[EM] The IRV-Disease has reached my town.

classic Classic list List threaded Threaded
18 messages Options
Reply | Threaded
Open this post in threaded view
|

[EM] The IRV-Disease has reached my town.

walabio
        ⸘Howdy‽

        The article is here:

        http://petaluma360.com/news/9331748-181/ranked-choice-voting-gains-traction

        You need not read it because it says nothing new.  I shall post again later today about a way to make the ranked-choice crowd happy with Score-Voting.

        ¡Peace!

--

        “⸘Ŭalabio‽” <[hidden email]>

Skype:
        Walabio

An IntactWiki:
        http://intactwiki.org

        “You are entitled to your own opinion, but you are not entitled to your own facts.”
        ——
        Senator Daniel Patrick Moynihan

----
Election-Methods mailing list - see https://electorama.com/em for list info
Reply | Threaded
Open this post in threaded view
|

Re: [EM] The IRV-Disease has reached my town.

Chris Benham-2
I think voting reform activists in the US should welcome IRV and push to
make sure that there are
no abominable "details" like restricted ranking or eliminating
all-at-once all but the top two candidates.

Voters must be able to strictly rank from the top however many
candidates they wish, and the eliminations
must be one-at-a-time.

In my judgement this is better than Approval  (or something
strategically equivalent that uses ratings ballots
with more than 2 slots) because it doesn't have any annoying defection
incentive and properly meets
"Mutual Relative Majority".

If  you are a fan of the Condorcet criterion, then I think it is fine to
modify IRV by before each elimination
we check for a Condorcet winner (among the so far remaining candidates)
and when we find one we stop
and declare that candidate the winner.

If you want something more simple then I think Condorcet//Approval is
acceptable.

Voters simply rank the candidates they approve. Equal-ranking should
preferably be allowed.
A candidate that pairwise beats all the others wins. If there is no such
candidate then the most approved candidate
wins.

A bit better, but equivalent in 3-candidate elections and harder to
explain, are Smith//Approval and (one of
my favourites) Max Covered Approval.

Chris Benham

On 4/03/2019 5:48 am, ⸘Ŭalabio‽ wrote:

> ⸘Howdy‽
>
> The article is here:
>
> http://petaluma360.com/news/9331748-181/ranked-choice-voting-gains-traction
>
> You need not read it because it says nothing new.  I shall post again later today about a way to make the ranked-choice crowd happy with Score-Voting.
>
> ¡Peace!
>

---
This email has been checked for viruses by AVG.
https://www.avg.com

----
Election-Methods mailing list - see https://electorama.com/em for list info
Reply | Threaded
Open this post in threaded view
|

Re: [EM] The IRV-Disease has reached my town.

robert bristow-johnson



---------------------------- Original Message ----------------------------
Subject: Re: [EM] The IRV-Disease has reached my town.
From: "Chris Benham" <[hidden email]>
Date: Mon, March 4, 2019 7:41 am
To: [hidden email]
--------------------------------------------------------------------------

> I think voting reform activists in the US should welcome IRV and push to
> make sure that there are
> no abominable "details" like restricted ranking or eliminating
> all-at-once all but the top two candidates.

i dunno what that means.

> Voters must be able to strictly rank from the top however many
> candidates they wish, and the eliminations
> must be one-at-a-time.

then you need tougher ballot access restrictions.  SF had more than a dozen candidates but only 3 levels of ranking.  i think that it is reasonable for the law to assign a fixed number of ranking levels, but it should be more than 3.  then the number of signatures required to get a candidate's name on the ballot should be increased high enough that *typically* the number of candidates is no more than the number of ranking levels.  in Burlington Vermont, we had 5 levels and rarely more than 5 candidates on the ballot for mayor.

> In my judgement this is better than Approval  (or something
> strategically equivalent that uses ratings ballots
> with more than 2 slots) because it doesn't have any annoying defection
> incentive and properly meets
> "Mutual Relative Majority".

the problem with both Approval and Score Voting is that the voter must make a tactical decision about how high to score (or whether to approve) their second-choice.  it is inherently flawed in that manner.


> If  you are a fan of the Condorcet criterion, then I think it is fine to
> modify IRV by before each elimination
> we check for a Condorcet winner (among the so far remaining candidates)
> and when we find one we stop
> and declare that candidate the winner.

one way to do this is STV-BTR  ("Bottom Two Runoff").  so it's just like IRV, except when a candidate needs to be eliminated, it is not simply the guy on the bottom (of first-choice votes), the two bottom candidates are runoff against each other, counting only how voters rank them relative to each other (like in the IRV final round), only the winner of that Bottom Runoff advances to the next round.  this STV-BTR is Condorcet compliant.


> If you want something more simple then I think Condorcet//Approval is
> acceptable.
>
> Voters simply rank the candidates they approve. Equal-ranking should
> preferably be allowed.

Yes!

> A candidate that pairwise beats all the others wins. If there is no such
> candidate then the most approved candidate wins.

in a ranked ballot, what defines an "approved" candidate?  all unranked candidates are tied for last place on a ballot.  is any candidate that is ranked at all "approved"?  that would change and complicate the meaning of the ranked ballot.


> A bit better, but equivalent in 3-candidate elections and harder to
> explain, are Smith//Approval and (one of
> my favourites) Max Covered Approval.

i still think that Tideman RP using margins is the simplest meaningful method that elects the same candidate that Schulze (based on margins) elects when there are 3 in the Smith set.

both the rules and meaning for marking the ballots (Ranking Candidate A above Candidate B only means that if the election were held between only those two candidates, your vote is for Candidate A) and the decision algorithm (If more voters mark their ballots preferring Candidate A over Candidate B than the number of voters marking their ballots to the contrary, then Candidate B is not elected)m should be simple for everyone to understand.  as simple as possible.

i still see no advantage of any method that is not Condorcet compliant over one that is.

--

r b-j                         [hidden email]

"Imagination is more important than knowledge."
 

 

 

 


----
Election-Methods mailing list - see https://electorama.com/em for list info
Reply | Threaded
Open this post in threaded view
|

Re: [EM] The IRV-Disease has reached my town.

Chris Benham-2

Robert,

I am strongly of the view that voters should be allowed to bullet-vote
or rank every single candidate or anything in between.

(For IRV and my suggested IRV-Condorcet hybrid I am opposed to allowing
above-bottom equal-ranking without a relatively complicated mechanism
I invented.)

i think that it is reasonable for the law to assign a fixed number of ranking levels, but it should be more than 3. 

What is "reasonable" about it?  There is no such restriction in elections for the President of Ireland or seats in
the legislature of the Australian state of New South Wales and other places.

"..eliminating all-at-once all but the top two candidates."

i dunno what that means.
 
In other words I support the Alternative Vote and not what is called the "Contingent Vote" or "Top-Two IRV".

https://en.wikipedia.org/wiki/Contingent_vote
The contingent vote differs from the alternative vote which allows for many rounds of counting, eliminating only one weakest candidate each round.

in a ranked ballot, what defines an "approved" candidate?  all unranked candidates are tied for last place on a ballot.  is any candidate that is ranked at all "approved"?
Yes.

that would change and complicate the meaning of the ranked ballot.

Arguably "change" somewhat but I don't see how (overly) "complicate". Allowing voters to rank among unapproved
candidates makes the method more vulnerable to strategy and a lot more complicated.

i still see no advantage of any method that is not Condorcet compliant over one that is.

All Condorcet methods are vulnerable to Burial strategy and fail the Favorite Betrayal Criterion.

As has been pointed out to you on this list before, Margins is especially vulnerable to Burial.

Suppose sincere is:

46 A
44 B
10 C


I hope we are clear that if everyone bullet-votes like this then A is the Condorcet (and every other
type of) winner.

Supposing this scenario is accurately reflected in the pre-election polls, and so the B voters (perhaps
following advice from their party) decide "It looks like we are going to lose to A, maybe something good
will happen if we rank C" and so the votes cast are:

46 A
44 B>C
10 C

A>B 46-44 (margin=2)     B>C 44-10 (margin=34)   C>A 54-46 (margin=8)

Now Margins elects B,  rewarding the outrageous Burial strategy.

I can't tolerate any method that elects B in this scenario. Even assuming that all the votes are sincere,
B is clearly the weakest candidate (the least "approved" and positionally dominated and pairwise-beaten
by A.)

IRV and the IRV-Condorcet method I like elects A.

The Condorcet//Approval or Smith//Approval methods I like elect C.  The B voters "approved" C and they got C.

If you think it is "reasonable" to restrict the number of "ranking levels" and you want to allow voters to equal-rank
and presumably also skip ranking levels, then in my book you are talking about ratings rather than rankings
and you have something that stuffs up IRV or anything similar.

But it is fine for Condorcet//Approval or Smith//Approval or Smith//Top Ratings and several other voting methods
I like.

Chris Benham




On 5/03/2019 10:02 am, robert bristow-johnson wrote:



---------------------------- Original Message ----------------------------
Subject: Re: [EM] The IRV-Disease has reached my town.
From: "Chris Benham" [hidden email]
Date: Mon, March 4, 2019 7:41 am
To: [hidden email]
--------------------------------------------------------------------------

> I think voting reform activists in the US should welcome IRV and push to
> make sure that there are
> no abominable "details" like restricted ranking or eliminating
> all-at-once all but the top two candidates.

i dunno what that means.

