Hello.
Well, I tried, I really did. I read the wikipedia entries. In English *and* in French. But I still don't understand the Schulze method. I mean, I kind of understand that it's good (TM) and that it probably will end up choosing the best person for the position. But I don't understand the implications of what I vote and how I vote for some things.
I am hoping that someone can make this clearer to me.
So here are my questions:
The explanation says (and I quote): "You may give the same preference to more than one candidate and may keep candidates unranked. It is presumed that you prefer all ranked candidates to all not ranked candidates and that you are indifferent between all not ranked candidates."
#Question 1 Does "you may give the same preference to more than one candidate" mean that I can rank three candidates with rank 1, three with rank 2 , one with rank 3 and five with rank 4 (and so forth)?
#Question 2 Can I actually rank one candidate with rank 1, three candidates with rank 2 and 5 candidates with rank 15? That is, does the rank (1, 2, 3 etc.) actually matter in the overall results, or is rank always relative? (ie. If I rank 2 people with rank 1 and 10 with rank 15, the 10 will be counted as being my second choice, not as being "of rank 15")
#Question 3 What's the best way to go about making sure that a candidate is ranked as low as possible? Rank them at the lowest possible rank (this will of course depend on answers to question 2)? Or not rank them at all?
I am not sure that my questions are clear. I hope so :-)
Thank you for your help.
Delphine
Hi Delphine,
Your questions make sense and I hope my answers will be of help:
Yes, you can rank any number of candidates any way you like. What matters is the '''relative order''' of the numbers you assign to the candidates, as the Schulze method only cares about relative rank. The numbers are only a way for you to express this relative order of ranking. For example, giving the candidate you dislike the most 50 or 173 does not matter, as long as you assign better (smaller) grades to the others. Unranked candidates are implicitly assumed to be less favored than all the ranked ones, and the system assigns them the rank 100 (or was it 99?).
Harel
On Fri, Jun 6, 2008 at 5:36 PM, Delphine Ménard notafishz@gmail.com wrote:
Hello.
Well, I tried, I really did. I read the wikipedia entries. In English *and* in French. But I still don't understand the Schulze method. I mean, I kind of understand that it's good (TM) and that it probably will end up choosing the best person for the position. But I don't understand the implications of what I vote and how I vote for some things.
I am hoping that someone can make this clearer to me.
So here are my questions:
The explanation says (and I quote): "You may give the same preference to more than one candidate and may keep candidates unranked. It is presumed that you prefer all ranked candidates to all not ranked candidates and that you are indifferent between all not ranked candidates."
#Question 1 Does "you may give the same preference to more than one candidate" mean that I can rank three candidates with rank 1, three with rank 2 , one with rank 3 and five with rank 4 (and so forth)?
#Question 2 Can I actually rank one candidate with rank 1, three candidates with rank 2 and 5 candidates with rank 15? That is, does the rank (1, 2, 3 etc.) actually matter in the overall results, or is rank always relative? (ie. If I rank 2 people with rank 1 and 10 with rank 15, the 10 will be counted as being my second choice, not as being "of rank 15")
#Question 3 What's the best way to go about making sure that a candidate is ranked as low as possible? Rank them at the lowest possible rank (this will of course depend on answers to question 2)? Or not rank them at all?
I am not sure that my questions are clear. I hope so :-)
Thank you for your help.
Delphine
~notafish
NB. This gmail address is used for mailing lists. For Wikimedia related correspondence, use my dmenard(at)wikimedia(point)org address. http://blog.notanendive.org
foundation-l mailing list foundation-l@lists.wikimedia.org Unsubscribe: https://lists.wikimedia.org/mailman/listinfo/foundation-l
Harel appropriately answers your questions, Delphine. The only thing that I would clarify is that the system does, indeed, rank "unranked" candidates with a 100 - they are assumed to be the least favored.
To be clear: you may rank as many candidates as you want with a 1, whether that's three candidates or one candidate. Your rankings do not need to be numerically ordered (you don't need to do 1, 2, 3, 4, 5.... you can do 1, 15, 27, 32, and unranked, if you want).
Philippe
-------------------------------------------------- From: "Harel Cain" harel.cain@gmail.com Sent: Friday, June 06, 2008 9:49 AM To: "Wikimedia Foundation Mailing List" foundation-l@lists.wikimedia.org Subject: Re: [Foundation-l] Board vote, need a bit of help
Hi Delphine,
Your questions make sense and I hope my answers will be of help:
Yes, you can rank any number of candidates any way you like. What matters is the '''relative order''' of the numbers you assign to the candidates, as the Schulze method only cares about relative rank. The numbers are only a way for you to express this relative order of ranking. For example, giving the candidate you dislike the most 50 or 173 does not matter, as long as you assign better (smaller) grades to the others. Unranked candidates are implicitly assumed to be less favored than all the ranked ones, and the system assigns them the rank 100 (or was it 99?).
