such a crazy idea if one doesn't know much about a given candidate. I wonder if more can be done with the concept. My main concern with this type of method is the situation that a bloc of voters (supporting candidate A let's say) "defects" because they don't actually like candidate B. Even if B supporters know this is the situation, and even if B supporters would be willing to support A anyway, they don't have a (sincere) way of doing this. Kevin De : Andy Jennings <elections@jenningsstory.com> À : Election Methods <election-methods@electorama.com> Envoyé le : Vendredi 16 octobre 2015 13h28 Objet : [EM] Cooperative voting Cooperative voting ------------------ Voters indicate one favorite candidate and separately indicate all other candidates they approve of. Approvals are an agreement to "conditionally cooperate".  If A is my favorite but I also approved B, the approval for B will not count unless there is a corresponding B-voter who approved A. Counting -------- 1. Candidates start with their plurality totals. 2. Each pair of candidates (A,B) is examined.  Take the smaller of the number of A-voters who approved B and the number of B-voters who approved A and add it to both A's and B's totals. Motivation ---------- Having run the straw poll for the Repulican Liberty Caucus last weekend, I've been thinking about approval voting, the chicken (Burr) dilemma, and the prisoners' dilemma. One solution to a prisoners' dilemma situation is to offer "conditional cooperation". Alternatives ------------ 1) Use percentages instead.  In step 2, let p equal the smaller of the fraction of A-voters who approved B and the fraction of B-voters who approved A.  p times the number of A-voters is added to B's total.  p times the number of B-voters is added to A's total. 2) Use non-approval instead.  In step 2, let x equal the larger of the number of A-voters who did NOT approve B and the number of B-voters who did NOT approve A.  The number of A-voters minus x (if positive) is added to B's total.  The number of B-voters minus x (if positive) is added to A's total. Questions --------- Has this system been proposed before by another name? How can it be abused? ---- Election-Methods mailing list - see http://electorama.com/em for list info
AJ
Andy Jennings
Wed, Oct 28, 2015 12:21 AM

On Fri, Oct 16, 2015 at 3:52 PM, Kevin Venzke stepjak@yahoo.fr wrote:

From the perspective of a sincere voter, the "approval" option signifies
conditioning their opinion of a candidate on how that candidate's
supporters vote. It's not such a crazy idea if one doesn't know much about
a given candidate. I wonder if more can be done with the concept.

My main concern with this type of method is the situation that a bloc of
voters (supporting candidate A let's say) "defects" because they don't
actually like candidate B. Even if B supporters know this is the situation,
and even if B supporters would be willing to support A anyway, they don't
have a (sincere) way of doing this.

You could have people mark their favorite in one column, mark everyone they
unconditionally approve in a second column, and mark everyone they
conditionally approve in a third column.  It seems like that creates more
incentives for "chicken", where A's supporters say, "we're not going to
approve B at all, so you'd better unconditionally approve A or someone else
will win."  I'm trying to decide if that's worse than the original system.
In the original system they could say, "we're not going to conditionally
approve B, so you'd better just put A as your favorite or someone else will
win."

We could combine the first two columns, so they just marked everyone they
unconditionally approve in one column and mark everyone they conditionally
approve in another column.  Then the ballot is similar to other "3-grade"
systems like MCA.  What does conditional approval mean, in that case?  Is
my approval activated when someone is found who approves ANY of my
favorites?  When someone approves ALL of my favorites?  Or it is
fractionally activated depending on what fraction of my favorites they
approve?  (I'm not sure if any of these play nicely with reciprocity.  I'll
have to think about it...)

On Fri, Oct 16, 2015 at 3:52 PM, Kevin Venzke <stepjak@yahoo.fr> wrote: > From the perspective of a sincere voter, the "approval" option signifies > conditioning their opinion of a candidate on how that candidate's > supporters vote. It's not such a crazy idea if one doesn't know much about > a given candidate. I wonder if more can be done with the concept. > > My main concern with this type of method is the situation that a bloc of > voters (supporting candidate A let's say) "defects" because they don't > actually like candidate B. Even if B supporters know this is the situation, > and even if B supporters would be willing to support A anyway, they don't > have a (sincere) way of doing this. > You could have people mark their favorite in one column, mark everyone they unconditionally approve in a second column, and mark everyone they conditionally approve in a third column. It seems like that creates more incentives for "chicken", where A's supporters say, "we're not going to approve B at all, so you'd better unconditionally approve A or someone else will win." I'm trying to decide if that's worse than the original system. In the original system they could say, "we're not going to conditionally approve B, so you'd better just put A as your favorite or someone else will win." We could combine the first two columns, so they just marked everyone they unconditionally approve in one column and mark everyone they conditionally approve in another column. Then the ballot is similar to other "3-grade" systems like MCA. What does conditional approval mean, in that case? Is my approval activated when someone is found who approves ANY of my favorites? When someone approves ALL of my favorites? Or it is fractionally activated depending on what fraction of my favorites they approve? (I'm not sure if any of these play nicely with reciprocity. I'll have to think about it...)