> Voters must be able to strictly rank from the top however many
> candidates they wish, and the eliminations
> must be one-at-a-time.

then you need tougher ballot access restrictions.  SF had more than a dozen candidates but only 3 levels of ranking.  i think that it is reasonable for the law to assign a fixed number of ranking levels, but it should be more than 3.  then the number of signatures required to get a candidate's name on the ballot should be increased high enough that *typically* the number of candidates is no more than the number of ranking levels.  in Burlington Vermont, we had 5 levels and rarely more than 5 candidates on the ballot for mayor.

> In my judgement this is better than Approval  (or something
> strategically equivalent that uses ratings ballots
> with more than 2 slots) because it doesn't have any annoying defection
> incentive and properly meets
> "Mutual Relative Majority".

the problem with both Approval and Score Voting is that the voter must make a tactical decision about how high to score (or whether to approve) their second-choice.  it is inherently flawed in that manner.


> If  you are a fan of the Condorcet criterion, then I think it is fine to
> modify IRV by before each elimination
> we check for a Condorcet winner (among the so far remaining candidates)
> and when we find one we stop
> and declare that candidate the winner.

one way to do this is STV-BTR  ("Bottom Two Runoff").  so it's just like IRV, except when a candidate needs to be eliminated, it is not simply the guy on the bottom (of first-choice votes), the two bottom candidates are runoff against each other, counting only how voters rank them relative to each other (like in the IRV final round), only the winner of that Bottom Runoff advances to the next round.  this STV-BTR is Condorcet compliant.


> If you want something more simple then I think Condorcet//Approval is
> acceptable.
>
> Voters simply rank the candidates they approve. Equal-ranking should
> preferably be allowed.

Yes!

> A candidate that pairwise beats all the others wins. If there is no such
> candidate then the most approved candidate wins.

in a ranked ballot, what defines an "approved" candidate?  all unranked candidates are tied for last place on a ballot.  is any candidate that is ranked at all "approved"?  that would change and complicate the meaning of the ranked ballot.


> A bit better, but equivalent in 3-candidate elections and harder to
> explain, are Smith//Approval and (one of
> my favourites) Max Covered Approval.

i still think that Tideman RP using margins is the simplest meaningful method that elects the same candidate that Schulze (based on margins) elects when there are 3 in the Smith set.

both the rules and meaning for marking the ballots (Ranking Candidate A above Candidate B only means that if the election were held between only those two candidates, your vote is for Candidate A) and the decision algorithm (If more voters mark their ballots preferring Candidate A over Candidate B than the number of voters marking their ballots to the contrary, then Candidate B is not elected)m should be simple for everyone to understand.  as simple as possible.

i still see no advantage of any method that is not Condorcet compliant over one that is.

--

r b-j                         [hidden email]

"Imagination is more important than knowledge."
 

 

 

 


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

Virus-free. www.avg.com

----
Election-Methods mailing list - see https://electorama.com/em for list info
Reply | Threaded
Open this post in threaded view
|

Re: [EM] The IRV-Disease has reached my town.

Andrew Myers
In reply to this post by Chris Benham-2
Chris Benham wrote:
> If  you are a fan of the Condorcet criterion, then I think it is fine
> to modify IRV by before each elimination
> we check for a Condorcet winner (among the so far remaining
> candidates) and when we find one we stop
> and declare that candidate the winner.
This is one of the algorithms supported by CIVS, the one called
"Condorcet-IRV" though to my understanding it was originally proposed by
Thomas Hill.

-- Andrew

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

andru.vcf (209 bytes) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Re: [EM] [ApprovalVoting] Re: The IRV-Disease has reached my town.

Juho Laatu-4
In reply to this post by Chris Benham-2
On 05 Mar 2019, at 07:45, Chris Benham [hidden email] [ApprovalVoting] <[hidden email]> wrote:

Robert,



in a ranked ballot, what defines an "approved" candidate?  all unranked candidates are tied for last place on a ballot.  is any candidate that is ranked at all "approved"?
Yes.

that would change and complicate the meaning of the ranked ballot.

Arguably "change" somewhat but I don't see how (overly) "complicate". Allowing voters to rank among unapproved
candidates makes the method more vulnerable to strategy and a lot more complicated.

One might face problems sooner with sincere voting than with strategic voting.

First preferences could be as follows.

30: far-left
21: left
19: right
30: far-right

If we assume that left hates right, and right hates left, the natural approval limit would be between the left wing and right wing parties. We would get mostly votes that rank only left wing or only right wing candidates. And the winner would be with good probability the far-left candidate, not the expected Condorcet winner (left).

The problem with "implicit approval cutoff after the ranked candidates" is that voters would be encouraged not to rank all the major candidates. Not good for Condorcet. That would remove some important information. In this example the sincere Condorcet winner could not be identified anymore.

46 A
44 B>C
10 C

A>B 46-44 (margin=2)     B>C 44-10 (margin=34)   C>A 54-46 (margin=8)

Now Margins elects B,  rewarding the outrageous Burial strategy.

I can't tolerate any method that elects B in this scenario. Even assuming that all the votes are sincere,
B is clearly the weakest candidate (the least "approved" and positionally dominated and pairwise-beaten
by A.)

Also here you assume that there is an implicit approval cutoff after the ranked candidates. What if the votes are sincere and there is no implicit approval cutoff?

Juho


P.S. I think the STV-BTR method that Robert proposed could make a lot of sense in societies where IRV way of thinking is strong.

P.P.S. Limiting the number of ranking levels or number of ranked candidates could make sense when the number of candidates is very high, or just to keep things simple for the vote counting process, or to keep things simple enough for the voters (not to frighten them with the idea of ranking all 100 candidates). I.e. not theoretically ideal, but in practical situations ranking some candidates may be much better than ranking only one, or not bothering to vote at all.

----
Election-Methods mailing list - see https://electorama.com/em for list info
Reply | Threaded
Open this post in threaded view
|

Re: [EM] The IRV-Disease has reached my town.

robert bristow-johnson
In reply to this post by Chris Benham-2

i am not on the other two lists, so i can't post to them.

---------------------------- Original Message ----------------------------
Subject: Re: [EM] The IRV-Disease has reached my town.
From: "Chris Benham" <[hidden email]>
Date: Mon, March 4, 2019 9:45 pm
To: [hidden email]
Cc: [hidden email]
[hidden email]
"robert bristow-johnson" <[hidden email]>
--------------------------------------------------------------------------

> I am strongly of the view that voters should be allowed to bullet-vote
> or rank every single candidate or anything in between.

Fine. 

 

>> i think that it is reasonable for the law to assign a fixed number of
>> ranking levels, but it should be more than 3.
>
> What is "reasonable" about it? 

finite amount of real-estate on the paper ballot.


>
> https://en.wikipedia.org/wiki/Contingent_vote
>> The contingent vote differs from the alternative vote
>> <https://en.wikipedia.org/wiki/Instant-runoff_voting>which allows for
>> many rounds of counting, eliminating only one weakest candidate each
>> round.
>
>> in a ranked ballot, what defines an "approved" candidate?  all
>> unranked candidates are tied for last place on a ballot.  is any
>> candidate that is ranked at all "approved"?
> Yes.
>
>> that would change and complicate the meaning of the ranked ballot.
>
> Arguably "change" somewhat but I don't see how (overly) "complicate".
> Allowing voters to rank among unapproved
> candidates makes the method more vulnerable to strategy and a lot more
> complicated.

it adds rules.  without this addition rank/approval thing, the rules of tabulation are simpler.


>> i still see no advantage of any method that is not Condorcet compliant
>> over one that is.
>
> All Condorcet methods are vulnerable to Burial strategy and fail the
> Favorite Betrayal Criterion.

yeah, but if a CW exists and, for whatever reason, you do not elect the CW (as what happened to us in Burlington in 2009), you are electing a candidate when **more** of us marked our ballots that we preferred someone else.  that's hardly democratic.


> As has been pointed out to you on this list before, Margins is
> especially vulnerable to Burial.

yes it has been pointed out to me.  with anecdotes.  with "suppose this..."

> Suppose sincere is:
>
> 46 A
> 44 B
> 10 C
>
>
> I hope we are clear that if everyone bullet-votes like this then A is
> the Condorcet (and every other
> type of) winner.
>
> Supposing this scenario is accurately reflected in the pre-election
> polls, and so the B voters (perhaps
> following advice from their party) decide "It looks like we are going to
> lose to A, maybe something good
> will happen if we rank C" and so the votes cast are:
>
> 46 A
> 44 B>C
> 10 C
>
> A>B 46-44 (margin=2)     B>C 44-10 (margin=34)   C>A 54-46 (margin=8)
>
> Now Margins elects B,  rewarding the outrageous Burial strategy.

yup it's the impossible Arrow.

how would we differentiate that scenario with 44 B>C that was strategic from the same exact set of ballots cast that are sincere?  if it's sincere, then B should win.

but it's a good example of how, if the election was close and you could convince enough B voters to vote tactically, that it rewards Burial.  but it's still a contrived example.

maybe Winning Votes is better than Margins.  i dunno.

> I can't tolerate any method that elects B in this scenario. Even
> assuming that all the votes are sincere,
> B is clearly the weakest candidate (the least "approved" and
> positionally dominated and pairwise-beaten by A.)

but only very closely beaten by A where it appears that otherwise voters more strongly prefer B over C and C over A.  if the vote is sincere, it looks to me that A is *barely* a plurality winner.

>
> IRV and the IRV-Condorcet method I like elects A.

is that IRV-Condorcet the same as STV-BTR?