Harel
On Fri, Jun 6, 2008 at 5:36 PM, Delphine Ménard notafishz@gmail.com wrote:
Hello.
Well, I tried, I really did. I read the wikipedia entries. In English *and* in French. But I still don't understand the Schulze method. I mean, I kind of understand that it's good (TM) and that it probably will end up choosing the best person for the position. But I don't understand the implications of what I vote and how I vote for some things.
I am hoping that someone can make this clearer to me.
So here are my questions:
The explanation says (and I quote): "You may give the same preference to more than one candidate and may keep candidates unranked. It is presumed that you prefer all ranked candidates to all not ranked candidates and that you are indifferent between all not ranked candidates."
#Question 1 Does "you may give the same preference to more than one candidate" mean that I can rank three candidates with rank 1, three with rank 2 , one with rank 3 and five with rank 4 (and so forth)?
#Question 2 Can I actually rank one candidate with rank 1, three candidates with rank 2 and 5 candidates with rank 15? That is, does the rank (1, 2, 3 etc.) actually matter in the overall results, or is rank always relative? (ie. If I rank 2 people with rank 1 and 10 with rank 15, the 10 will be counted as being my second choice, not as being "of rank 15")
#Question 3 What's the best way to go about making sure that a candidate is ranked as low as possible? Rank them at the lowest possible rank (this will of course depend on answers to question 2)? Or not rank them at all?
I am not sure that my questions are clear. I hope so :-)
Thank you for your help.
Delphine
~notafish
NB. This gmail address is used for mailing lists. For Wikimedia related correspondence, use my dmenard(at)wikimedia(point)org address. http://blog.notanendive.org
foundation-l mailing list foundation-l@lists.wikimedia.org Unsubscribe: https://lists.wikimedia.org/mailman/listinfo/foundation-l
-- Quidquid latine dictum sit, altum viditur.
foundation-l mailing list foundation-l@lists.wikimedia.org Unsubscribe: https://lists.wikimedia.org/mailman/listinfo/foundation-l
On Fri, Jun 6, 2008 at 9:45 AM, Philippe Beaudette philippebeaudette@gmail.com wrote:
Harel appropriately answers your questions, Delphine. The only thing that I would clarify is that the system does, indeed, rank "unranked" candidates with a 100 - they are assumed to be the least favored.
To be clear: you may rank as many candidates as you want with a 1, whether that's three candidates or one candidate. Your rankings do not need to be numerically ordered (you don't need to do 1, 2, 3, 4, 5.... you can do 1, 15, 27, 32, and unranked, if you want).
Philippe
But 1, 2, 3, 4, 5 and 1, 15, 27, 32, unranked are equivalent as far as the voting software is concerned? I didn't understand this part either.
I might have to revote :) -- phoebe
On Fri, 2008-06-06 at 10:24 -0700, phoebe ayers wrote:
But 1, 2, 3, 4, 5 and 1, 15, 27, 32, unranked are equivalent as far as the voting software is concerned? I didn't understand this part either.
I might have to revote :) -- phoebe
Assuming there are 5 and only 5 candidates in your example, yes they are the same.
KTC
is there already an article about this voting system on simple: ? Would someone who really understands it please do that? That would be of great help I think. Maybe an addendum on "if I want this and this to happen, how should I vote?" would be useful.
br, lodewijk
2008/6/6 Kwan Ting Chan ktc@ktchan.info:
On Fri, 2008-06-06 at 10:24 -0700, phoebe ayers wrote:
But 1, 2, 3, 4, 5 and 1, 15, 27, 32, unranked are equivalent as far as the voting software is concerned? I didn't understand this part either.
I might have to revote :) -- phoebe
Assuming there are 5 and only 5 candidates in your example, yes they are the same.
KTC
-- Experience is a good school but the fees are high.
- Heinrich Heine
foundation-l mailing list foundation-l@lists.wikimedia.org Unsubscribe: https://lists.wikimedia.org/mailman/listinfo/foundation-l
Philippe Beaudette wrote:
Harel appropriately answers your questions, Delphine. The only thing that I would clarify is that the system does, indeed, rank "unranked" candidates with a 100 - they are assumed to be the least favored.
That prompts a question. Suppose I happened to rank one candidate at 1, one candidate at 2, one candidate at 101, one candidate at 102, one candidate at 201, one candidate at 202, and so on, but left other candidates unranked. Does the system allow these numbers? If the default for unranked candidates is 100, is the software still able to correctly sort out the order in which I preferred the candidates?