> The Condorcet//Approval or Smith//Approval methods I like elect C.  The
> B voters "approved" C and they got C.

looks weak to me (a lot more voters prefer B to C).  i think you would have some pitchforks in the street if you gave that election to C.

 

> If you think it is "reasonable" to restrict the number of "ranking
> levels" and you want to allow voters to equal-rank
> and presumably also skip ranking levels, then in my book you are talking
> about ratings rather than rankings

No.  If it's scoring (or "rating"), the voter must tactically decide how much they will score their second choice.  It doesn't make a difference in ranking.  And every voter's voting power should be equal.

> and you have something that stuffs up IRV or anything similar.

dunno what that means, either.

> But it is fine for Condorcet//Approval or Smith//Approval or Smith//Top
> Ratings and several other voting methods I like.

I just want it so that when A is compared to B, that everyone who weighs in on that choice have an equal say (that's One-Person-One-Vote).

And, excepting a cycle (which is an "impossible" mess), I just want it that if more voters mark their ballots preferring A to B than the number of voters marking their ballots to the contrary, then B is not elected.

Them's are my dogma.  And eliminating obvious tactical burdens from voters follows close behind to this dogma.  Simplicity in voting and tabulation comes next.


--

r b-j                         [hidden email]

"Imagination is more important than knowledge."
 

 

 

 


----
Election-Methods mailing list - see https://electorama.com/em for list info
Reply | Threaded
Open this post in threaded view
|

Re: [EM] [ApprovalVoting] Re: The IRV-Disease has reached my town.

robert bristow-johnson
In reply to this post by Juho Laatu-4



---------------------------- Original Message ----------------------------
Subject: Re: [EM] [ApprovalVoting] Re: The IRV-Disease has reached my town.
From: "Juho Laatu" <[hidden email]>
Date: Tue, March 5, 2019 12:12 am
To: "EM" <[hidden email]>
--------------------------------------------------------------------------


> P.S. I think the STV-BTR method that Robert proposed could make a lot of sense in societies where IRV way of thinking is strong.

i can't take any credit for proposing that.  i dunno who thought of it first.  wasn't me.

the thing that i sorta don't like about it is that it is still IRV and not strictly precinct-summable.  otherwise it's okay, i guess.  and we could still precinct-sum each defeat pair as a check on the election if there *is* a CW.  But it would be necessary that a physical instrument representing each ballot be transferred from the precinct to the central tabulation location to tally the election and determine the winner.  if transparency is salient, then STV-BTR is still more like IRV and less like Condorcet.


> P.P.S. Limiting the number of ranking levels or number of ranked candidates could make sense when the number of candidates is very high, or just to keep things simple for the vote counting process, or to keep things simple enough for the voters (not to frighten them with the idea of ranking all 100 candidates). I.e. not theoretically ideal, but in practical situations ranking some candidates may be much better than ranking only one, or not bothering to vote at all.----
 

the problem is, in 2010, when the IRV opponents were gearing up their rhetoric in Burlington, they **falsely** claimed there was "voter disenfranchisement" with IRV in Burlington.  because they read some propaganda that pointed to *real* disenfranchisement in San Francisco for voters that didn't realize who the race was really gonna end up between.  these "disenfranchised" voters marked their ballots 1,2,3 and none of those candidates were in the IRV final round.  they had no say in who was ultimately elected.  and, like plurality, they would have to guess who is likely to win and vote insincerely for the major candidate that they dislike the least.

in Burlington, we had 5 candidates and 5 ranking levels.  i think 5 or 6 levels is enough and we should make our ballot access laws tough enough to *approximately* limit the number of candidates on the ballot to that number.  maybe there's enough room on the ballot for 8 or 9 levels, but i don't think we would need them.


--

r b-j                         [hidden email]

"Imagination is more important than knowledge."
 

 

 

 


----
Election-Methods mailing list - see https://electorama.com/em for list info
Reply | Threaded
Open this post in threaded view
|

Re: [EM] [ApprovalVoting] Re: The IRV-Disease has reached my town.

Juho Laatu-4
On 05 Mar 2019, at 10:56, robert bristow-johnson <[hidden email]> wrote:



---------------------------- Original Message ----------------------------
Subject: Re: [EM] [ApprovalVoting] Re: The IRV-Disease has reached my town.
From: "Juho Laatu" <[hidden email]>
Date: Tue, March 5, 2019 12:12 am
To: "EM" <[hidden email]>
--------------------------------------------------------------------------


> P.S. I think the STV-BTR method that Robert proposed could make a lot of sense in societies where IRV way of thinking is strong.

i can't take any credit for proposing that.  i dunno who thought of it first.  wasn't me.

the thing that i sorta don't like about it is that it is still IRV and not strictly precinct-summable.  otherwise it's okay, i guess.  and we could still precinct-sum each defeat pair as a check on the election if there *is* a CW.  But it would be necessary that a physical instrument representing each ballot be transferred from the precinct to the central tabulation location to tally the election and determine the winner.  if transparency is salient, then STV-BTR is still more like IRV and less like Condorcet.

Precinct-summability could be the next evolutionary step (in an IRV oriented society) after the basic idea of STV-BTR has been accepted.


> P.P.S. Limiting the number of ranking levels or number of ranked candidates could make sense when the number of candidates is very high, or just to keep things simple for the vote counting process, or to keep things simple enough for the voters (not to frighten them with the idea of ranking all 100 candidates). I.e. not theoretically ideal, but in practical situations ranking some candidates may be much better than ranking only one, or not bothering to vote at all.----
 

the problem is, in 2010, when the IRV opponents were gearing up their rhetoric in Burlington, they **falsely** claimed there was "voter disenfranchisement" with IRV in Burlington.  because they read some propaganda that pointed to *real* disenfranchisement in San Francisco for voters that didn't realize who the race was really gonna end up between.  these "disenfranchised" voters marked their ballots 1,2,3 and none of those candidates were in the IRV final round.  they had no say in who was ultimately elected.  and, like plurality, they would have to guess who is likely to win and vote insincerely for the major candidate that they dislike the least.

in Burlington, we had 5 candidates and 5 ranking levels.  i think 5 or 6 levels is enough and we should make our ballot access laws tough enough to *approximately* limit the number of candidates on the ballot to that number.  maybe there's enough room on the ballot for 8 or 9 levels, but i don't think we would need them.

There may be many kind of elections, with different number of candidates, with or without reliable polls beforehand, and voters may be different. One should take all this into account when planning an ideal election method for each society, and their different elections. Where I live, we could have some 200 candidates in the parliamentary elections. The system might evolve towards STV or something like that. Most people would not like to rank all the candidates anyway. Having all the candidates named on the ballot paper would be tedious too. The best idea might be to continue the tradition of writing the number of the favourite candidate on the (blank) ballot sheet, and allow voters to write more than one number on that ballot sheet. But how many numbers should one allow? 200 sounds excessive. And what if people just want to vote for their favourite party? Maybe one should have some inheritance rules that allow the party to inherit the vote when all the listed candidates have been eliminated. This example is a proportional election, and that complicates things a bit. But my point is just that there are different societies with different traditions, different preferences, different needs and different elections, and we just need to take that into account, and adapt to whatever the real-life situation is.

Juho


--

r b-j                         [hidden email]

"Imagination is more important than knowledge."
 

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


----
Election-Methods mailing list - see https://electorama.com/em for list info
Reply | Threaded
Open this post in threaded view
|

Re: [EM] [ApprovalVoting] Re: The IRV-Disease has reached my town.

Chris Benham-2
In reply to this post by Juho Laatu-4

Juho,

46 A
44 B>C
10 C

A>B 46-44 (margin=2)���� B>C 44-10 (margin=34)�� C>A 54-46 (margin=8)

Now Margins elects B,� rewarding the outrageous Burial strategy.

I can't tolerate any method that elects B in this scenario. Even assuming that all the votes are sincere,
B is clearly the weakest candidate (the least "approved" and positionally dominated and pairwise-beaten
by A.)

Also here you assume that there is an implicit approval cutoff after the ranked candidates. What if the votes are sincere and there is no implicit approval cutoff?

No, I'm only "assuming" that positional information is more meaningful than which candidate is "closer" to being the Condorcet winner
according to Margins or which candidate needs the fewest additional bullet-votes to become the CW.

Given that Margins is very vulnerable to Burial strategy the argument that it's worth putting up with that because with sincere votes B
in this scenario is the best (or a good or even acceptable) candidate is .. what??

Or to put it another way, assuming the votes are sincere, you arguments against electing A or C are what?

One might face problems sooner with sincere voting than with strategic voting.

First preferences could be as follows.

30: far-left
21: left
19: right
30: far-right

If we assume that left hates right, and right hates left, the natural approval limit would be between the left wing and right wing parties. We would get mostly votes that rank only left wing or only right wing candidates. And the winner would be with good probability the far-left candidate, not the expected Condorcet winner (left).

That doesn't bother me much because (a) far-left may be higher "Social Utility" than left and (b) probably enough right voters would
be aware that the result is unlikely to be decided by Approval and so they would not be taking a huge risk by sincerely ranking left
over far-left.

But having said that, Smith//Approval using ballots that� allow voters to rank among candidates they don't approve would not be
in my book too bad (and much better than Margins).

P.S. I think the STV-BTR method that Robert proposed could make a lot of sense in societies where IRV way of thinking is strong.
That method doesn't have good criterion compliances. It's just a gimmick to smuggle Condorcet compliance past IRV enthusiasts.
The alternative of just checking for a CW (among remaining candidates) before each elimination is much better.

P.P.S. Limiting the number of ranking levels or number of ranked candidates could make sense when the number of candidates is very high, or just to keep things simple for the vote counting process, or to keep things simple enough for the voters (not to frighten them with the idea of ranking all 100 candidates). I.e. not theoretically ideal, but in practical situations ranking some candidates may be much better than ranking only one, or not bothering to vote at all.

Limiting the number of candidates the voter is allowed to rank makes no sense. What has happened to your concern
about "removing information" on who the sincere/ "expected" CW is?

There would be nothing "frightening" about ranking all the candidates if doing so is purely optional. But voters who wish
to vote a full ranking should be allowed to.

Chris� Benham


On 5/03/2019 6:42 pm, Juho Laatu wrote:
On 05 Mar 2019, at 07:45, Chris Benham [hidden email] [ApprovalVoting] <[hidden email]> wrote:

Robert,



in a ranked ballot, what defines an "approved" candidate?� all unranked candidates are tied for last place on a ballot.� is any candidate that is ranked at all "approved"?
Yes.

that would change and complicate the meaning of the ranked ballot.

Arguably "change" somewhat but I don't see how (overly) "complicate". Allowing voters to rank among unapproved
candidates makes the method more vulnerable to strategy and a lot more complicated.

One might face problems sooner with sincere voting than with strategic voting.

First preferences could be as follows.

30: far-left
21: left
19: right
30: far-right

If we assume that left hates right, and right hates left, the natural approval limit would be between the left wing and right wing parties. We would get mostly votes that rank only left wing or only right wing candidates. And the winner would be with good probability the far-left candidate, not the expected Condorcet winner (left).

The problem with "implicit approval cutoff after the ranked candidates" is that voters would be encouraged not to rank all the major candidates. Not good for Condorcet. That would remove some important information. In this example the sincere Condorcet winner could not be identified anymore.

46 A
44 B>C
10 C

A>B 46-44 (margin=2)���� B>C 44-10 (margin=34)�� C>A 54-46 (margin=8)

Now Margins elects B,� rewarding the outrageous Burial strategy.

I can't tolerate any method that elects B in this scenario. Even assuming that all the votes are sincere,
B is clearly the weakest candidate (the least "approved" and positionally dominated and pairwise-beaten
by A.)

Also here you assume that there is an implicit approval cutoff after the ranked candidates. What if the votes are sincere and there is no implicit approval cutoff?

Juho


P.S. I think the STV-BTR method that Robert proposed could make a lot of sense in societies where IRV way of thinking is strong.

P.P.S. Limiting the number of ranking levels or number of ranked candidates could make sense when the number of candidates is very high, or just to keep things simple for the vote counting process, or to keep things simple enough for the voters (not to frighten them with the idea of ranking all 100 candidates). I.e. not theoretically ideal, but in practical situations ranking some candidates may be much better than ranking only one, or not bothering to vote at all.

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

Virus-free. www.avg.com

----
Election-Methods mailing list - see https://electorama.com/em for list info
Reply | Threaded
Open this post in threaded view
|

Re: [EM] The IRV-Disease has reached my town.

Chris Benham-2
In reply to this post by robert bristow-johnson

> What is "reasonable" about it? 

finite amount of real-estate on the paper ballot.

I can see that as an argument against compelling voters to rank all the
candidates, as happens in most Australian elections, but not for denying
anyone the right to rank as many candidates as they wish.

yeah, but if a CW exists and, for whatever reason, you do not elect the CW (as what happened to us in Burlington in 2009), you are electing a candidate when **more** of us marked our ballots that we preferred someone else.  that's hardly democratic.
Yes that is arguable and I like Condorcet criterion compliance, but that isn't the same as saying there
can be "no advantage" to doing without it.

IRV and the IRV-Condorcet method I like elects A.

is that IRV-Condorcet the same as STV-BTR?
No. It checks for a CW before the first normal IRV-style elimination and if there isn't one does so before each
elimination among remaining candidates and elects the first one it finds.

STV-BTR doesn't have good criterion compliances but I forget the details.

how would we differentiate that scenario with 44 B>C that was strategic from the same exact set of ballots cast that are sincere?  if it's sincere, then B should win.
Because ..??

Chris Benham

On 5/03/2019 7:09 pm, robert bristow-johnson wrote:

i am not on the other two lists, so i can't post to them.

---------------------------- Original Message ----------------------------
Subject: Re: [EM] The IRV-Disease has reached my town.
From: "Chris Benham" [hidden email]
Date: Mon, March 4, 2019 9:45 pm
To: [hidden email]
Cc: [hidden email]
[hidden email]
"robert bristow-johnson" [hidden email]
--------------------------------------------------------------------------

> I am strongly of the view that voters should be allowed to bullet-vote
> or rank every single candidate or anything in between.

Fine. 

 

>> i think that it is reasonable for the law to assign a fixed number of
>> ranking levels, but it should be more than 3.
>
> What is "reasonable" about it? 

finite amount of real-estate on the paper ballot.


>
> https://en.wikipedia.org/wiki/Contingent_vote
>> The contingent vote differs from the alternative vote
>> <https://en.wikipedia.org/wiki/Instant-runoff_voting>which allows for
>> many rounds of counting, eliminating only one weakest candidate each
>> round.
>
>> in a ranked ballot, what defines an "approved" candidate?  all
>> unranked candidates are tied for last place on a ballot.  is any
>> candidate that is ranked at all "approved"?
> Yes.
>
>> that would change and complicate the meaning of the ranked ballot.
>
> Arguably "change" somewhat but I don't see how (overly) "complicate".
> Allowing voters to rank among unapproved
> candidates makes the method more vulnerable to strategy and a lot more
> complicated.

it adds rules.  without this addition rank/approval thing, the rules of tabulation are simpler.


>> i still see no advantage of any method that is not Condorcet compliant
>> over one that is.
>
> All Condorcet methods are vulnerable to Burial strategy and fail the
> Favorite Betrayal Criterion.

yeah, but if a CW exists and, for whatever reason, you do not elect the CW (as what happened to us in Burlington in 2009), you are electing a candidate when **more** of us marked our ballots that we preferred someone else.  that's hardly democratic.


> As has been pointed out to you on this list before, Margins is
> especially vulnerable to Burial.

yes it has been pointed out to me.  with anecdotes.  with "suppose this..."

> Suppose sincere is:
>
> 46 A
> 44 B
> 10 C
>
>
> I hope we are clear that if everyone bullet-votes like this then A is
> the Condorcet (and every other
> type of) winner.
>
> Supposing this scenario is accurately reflected in the pre-election
> polls, and so the B voters (perhaps
> following advice from their party) decide "It looks like we are going to
> lose to A, maybe something good
> will happen if we rank C" and so the votes cast are:
>
> 46 A
> 44 B>C
> 10 C
>
> A>B 46-44 (margin=2)     B>C 44-10 (margin=34)   C>A 54-46 (margin=8)
>
> Now Margins elects B,  rewarding the outrageous Burial strategy.

yup it's the impossible Arrow.

how would we differentiate that scenario with 44 B>C that was strategic from the same exact set of ballots cast that are sincere?  if it's sincere, then B should win.

but it's a good example of how, if the election was close and you could convince enough B voters to vote tactically, that it rewards Burial.  but it's still a contrived example.

maybe Winning Votes is better than Margins.  i dunno.

> I can't tolerate any method that elects B in this scenario. Even
> assuming that all the votes are sincere,
> B is clearly the weakest candidate (the least "approved" and
> positionally dominated and pairwise-beaten by A.)

but only very closely beaten by A where it appears that otherwise voters more strongly prefer B over C and C over A.  if the vote is sincere, it looks to me that A is *barely* a plurality winner.

>
> IRV and the IRV-Condorcet method I like elects A.

is that IRV-Condorcet the same as STV-BTR?


> The Condorcet//Approval or Smith//Approval methods I like elect C.  The
> B voters "approved" C and they got C.

looks weak to me (a lot more voters prefer B to C).  i think you would have some pitchforks in the street if you gave that election to C.

 

> If you think it is "reasonable" to restrict the number of "ranking
> levels" and you want to allow voters to equal-rank
> and presumably also skip ranking levels, then in my book you are talking
> about ratings rather than rankings

No.  If it's scoring (or "rating"), the voter must tactically decide how much they will score their second choice.  It doesn't make a difference in ranking.  And every voter's voting power should be equal.

> and you have something that stuffs up IRV or anything similar.

dunno what that means, either.

> But it is fine for Condorcet//Approval or Smith//Approval or Smith//Top
> Ratings and several other voting methods I like.

I just want it so that when A is compared to B, that everyone who weighs in on that choice have an equal say (that's One-Person-One-Vote).

And, excepting a cycle (which is an "impossible" mess), I just want it that if more voters mark their ballots preferring A to B than the number of voters marking their ballots to the contrary, then B is not elected.

Them's are my dogma.  And eliminating obvious tactical burdens from voters follows close behind to this dogma.  Simplicity in voting and tabulation comes next.


--

r b-j                         [hidden email]

"Imagination is more important than knowledge."
 

 

 

 


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

Virus-free. www.avg.com

----
Election-Methods mailing list - see https://electorama.com/em for list info
Reply | Threaded
Open this post in threaded view
|

Re: [EM] The IRV-Disease has reached my town.