--Michael Snow
On Fri, 2008-06-06 at 17:36 -0700, Michael Snow wrote:
Philippe Beaudette wrote:
Harel appropriately answers your questions, Delphine. The only thing that I would clarify is that the system does, indeed, rank "unranked" candidates with a 100 - they are assumed to be the least favored.
That prompts a question. Suppose I happened to rank one candidate at 1, one candidate at 2, one candidate at 101, one candidate at 102, one candidate at 201, one candidate at 202, and so on, but left other candidates unranked. Does the system allow these numbers? If the default for unranked candidates is 100, is the software still able to correctly sort out the order in which I preferred the candidates?
The system only allow rankings between 1 and 99 inclusive, by both limiting a user with only being allowed to enter 2 characters with HTML form attribute and by post submission checking for those cheeky enough to deliberately disable HTML form limits.
If someone submit the form with anything other than the number 1 to 99 (with no leading zero for 1 to 9) in any of the fields, then it return the user to the voting page with an error message at the top.
Regards,
KTC
on 6/6/08 10:36 AM, Delphine Ménard at notafishz@gmail.com wrote:
Hello.
Well, I tried, I really did.
And so did I. I went to the Wikipedia: Board of Trustees election Page, clicked on the "Click here" and got this message: "Security Failure. Data decryption error".
Help!
Marc Riddell
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
I voted in Palm Beach County and I think I voted for Pat Robertson when I tried to vote for Dan Rosenthal.
Cary
Delphine Ménard wrote: | Hello. | | Well, I tried, I really did. I read the wikipedia entries. In English | *and* in French. But I still don't understand the Schulze method. I | mean, I kind of understand that it's good (TM) and that it probably | will end up choosing the best person for the position. But I don't | understand the implications of what I vote and how I vote for some | things. | | I am hoping that someone can make this clearer to me. | | So here are my questions: | | The explanation says (and I quote): "You may give the same preference | to more than one candidate and may keep candidates unranked. It is | presumed that you prefer all ranked candidates to all not ranked | candidates and that you are indifferent between all not ranked | candidates." | | #Question 1 | Does "you may give the same preference to more than one candidate" | mean that I can rank three candidates with rank 1, three with rank 2 , | one with rank 3 and five with rank 4 (and so forth)? | | #Question 2 | Can I actually rank one candidate with rank 1, three candidates with | rank 2 and 5 candidates with rank 15? That is, does the rank (1, 2, 3 | etc.) actually matter in the overall results, or is rank always | relative? (ie. If I rank 2 people with rank 1 and 10 with rank 15, the | 10 will be counted as being my second choice, not as being "of rank | 15") | | #Question 3 | What's the best way to go about making sure that a candidate is ranked | as low as possible? Rank them at the lowest possible rank (this will | of course depend on answers to question 2)? Or not rank them at all? | | I am not sure that my questions are clear. I hope so :-) | | Thank you for your help. | | Delphine
I've spent the entire day driving minorities to polling stations!
On 06/06/2008, Cary Bass cbass@wikimedia.org wrote:
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
I voted in Palm Beach County and I think I voted for Pat Robertson when I tried to vote for Dan Rosenthal.
Cary
Delphine Ménard wrote: | Hello. | | Well, I tried, I really did. I read the wikipedia entries. In English | *and* in French. But I still don't understand the Schulze method. I | mean, I kind of understand that it's good (TM) and that it probably | will end up choosing the best person for the position. But I don't | understand the implications of what I vote and how I vote for some | things. | | I am hoping that someone can make this clearer to me. | | So here are my questions: | | The explanation says (and I quote): "You may give the same preference | to more than one candidate and may keep candidates unranked. It is | presumed that you prefer all ranked candidates to all not ranked | candidates and that you are indifferent between all not ranked | candidates." | | #Question 1 | Does "you may give the same preference to more than one candidate" | mean that I can rank three candidates with rank 1, three with rank 2 , | one with rank 3 and five with rank 4 (and so forth)? | | #Question 2 | Can I actually rank one candidate with rank 1, three candidates with | rank 2 and 5 candidates with rank 15? That is, does the rank (1, 2, 3 | etc.) actually matter in the overall results, or is rank always | relative? (ie. If I rank 2 people with rank 1 and 10 with rank 15, the | 10 will be counted as being my second choice, not as being "of rank | 15") | | #Question 3 | What's the best way to go about making sure that a candidate is ranked | as low as possible? Rank them at the lowest possible rank (this will | of course depend on answers to question 2)? Or not rank them at all? | | I am not sure that my questions are clear. I hope so :-) | | Thank you for your help. | | Delphine -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.8 (Darwin) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org
iEYEARECAAYFAkhJggEACgkQyQg4JSymDYkyAQCfQAqwHxXSlpM9lqz3PR9LEwUM /P8AoLkYhlgBOCRtBFNsp7EAg84p8t72 =SyAn -----END PGP SIGNATURE-----
foundation-l mailing list foundation-l@lists.wikimedia.org Unsubscribe: https://lists.wikimedia.org/mailman/listinfo/foundation-l
On Fri, Jun 6, 2008 at 10:36 AM, Delphine Ménard notafishz@gmail.com wrote:
Hello.