Chris Benham-2
In reply to this post by Andrew Myers

Andrew,

An author of an article on different versions of Condorcet-IRV labeled it after me but I would prefer it
was named something (or after someone) else.

http://www.votingmatters.org.uk/ISSUE29/I29P1.PDF

https://www.accuratedemocracy.com/l_lor1.htm

Condorcet + IRV, the Best of Both ?

Several people have invented voting rules that resolve�voting cycles�by combining Condorcet's rule with Hare's rule in various ways.� Each of these hybrid rules enacts the Condorcet winner when there is one.� When a cycle occurs, it uses the Hare process of eliminations and transfers until one option tops each of the other remaining options.

David Hill, formerly of England's Electoral Reform Society, proposed in 1988 making the current Condorcet winner exempt from elimination at each step in an IRV tally. His goal was to make one excellent rule for both single- and multi- winner elections. (Ref.�H)

Robert Loring, formerly of FairVote, proposed in 1990 changing the criterion for winning Hare from a majority of ballots to the Condorcet criterion.� This eliminates the option which holds the top rank on the fewest ballots - until one option can top each of the others 1-against-1.� To�focus on the tally process, we might call Hare's rule "Majority-IRV" (M-IRV) and Loring's variation rule "Condorcet-IRV" (C-IRV).

Thanks.

Chris Benham

On 5/03/2019 4:20 pm, Andrew Myers wrote:
Chris Benham wrote:
If� you are a fan of the Condorcet criterion, then I think it is fine to modify IRV by before each elimination
we check for a Condorcet winner (among the so far remaining candidates) and when we find one we stop
and declare that candidate the winner.
This is one of the algorithms supported by CIVS, the one called "Condorcet-IRV" though to my understanding it was originally proposed by Thomas Hill.

-- Andrew

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

Virus-free. www.avg.com

----
Election-Methods mailing list - see https://electorama.com/em for list info
Reply | Threaded
Open this post in threaded view
|

Re: [EM] [ApprovalVoting] Re: The IRV-Disease has reached my town.

Juho Laatu-4
In reply to this post by Chris Benham-2
On 05 Mar 2019, at 16:04, Chris Benham [hidden email] [ApprovalVoting] <[hidden email]> wrote:

Also here you assume that there is an implicit approval cutoff after the ranked candidates. What if the votes are sincere and there is no implicit approval cutoff?

No, I'm only "assuming" that positional information is more meaningful than which candidate is "closer" to being the Condorcet winner
according to Margins or which candidate needs the fewest additional bullet-votes to become the CW.

If the added implicit information is not approval of all the ranked candidates, does the "positional information" maybe refer to some Borda like rating information? (or maybe implicit ratings) (I don't really like Borda ratings as additional information either because of the associated nomination related problems.)


Given that Margins is very vulnerable to Burial strategy the argument that it's worth putting up with that because with sincere votes B
in this scenario is the best (or a good or even acceptable) candidate is .. what??

Or to put it another way, assuming the votes are sincere, you arguments against electing A or C are what?

I just thought that B might be ok if we take the votes to be plain rankings (= pairwise preferences) with no additional approval or positional information assumed. The mentioned votes are of course very extreme (only three kind if voters), so they are not a typical set of votes form a real life election.

46 A
44 B>C
10 C

One could think that A is a left wing candidate, C is moderate right, and B is far right (because there are 44 voters that rank them in such linear order). A voters don't seem to care which one of the right wing candidates wins (they are tied in every single vote). The B voters have a clear and understandable position, with full rankings. The C voters have not ranked B. Maybe they are so centrist that A and B are equally good to them.

Condorcet methods can be said to elect a compromise candidate that is not too bad for anyone (sometimes the Condorcet winner might have no first preference votes). I therefore try to find an explanation to why B might win, from this point of view (= elect he best compromise (that is not very disliked)).

Since the A voters did not rank C above B, we must assume that they are perfectly ok with electing B, if A does not win. Same with C voters. B voters have a clear preference C>A (if B can not win).

If A wins, B and C supporters (54, majority) clearly think that C would have been a better choice (and 44 voters would prefer B). If C wins, 46 voters would prefer A, and 44 voters would prefer B. If B wins, 46 voters would prefer A, and 10 voters would prefer C. B doesn't look too bad in this comparison. B might be a better compromise than A or C. I.e. less complaints and rebellions after the election.

(I tried to avoid the "few votes short of being a Condorcet winner" argument since you might not appreciate it.)


One might face problems sooner with sincere voting than with strategic voting.

First preferences could be as follows.

30: far-left
21: left
19: right
30: far-right

If we assume that left hates right, and right hates left, the natural approval limit would be between the left wing and right wing parties. We would get mostly votes that rank only left wing or only right wing candidates. And the winner would be with good probability the far-left candidate, not the expected Condorcet winner (left).

That doesn't bother me much because (a) far-left may be higher "Social Utility" than left

I don't know what "Social Utility" means here. I guess any of the candidates could have that. It might not show up in rankings nor in approvals.

and (b) probably enough right voters would
be aware that the result is unlikely to be decided by Approval and so they would not be taking a huge risk by sincerely ranking left
over far-left.

I guess this depends on the method. I can't tell if all "ranking + implicit approval" methods would behave well in this respect. I also don't like very much the idea of voters having to cast strategic votes instead of sincere votes (i.e. implicit approvals in the ballots would not mean that the voter would approve the candidate, but something strategic instead).


But having said that, Smith//Approval using ballots that� allow voters to rank among candidates they don't approve would not be
in my book too bad (and much better than Margins).

P.S. I think the STV-BTR method that Robert proposed could make a lot of sense in societies where IRV way of thinking is strong.
That method doesn't have good criterion compliances. It's just a gimmick to smuggle Condorcet compliance past IRV enthusiasts.
The alternative of just checking for a CW (among remaining candidates) before each elimination is much better.

I'm not a big believer in criterion compliance in real life election methods. In theoretical studies different criteria are excellent measurement tools, but in real life elections nobody cares if the method performs well in some theoretical situations. Often slightly modified (heuristic style, not necessarily "mathematically clean") methods are close enough to meeting those criteria on practice anyway. Also my theoretically ideal "maximally strategy resistant method" might be one that fails to meet most of the named criteria, but does so intentionally in order to violate each one (or many) of them just a little bit, so that it can maximise resistance against all kind of strategies (and keep its worst vulnerability least bad).


P.P.S. Limiting the number of ranking levels or number of ranked candidates could make sense when the number of candidates is very high, or just to keep things simple for the vote counting process, or to keep things simple enough for the voters (not to frighten them with the idea of ranking all 100 candidates). I.e. not theoretically ideal, but in practical situations ranking some candidates may be much better than ranking only one, or not bothering to vote at all.

Limiting the number of candidates the voter is allowed to rank makes no sense. What has happened to your concern
about "removing information" on who the sincere/ "expected" CW is?

Typically the intention is not to limit the number of candidates that can be ranked (when compared to the situation before he change) but to make that number higher, while allowing it not to be very high. I am still worried about removing information, but I can accept some limitations sometimes (when full ranking is not feasible, or when most voters would not rank all candidates anyway, or when other solutions are not politically possible). The limits should be such that they probably do not lead to not electing the sincere Condorcet winner (or the best candidate when there is no Condorcet winner).

It is for example possible that there is an election with 5 serious candidates (based on some good polling information) and 50 other candidates with no chances to win. In that situation it would not be a big problem to limit the number of ranked candidates to say 7. The ballots could be simpler that way, and voting would not be too tedious. I would not mind someone starting even from 3, if that is an improvement e.g. to the earlier FPTP.


There would be nothing "frightening" about ranking all the candidates if doing so is purely optional. But voters who wish
to vote a full ranking should be allowed to.

That is possible, and a positive thing to do, but I do understand that sometimes also less perfect methods can be "perfect" or "sufficient" for the current real life situation.

Juho



Chris� Benham


On 5/03/2019 6:42 pm, Juho Laatu wrote:
On 05 Mar 2019, at 07:45, Chris Benham [hidden email] [ApprovalVoting] <[hidden email]> wrote:

Robert,



in a ranked ballot, what defines an "approved" candidate?� all unranked candidates are tied for last place on a ballot.� is any candidate that is ranked at all "approved"?
Yes.

that would change and complicate the meaning of the ranked ballot.

Arguably "change" somewhat but I don't see how (overly) "complicate". Allowing voters to rank among unapproved
candidates makes the method more vulnerable to strategy and a lot more complicated.

One might face problems sooner with sincere voting than with strategic voting.

First preferences could be as follows.

30: far-left
21: left
19: right
30: far-right

If we assume that left hates right, and right hates left, the natural approval limit would be between the left wing and right wing parties. We would get mostly votes that rank only left wing or only right wing candidates. And the winner would be with good probability the far-left candidate, not the expected Condorcet winner (left).

The problem with "implicit approval cutoff after the ranked candidates" is that voters would be encouraged not to rank all the major candidates. Not good for Condorcet. That would remove some important information. In this example the sincere Condorcet winner could not be identified anymore.

46 A
44 B>C
10 C

A>B 46-44 (margin=2)���� B>C 44-10 (margin=34)�� C>A 54-46 (margin=8)

Now Margins elects B,� rewarding the outrageous Burial strategy.

I can't tolerate any method that elects B in this scenario. Even assuming that all the votes are sincere,
B is clearly the weakest candidate (the least "approved" and positionally dominated and pairwise-beaten
by A.)

Also here you assume that there is an implicit approval cutoff after the ranked candidates. What if the votes are sincere and there is no implicit approval cutoff?

Juho


P.S. I think the STV-BTR method that Robert proposed could make a lot of sense in societies where IRV way of thinking is strong.

P.P.S. Limiting the number of ranking levels or number of ranked candidates could make sense when the number of candidates is very high, or just to keep things simple for the vote counting process, or to keep things simple enough for the voters (not to frighten them with the idea of ranking all 100 candidates). I.e. not theoretically ideal, but in practical situations ranking some candidates may be much better than ranking only one, or not bothering to vote at all.

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

Virus-free. www.avg.com
<a href="x-msg://6/#DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2" width="1" height="1" class="">


__._,_.___

Posted by: Chris Benham <[hidden email]>



__,_._,___


----
Election-Methods mailing list - see https://electorama.com/em for list info
Reply | Threaded
Open this post in threaded view
|

Re: [EM] The IRV-Disease has reached my town.

Richard Lung
In reply to this post by Andrew Myers
The late Dr David Hill invented an STV-Condorcet hybrid called Sequential STV, which uses Condorcet pairing to alleviate premature exclusion of candidates. Tho, this version of STV is still non-monotonic. (FAB STV is monotonic.) David Hill's great great great grandfather (if that's the right number of greats) was Thomas Wright Hill the first known inventor of the prototype of the single transferable vote, or the quota-preferential method, as the Australians call it. This year 2019 is its bi-centenary. (More information on the STV Action site of Anthony Tufffin, who promised his friend David, to commemorate the occasion.)

from Richard Lung.

On 05/03/2019 05:50, Andrew Myers wrote:
Chris Benham wrote:
If  you are a fan of the Condorcet criterion, then I think it is fine to modify IRV by before each elimination
we check for a Condorcet winner (among the so far remaining candidates) and when we find one we stop
and declare that candidate the winner.
This is one of the algorithms supported by CIVS, the one called "Condorcet-IRV" though to my understanding it was originally proposed by Thomas Hill.

-- Andrew

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



----
Election-Methods mailing list - see https://electorama.com/em for list info
Reply | Threaded
Open this post in threaded view
|

Re: [EM] [ApprovalVoting] Re: The IRV-Disease has reached my town.

Kristofer Munsterhjelm-3
In reply to this post by robert bristow-johnson
On 05/03/2019 09.56, robert bristow-johnson wrote:

>
>
> ---------------------------- Original Message ----------------------------
> Subject: Re: [EM] [ApprovalVoting] Re: The IRV-Disease has reached my town.
> From: "Juho Laatu" <[hidden email]>
> Date: Tue, March 5, 2019 12:12 am
> To: "EM" <[hidden email]>
> --------------------------------------------------------------------------
>
>
>> P.S. I think the STV-BTR method that Robert proposed could make a lot
> of sense in societies where IRV way of thinking is strong.
>
> i can't take any credit for proposing that.  i dunno who thought of it
> first.  wasn't me.

That seems to have been Rob LeGrand:
http://lists.electorama.com/pipermail/election-methods-electorama.com/2005-August/081989.html

That's the first mention I can find of it (there under the name BTR-IRV)
on this list.
----
Election-Methods mailing list - see https://electorama.com/em for list info
Reply | Threaded
Open this post in threaded view
|

Re: [EM] [ApprovalVoting] Re: The IRV-Disease has reached my town.

Chris Benham-2
In reply to this post by Juho Laatu-4

46 A
44 B>C
10 C

The positional information is on the ballots.� You can throw it away by refusing to look at anything but a pairwise
matrix, in this case� A>B 46-44,� B>C 44-10, C>A 54-46 and then say "oh well, I suppose we just have to break this
cycle at its weakest link".

A is voted in top position, which in the case of ranked ballots I would interpret as being ranked below no other
candidate, in this case strictly above all other candidates, on more ballots than B is ranked above bottom on any
ballots.

Plus of course A pairwise beats B.�� Given that A is so obviously dominant over B, the contention that "B might be
an ok winner" is absurd and not serious.� What shred of sane common-sense explanation could you possibly give
to the post-election complaining A supporters as to why their candidate should have been beaten by B ??!

does the "positional information" maybe refer to some Borda like rating information? (or maybe implicit ratings)

I infer some rating from rankings thus: candidates ranked above at least one candidate and below no candidate are
voted in the "top (or first) position", candidates not in the top position but also ranked above at least one other candidate
and below no candidate except those in top position are voted in the "second-from-the-top position" and so on
down to those candidates who are voted above no other candidate are in the "bottom position".

I consider that A "positionally dominates" B if A� has more first position votes and more first plus second position votes
and so on down to more above bottom votes.

If A wins, B and C supporters (54, majority) clearly think that C would have been a better choice (and 44 voters would prefer B).
In the scenario as I described it the B voters don't really care about C, they were just using C to benefit from Margins' gross failure
of the Later-no-Help criterion.

The A supporters have an undeniable complaint against B but not C.�

The B supporters also have no reasonable complaint against C. They all voted (above bottom) for C and C is voted above bottom
on more ballots than B (or A) and C� has a beatpath to B.� They can't claim to have been stung by a failure of Later-no-Harm
because clearly if they had� truncated then A would have won.

The C supporters have some� possible (relatively weak) complaint against A. But you want to elect the only candidate whose
supporters can have no remotely reasonable complaint against the election of either of the other candidates.

It is for example possible that there is an election with 5 serious candidates (based on some good polling information) and 50 other candidates with no chances to win. In that situation it would not be a big problem to limit the number of ranked candidates to say 7
If I sincerely prefer 7 minor candidates before any of the "serious" ones, how is this method better for me than plurality?� I will either
have to vote insincerely or put up with my vote having no effect on the result.

If Ireland and Australia have no problem allowing voters to fully rank then why should the US?

Chris Benham

On 6/03/2019 4:31 am, Juho Laatu wrote:
On 05 Mar 2019, at 16:04, Chris Benham [hidden email] [ApprovalVoting] <[hidden email]> wrote:

Also here you assume that there is an implicit approval cutoff after the ranked candidates. What if the votes are sincere and there is no implicit approval cutoff?

No, I'm only "assuming" that positional information is more meaningful than which candidate is "closer" to being the Condorcet winner
according to Margins or which candidate needs the fewest additional bullet-votes to become the CW.

If the added implicit information is not approval of all the ranked candidates, does the "positional information" maybe refer to some Borda like rating information? (or maybe implicit ratings) (I don't really like Borda ratings as additional information either because of the associated nomination related problems.)



Given that Margins is very vulnerable to Burial strategy the argument that it's worth putting up with that because with sincere votes B
in this scenario is the best (or a good or even acceptable) candidate is .. what??

Or to put it another way, assuming the votes are sincere, you arguments against electing A or C are what?

I just thought that B might be ok if we take the votes to be plain rankings (= pairwise preferences) with no additional approval or positional information assumed. The mentioned votes are of course very extreme (only three kind if voters), so they are not a typical set of votes form a real life election.

46 A
44 B>C
10 C

One could think that A is a left wing candidate, C is moderate right, and B is far right (because there are 44 voters that rank them in such linear order). A voters don't seem to care which one of the right wing candidates wins (they are tied in every single vote). The B voters have a clear and understandable position, with full rankings. The C voters have not ranked B. Maybe they are so centrist that A and B are equally good to them.

Condorcet methods can be said to elect a compromise candidate that is not too bad for anyone (sometimes the Condorcet winner might have no first preference votes). I therefore try to find an explanation to why B might win, from this point of view (= elect he best compromise (that is not very disliked)).

Since the A voters did not rank C above B, we must assume that they are perfectly ok with electing B, if A does not win. Same with C voters. B voters have a clear preference C>A (if B can not win).

If A wins, B and C supporters (54, majority) clearly think that C would have been a better choice (and 44 voters would prefer B). If C wins, 46 voters would prefer A, and 44 voters would prefer B. If B wins, 46 voters would prefer A, and 10 voters would prefer C. B doesn't look too bad in this comparison. B might be a better compromise than A or C. I.e. less complaints and rebellions after the election.

(I tried to avoid the "few votes short of being a Condorcet winner" argument since you might not appreciate it.)


One might face problems sooner with sincere voting than with strategic voting.

First preferences could be as follows.

30: far-left
21: left
19: right
30: far-right

If we assume that left hates right, and right hates left, the natural approval limit would be between the left wing and right wing parties. We would get mostly votes that rank only left wing or only right wing candidates. And the winner would be with good probability the far-left candidate, not the expected Condorcet winner (left).

That doesn't bother me much because (a) far-left may be higher "Social Utility" than left

I don't know what "Social Utility" means here. I guess any of the candidates could have that. It might not show up in rankings nor in approvals.

and (b) probably enough right voters would
be aware that the result is unlikely to be decided by Approval and so they would not be taking a huge risk by sincerely ranking left
over far-left.

I guess this depends on the method. I can't tell if all "ranking + implicit approval" methods would behave well in this respect. I also don't like very much the idea of voters having to cast strategic votes instead of sincere votes (i.e. implicit approvals in the ballots would not mean that the voter would approve the candidate, but something strategic instead).


But having said that, Smith//Approval using ballots that� allow voters to rank among candidates they don't approve would not be
in my book too bad (and much better than Margins).

P.S. I think the STV-BTR method that Robert proposed could make a lot of sense in societies where IRV way of thinking is strong.
That method doesn't have good criterion compliances. It's just a gimmick to smuggle Condorcet compliance past IRV enthusiasts.
The alternative of just checking for a CW (among remaining candidates) before each elimination is much better.