Well, I tried, I really did. I read the wikipedia entries. In English *and* in French. But I still don't understand the Schulze method. I mean, I kind of understand that it's good (TM)
No one asked me, but if they did I would have supported Schulze because it has the cloning resistance property. A clone is a candidate whos platform is nearly indistinguishable from some other candidate.
Wikipedia says "In some (voting) systems, the introduction of a clone will tend to divide support between the similar candidates, making it less likely either will be elected. In some other systems, the presence of a clone will tend to reduce support for dissimilar candidates, making it more likely that one (or more) of the similar candidates will be elected. In yet other systems, the introduction of clones will not significantly affect the chances that one of the similar candidates will be elected." (http://en.wikipedia.org/wiki/Independence_of_clones_criterion)
I think cloning resistance is highly relevant
Otherwise, I generally don't like condorcet methods for these kinds of elections. (the normal failure to meet later-no-harm and invulnerability to burying being my primary concerns with most condorcet methods)
Approval voting is generally claimed to meet the Independence of clones criterion, but in most approval elections I've seen (Including Wikimedia's), many people vote insincerely/strategically and approve only a single person just like a plurality vote ... so ultimately cloning results in spoilage.
and that it probably will end up choosing the best person for the position. But I don't understand the implications of what I vote and how I vote for some things.
Hey! At least the pages now say WMF is using Schulze, rather than still saying it hasn't been selected yet!
#Question 1 Does "you may give the same preference to more than one candidate" mean that I can rank three candidates with rank 1, three with rank 2 , one with rank 3 and five with rank 4 (and so forth)?
Yes. Exactly.
When you rank two candidates the same you are not expressing a preference between them, but are expressing that both are strictly preferred to all the less prefered (higher rank number) candidates.
#Question 2 Can I actually rank one candidate with rank 1, three candidates with rank 2 and 5 candidates with rank 15? That is, does the rank (1, 2, 3 etc.) actually matter in the overall results, or is rank always relative? (ie. If I rank 2 people with rank 1 and 10 with rank 15, the 10 will be counted as being my second choice, not as being "of rank 15")
The rank is relative. Your ballot is effectively converted into something like: Jane > John, Bob, Tom > Sue, Mark, Jose, Igor, Lynn
#Question 3 What's the best way to go about making sure that a candidate is ranked as low as possible? Rank them at the lowest possible rank (this will of course depend on answers to question 2)? Or not rank them at all?
If there are 15 candidates and you want to make sure that one is ranked worse than all others you will have to provide a rank for AT LEAST the 14 others. You could leave Mr. Evil unranked, or you could simply rank them all and give him a worse rank than anyone else.
This is because all the unranked are treated as ranked with a high number which is equal for all unranked.
On Fri, Jun 6, 2008 at 3:43 PM, Gregory Maxwell gmaxwell@gmail.com wrote:
I think cloning resistance is highly relevant
Otherwise, I generally don't like condorcet methods for these kinds of elections. (the normal failure to meet later-no-harm and invulnerability to burying being my primary concerns with most condorcet methods)
Personally I see an argument that a voluntary organization has much less of a need to pick a compromise candidate than a government. If you're picking the capital of Tennessee, fine, you pick something in the middle. But change it to a vote for a Wikimedia board member and it might be better to fully satisfy as many people as possible and let those with radically different viewpoints fork off their own projects than to dilute the core project principles with compromise. I'm not sure if I agree with this or not, though.
I agree that cloning resistance is very important, though, so plurality voting seems out. Why not IRV?
Approval voting is generally claimed to meet the Independence of clones criterion, but in most approval elections I've seen (Including Wikimedia's), many people vote insincerely/strategically and approve only a single person just like a plurality vote ... so ultimately cloning results in spoilage.
Well, technically, if your two top candidates are really *identical*, there would be no strategic reason to not approve of both. Back to reality, though, it could be that people are voting strategically, or it could be that they just don't understand what they're doing. If the latter, this could happen with the Schulze method as well. People might rank one candidate "1" and leave all the rest blank. There wouldn't be a strategic reason to do this (would there?), but it still might happen.