I'm not a big believer in criterion compliance in real life election methods. In theoretical studies different criteria are excellent measurement tools, but in real life elections nobody cares if the method performs well in some theoretical situations. Often slightly modified (heuristic style, not necessarily "mathematically clean") methods are close enough to meeting those criteria on practice anyway. Also my theoretically ideal "maximally strategy resistant method" might be one that fails to meet most of the named criteria, but does so intentionally in order to violate each one (or many) of them just a little bit, so that it can maximise resistance against all kind of strategies (and keep its worst vulnerability least bad).


P.P.S. Limiting the number of ranking levels or number of ranked candidates could make sense when the number of candidates is very high, or just to keep things simple for the vote counting process, or to keep things simple enough for the voters (not to frighten them with the idea of ranking all 100 candidates). I.e. not theoretically ideal, but in practical situations ranking some candidates may be much better than ranking only one, or not bothering to vote at all.

Limiting the number of candidates the voter is allowed to rank makes no sense. What has happened to your concern
about "removing information" on who the sincere/ "expected" CW is?

Typically the intention is not to limit the number of candidates that can be ranked (when compared to the situation before he change) but to make that number higher, while allowing it not to be very high. I am still worried about removing information, but I can accept some limitations sometimes (when full ranking is not feasible, or when most voters would not rank all candidates anyway, or when other solutions are not politically possible). The limits should be such that they probably do not lead to not electing the sincere Condorcet winner (or the best candidate when there is no Condorcet winner).

It is for example possible that there is an election with 5 serious candidates (based on some good polling information) and 50 other candidates with no chances to win. In that situation it would not be a big problem to limit the number of ranked candidates to say 7. The ballots could be simpler that way, and voting would not be too tedious. I would not mind someone starting even from 3, if that is an improvement e.g. to the earlier FPTP.


There would be nothing "frightening" about ranking all the candidates if doing so is purely optional. But voters who wish
to vote a full ranking should be allowed to.

That is possible, and a positive thing to do, but I do understand that sometimes also less perfect methods can be "perfect" or "sufficient" for the current real life situation.

Juho



Chris� Benham


On 5/03/2019 6:42 pm, Juho Laatu wrote:
On 05 Mar 2019, at 07:45, Chris Benham [hidden email] [ApprovalVoting] <[hidden email]> wrote:

Robert,



in a ranked ballot, what defines an "approved" candidate?� all unranked candidates are tied for last place on a ballot.� is any candidate that is ranked at all "approved"?
Yes.

that would change and complicate the meaning of the ranked ballot.

Arguably "change" somewhat but I don't see how (overly) "complicate". Allowing voters to rank among unapproved
candidates makes the method more vulnerable to strategy and a lot more complicated.

One might face problems sooner with sincere voting than with strategic voting.

First preferences could be as follows.

30: far-left
21: left
19: right
30: far-right

If we assume that left hates right, and right hates left, the natural approval limit would be between the left wing and right wing parties. We would get mostly votes that rank only left wing or only right wing candidates. And the winner would be with good probability the far-left candidate, not the expected Condorcet winner (left).

The problem with "implicit approval cutoff after the ranked candidates" is that voters would be encouraged not to rank all the major candidates. Not good for Condorcet. That would remove some important information. In this example the sincere Condorcet winner could not be identified anymore.

46 A
44 B>C
10 C

A>B 46-44 (margin=2)���� B>C 44-10 (margin=34)�� C>A 54-46 (margin=8)

Now Margins elects B,� rewarding the outrageous Burial strategy.

I can't tolerate any method that elects B in this scenario. Even assuming that all the votes are sincere,
B is clearly the weakest candidate (the least "approved" and positionally dominated and pairwise-beaten
by A.)

Also here you assume that there is an implicit approval cutoff after the ranked candidates. What if the votes are sincere and there is no implicit approval cutoff?

Juho


P.S. I think the STV-BTR method that Robert proposed could make a lot of sense in societies where IRV way of thinking is strong.

P.P.S. Limiting the number of ranking levels or number of ranked candidates could make sense when the number of candidates is very high, or just to keep things simple for the vote counting process, or to keep things simple enough for the voters (not to frighten them with the idea of ranking all 100 candidates). I.e. not theoretically ideal, but in practical situations ranking some candidates may be much better than ranking only one, or not bothering to vote at all.

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

Virus-free. www.avg.com
<a href="x-msg://6/#DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2" width="1" height="1" class="" moz-do-not-send="true">


__._,_.___

Posted by: Chris Benham <[hidden email]>


Yahoo! Groups
� Privacy � [hidden email] � Terms of Use

__,_._,___


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

----
Election-Methods mailing list - see https://electorama.com/em for list info
Reply | Threaded
Open this post in threaded view
|

Re: [EM] [RangeVoting] [ApprovalVoting] Re: The IRV-Disease has reached my town.

Juho Laatu-4

On 05 Mar 2019, at 23:07, Chris Benham [hidden email] [RangeVoting] <[hidden email]> wrote:



46 A
44 B>C
10 C

The positional information is on the ballots.� You can throw it away by refusing to look at anything but a pairwise
matrix, in this case� A>B 46-44,� B>C 44-10, C>A 54-46 and then say "oh well, I suppose we just have to break this
cycle at its weakest link".

I guess the basic assumption is that ranked methods are based on rankings. I used also the actual ballots when trying to explain what would be a natural outcome with those votes (although the method was expected to use only rankings / the matrix). I didn't assume any approvals or positional information to be extracted from the ballots.

I btw don't favour the approach of "breaking a cycle" since that approach seems to assume that there is something wrong with having a cycle, and that they should be eliminated to form a preference order without cycles. I rather tend to think that one should just elect the best candidate (no need to assume that cycles should be somehow broken or removed in the process).


A is voted in top position, which in the case of ranked ballots I would interpret as being ranked below no other
candidate, in this case strictly above all other candidates, on more ballots than B is ranked above bottom on any
ballots.


Ok, that's some additional "positional information" that you may extract from the ballots. My default assumption in ranked methods is that vote X>A>B>C (where A is not in top position) would be the same as A>B>C if X is a totally irrelevant candidate. But of course, also additional information (other than rankings) can be included in the method if one so wants.


Plus of course A pairwise beats B.�� Given that A is so obviously dominant over B, the contention that "B might be
an ok winner" is absurd and not serious.� What shred of sane common-sense explanation could you possibly give
to the post-election complaining A supporters as to why their candidate should have been beaten by B ??!


I tried to address this from one point of view in my previous post.


does the "positional information" maybe refer to some Borda like rating information? (or maybe implicit ratings)

I infer some rating from rankings thus: candidates ranked above at least one candidate and below no candidate are
voted in the "top (or first) position", candidates not in the top position but also ranked above at least one other candidate
and below no candidate except those in top position are voted in the "second-from-the-top position" and so on
down to those candidates who are voted above no other candidate are in the "bottom position".

I consider that A "positionally dominates" B if A� has more first position votes and more first plus second position votes
and so on down to more above bottom votes.

If A wins, B and C supporters (54, majority) clearly think that C would have been a better choice (and 44 voters would prefer B).
In the scenario as I described it the B voters don't really care about C, they were just using C to benefit from Margins' gross failure
of the Later-no-Help criterion.

Yes, you presented a strategic scenario, but my question was, what should we do if those same votes were sincere votes. In that case the method should pick the best candidate under the assumption that the votes are sincere. There could thus be another election with exactly similar votes, but where the votes are sincere.


The A supporters have an undeniable complaint against B but not C.�

The B supporters also have no reasonable complaint against C. They all voted (above bottom) for C and C is voted above bottom
on more ballots than B (or A) and C� has a beatpath to B.� They can't claim to have been stung by a failure of Later-no-Harm
because clearly if they had� truncated then A would have won.

The C supporters have some� possible (relatively weak) complaint against A. But you want to elect the only candidate whose
supporters can have no remotely reasonable complaint against the election of either of the other candidates.

It is for example possible that there is an election with 5 serious candidates (based on some good polling information) and 50 other candidates with no chances to win. In that situation it would not be a big problem to limit the number of ranked candidates to say 7
If I sincerely prefer 7 minor candidates before any of the "serious" ones, how is this method better for me than plurality?� I will either
have to vote insincerely or put up with my vote having no effect on the result.

In that case you should understand that ranking only the 7 minor candidates is not a good idea in that method. You should satisfy with ranking only 3 minor candidates.


If Ireland and Australia have no problem allowing voters to fully rank then why should the US?

Yes, I'm sure there are many elections where allowing full rankings is the best approach. (I have no knowledge of the US scenario.)

Juho



Chris Benham

On 6/03/2019 4:31 am, Juho Laatu wrote:
On 05 Mar 2019, at 16:04, Chris Benham [hidden email] [ApprovalVoting] <[hidden email]> wrote:

Also here you assume that there is an implicit approval cutoff after the ranked candidates. What if the votes are sincere and there is no implicit approval cutoff?

No, I'm only "assuming" that positional information is more meaningful than which candidate is "closer" to being the Condorcet winner
according to Margins or which candidate needs the fewest additional bullet-votes to become the CW.

If the added implicit information is not approval of all the ranked candidates, does the "positional information" maybe refer to some Borda like rating information? (or maybe implicit ratings) (I don't really like Borda ratings as additional information either because of the associated nomination related problems.)



Given that Margins is very vulnerable to Burial strategy the argument that it's worth putting up with that because with sincere votes B
in this scenario is the best (or a good or even acceptable) candidate is .. what??

Or to put it another way, assuming the votes are sincere, you arguments against electing A or C are what?

I just thought that B might be ok if we take the votes to be plain rankings (= pairwise preferences) with no additional approval or positional information assumed. The mentioned votes are of course very extreme (only three kind if voters), so they are not a typical set of votes form a real life election.

46 A
44 B>C
10 C

One could think that A is a left wing candidate, C is moderate right, and B is far right (because there are 44 voters that rank them in such linear order). A voters don't seem to care which one of the right wing candidates wins (they are tied in every single vote). The B voters have a clear and understandable position, with full rankings. The C voters have not ranked B. Maybe they are so centrist that A and B are equally good to them.

Condorcet methods can be said to elect a compromise candidate that is not too bad for anyone (sometimes the Condorcet winner might have no first preference votes). I therefore try to find an explanation to why B might win, from this point of view (= elect he best compromise (that is not very disliked)).

Since the A voters did not rank C above B, we must assume that they are perfectly ok with electing B, if A does not win. Same with C voters. B voters have a clear preference C>A (if B can not win).

If A wins, B and C supporters (54, majority) clearly think that C would have been a better choice (and 44 voters would prefer B). If C wins, 46 voters would prefer A, and 44 voters would prefer B. If B wins, 46 voters would prefer A, and 10 voters would prefer C. B doesn't look too bad in this comparison. B might be a better compromise than A or C. I.e. less complaints and rebellions after the election.

(I tried to avoid the "few votes short of being a Condorcet winner" argument since you might not appreciate it.)


One might face problems sooner with sincere voting than with strategic voting.

First preferences could be as follows.

30: far-left
21: left
19: right
30: far-right

If we assume that left hates right, and right hates left, the natural approval limit would be between the left wing and right wing parties. We would get mostly votes that rank only left wing or only right wing candidates. And the winner would be with good probability the far-left candidate, not the expected Condorcet winner (left).

That doesn't bother me much because (a) far-left may be higher "Social Utility" than left

I don't know what "Social Utility" means here. I guess any of the candidates could have that. It might not show up in rankings nor in approvals.

and (b) probably enough right voters would
be aware that the result is unlikely to be decided by Approval and so they would not be taking a huge risk by sincerely ranking left
over far-left.

I guess this depends on the method. I can't tell if all "ranking + implicit approval" methods would behave well in this respect. I also don't like very much the idea of voters having to cast strategic votes instead of sincere votes (i.e. implicit approvals in the ballots would not mean that the voter would approve the candidate, but something strategic instead).


But having said that, Smith//Approval using ballots that� allow voters to rank among candidates they don't approve would not be
in my book too bad (and much better than Margins).

P.S. I think the STV-BTR method that Robert proposed could make a lot of sense in societies where IRV way of thinking is strong.
That method doesn't have good criterion compliances. It's just a gimmick to smuggle Condorcet compliance past IRV enthusiasts.
The alternative of just checking for a CW (among remaining candidates) before each elimination is much better.

I'm not a big believer in criterion compliance in real life election methods. In theoretical studies different criteria are excellent measurement tools, but in real life elections nobody cares if the method performs well in some theoretical situations. Often slightly modified (heuristic style, not necessarily "mathematically clean") methods are close enough to meeting those criteria on practice anyway. Also my theoretically ideal "maximally strategy resistant method" might be one that fails to meet most of the named criteria, but does so intentionally in order to violate each one (or many) of them just a little bit, so that it can maximise resistance against all kind of strategies (and keep its worst vulnerability least bad).


P.P.S. Limiting the number of ranking levels or number of ranked candidates could make sense when the number of candidates is very high, or just to keep things simple for the vote counting process, or to keep things simple enough for the voters (not to frighten them with the idea of ranking all 100 candidates). I.e. not theoretically ideal, but in practical situations ranking some candidates may be much better than ranking only one, or not bothering to vote at all.

Limiting the number of candidates the voter is allowed to rank makes no sense. What has happened to your concern
about "removing information" on who the sincere/ "expected" CW is?

Typically the intention is not to limit the number of candidates that can be ranked (when compared to the situation before he change) but to make that number higher, while allowing it not to be very high. I am still worried about removing information, but I can accept some limitations sometimes (when full ranking is not feasible, or when most voters would not rank all candidates anyway, or when other solutions are not politically possible). The limits should be such that they probably do not lead to not electing the sincere Condorcet winner (or the best candidate when there is no Condorcet winner).

It is for example possible that there is an election with 5 serious candidates (based on some good polling information) and 50 other candidates with no chances to win. In that situation it would not be a big problem to limit the number of ranked candidates to say 7. The ballots could be simpler that way, and voting would not be too tedious. I would not mind someone starting even from 3, if that is an improvement e.g. to the earlier FPTP.


There would be nothing "frightening" about ranking all the candidates if doing so is purely optional. But voters who wish
to vote a full ranking should be allowed to.

That is possible, and a positive thing to do, but I do understand that sometimes also less perfect methods can be "perfect" or "sufficient" for the current real life situation.

Juho



Chris� Benham


On 5/03/2019 6:42 pm, Juho Laatu wrote:
On 05 Mar 2019, at 07:45, Chris Benham [hidden email] [ApprovalVoting] <[hidden email]> wrote:

Robert,



in a ranked ballot, what defines an "approved" candidate?� all unranked candidates are tied for last place on a ballot.� is any candidate that is ranked at all "approved"?
Yes.

that would change and complicate the meaning of the ranked ballot.

Arguably "change" somewhat but I don't see how (overly) "complicate". Allowing voters to rank among unapproved
candidates makes the method more vulnerable to strategy and a lot more complicated.

One might face problems sooner with sincere voting than with strategic voting.

First preferences could be as follows.

30: far-left
21: left
19: right
30: far-right

If we assume that left hates right, and right hates left, the natural approval limit would be between the left wing and right wing parties. We would get mostly votes that rank only left wing or only right wing candidates. And the winner would be with good probability the far-left candidate, not the expected Condorcet winner (left).

The problem with "implicit approval cutoff after the ranked candidates" is that voters would be encouraged not to rank all the major candidates. Not good for Condorcet. That would remove some important information. In this example the sincere Condorcet winner could not be identified anymore.

46 A
44 B>C
10 C

A>B 46-44 (margin=2)���� B>C 44-10 (margin=34)�� C>A 54-46 (margin=8)

Now Margins elects B,� rewarding the outrageous Burial strategy.

I can't tolerate any method that elects B in this scenario. Even assuming that all the votes are sincere,
B is clearly the weakest candidate (the least "approved" and positionally dominated and pairwise-beaten
by A.)

Also here you assume that there is an implicit approval cutoff after the ranked candidates. What if the votes are sincere and there is no implicit approval cutoff?

Juho


P.S. I think the STV-BTR method that Robert proposed could make a lot of sense in societies where IRV way of thinking is strong.

P.P.S. Limiting the number of ranking levels or number of ranked candidates could make sense when the number of candidates is very high, or just to keep things simple for the vote counting process, or to keep things simple enough for the voters (not to frighten them with the idea of ranking all 100 candidates). I.e. not theoretically ideal, but in practical situations ranking some candidates may be much better than ranking only one, or not bothering to vote at all.

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

Virus-free. www.avg.com
<a href="x-msg://6/#DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2" width="1" height="1" class="" moz-do-not-send="true">


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


__._,_.___

Posted by: Chris Benham <[hidden email]>



__,_._,___


----
Election-Methods mailing list - see https://electorama.com/em for list info
Reply | Threaded
Open this post in threaded view
|

[EM] Landau vs Banks confusion

Forest Simmons
In reply to this post by Chris Benham-2
Chris,

Not so quick; don't throw the baby out with the bath water!

Max Covering based on an approval list or any other monotonically generated list of the candidates elects a candidate monotonically from the set of uncovered candidates, i.e. the Landau Set, but not necessarily from the Banks set, because (in general) some Landau candidates may not be Banks candidates. 

All uncovered candidates are members of Landau.

But only candidates that stand at the head of (one or more) maximal totally ordered chains (in the defeat relation) qualify as members of the Banks set. 

We have several methods (including Copeland) that elect monotonically from Landau, but so far only one from Banks (chain climbing).

Copeland elects monotonically from Landau, but is no more clone independent than Borda.  As far as Landau methods go, the Max Covering schema is the only one I know of (besides chain climbing itself) which is both monotone and clone free

If chain climbing satisfied Independence from Pareto Dominated Alternatives (IPDA), I would be content to stop looking for a better Banks method.

Max Covering does satisfy the IPDA criterion, so perhaps some tweak on it might get us where we want. We would start with the covering relation which is a partial order on the set of candidates. Then we would single out some maximal chain in the covering relation. as in the Max Cover procedure.  Finally beef up this chain with additional candidates while relaxing the order relation from covering to mere defeat until it become a maximal totally ordered chain with respect to the defeat relation.

In the resulting chain every candidate would defeat all of the candidates below it and would be defeated by all of the candidates above it, and no other candidate could be included into the chain without destroying this total order property.  The candidate at the head of the chain would beat all of the other candidates in the chain.  And since that chain was built up to be maximal with respect to the defeat relation that candidate would be a member of the Banks set.

Can this be done monotonically?  The devil is in the details.

Our recent attempt got pretty close, but did not (to our disappointment) actually satisfy monotonicity.as it first appeared.

Back to the drawiing board!




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