I'm interested in whether or not we'll even know if this happened. How much data are we going to be given about the votes?
On Fri, Jun 6, 2008 at 11:02 PM, Anthony wikimail@inbox.org wrote:
I agree that cloning resistance is very important, though, so plurality voting seems out. Why not IRV?
Hmm...IRV would pick the "compromise candidate" in the Tennessee capital election too, I guess...
Oh well.
On Fri, Jun 6, 2008 at 11:02 PM, Anthony wikimail@inbox.org wrote:
Personally I see an argument that a voluntary organization has much less of a need to pick a compromise candidate than a government. If you're picking the capital of Tennessee, fine, you pick something in the middle. But change it to a vote for a Wikimedia board member and it might be better to fully satisfy as many people as possible and let those with radically different viewpoints fork off their own projects than to dilute the core project principles with compromise. I'm not sure if I agree with this or not, though.
I don't think thats an unreasonable argument. Not quite sure if I share it, have to think about it more. ;) For one forking doesn't seem like a good solution for splits on details (details are important, but don't justify forks...)
I agree that cloning resistance is very important, though, so plurality voting seems out. Why not IRV?
My preference for this kind of selection is IRV.
In my view IRV's weakness for this sort use is distortion from people overrating hoping to cause earlier round elimination of their least liked choice. (I think this is a worse problem then the underrating of second choices usually done against cordeset methods in elections where there are usually a few reasonable choices)
Well, technically, if your two top candidates are really *identical*, there would be no strategic reason to not approve of both.
Sure, but not everyone will see two candidates to be clones. Some people see A/B the same, approve both. Others see A and B as largely similar but randomly prefer one or the other (perhaps they just want to have a favorite), dishonestly underrate their less preferred alternative in the hopes of helping their (slightly) preferred candidate win.. votes end up split and C wins.
Back to reality, though, it could be that people are voting strategically, or it could be that they just don't understand what they're doing.
It's very true.
If the latter, this could happen with the Schulze method as well. People might rank one candidate "1" and leave all the rest blank. There wouldn't be a strategic reason to do this (would there?), but it still might happen.
I'd love to see stats on that. My bet is that a many will rank only a few, because of lacking patience, or a belief that they are helping their primary choices win like on an approval vote. ::shrugs::
One improvement might be to force people to rank all the candidates (not uniquely rank, just rank). I'm sure some would give up voting if they couldn't just put a "1" next to their favorite, but would biasing the election towards the choices of people who are more patient be all that bad? ;)
I'm interested in whether or not we'll even know if this happened. How much data are we going to be given about the votes?
In the past information on how many were approved was released. I don't see why the complete decrypted ballots (which contain no personal information) couldn't be released. Especially since we'll have ranked data it would be interesting to see how different counting systems (IRV, approval of all ranked, approval of all 1s, etc) might have influenced the results... and it's pretty normal for the anonymous ballots in an election to be not treated as secret.
On Sat, Jun 7, 2008 at 1:57 AM, Gregory Maxwell gmaxwell@gmail.com wrote:
In the past information on how many were approved was released. I don't see why the complete decrypted ballots (which contain no personal information) couldn't be released. Especially since we'll have ranked data it would be interesting to see how different counting systems (IRV, approval of all ranked, approval of all 1s, etc) might have influenced the results... and it's pretty normal for the anonymous ballots in an election to be not treated as secret.
I think this information would be very, very interesting, so long as it was obviously completely independent of vote time (i.e. the order is randomized, not ranked by the order that the votes were made in). This is a minor thing to do, but one that might be overlooked.
On Sat, Jun 7, 2008 at 1:57 AM, Gregory Maxwell gmaxwell@gmail.com wrote:
On Fri, Jun 6, 2008 at 11:02 PM, Anthony wikimail@inbox.org wrote:
If the latter, this could happen with the Schulze method as well. People might rank one candidate "1" and leave all the rest blank. There wouldn't be a strategic reason to do this (would there?), but it still might happen.
I'd love to see stats on that. My bet is that a many will rank only a few, because of lacking patience, or a belief that they are helping their primary choices win like on an approval vote. ::shrugs::
One improvement might be to force people to rank all the candidates (not uniquely rank, just rank). I'm sure some would give up voting if they couldn't just put a "1" next to their favorite, but would biasing the election towards the choices of people who are more patient be all that bad? ;)
I dunno. Is it really worth it to waste thousands of person-hours having each voter ponder the order of their 5th and 6th choice, neither of whom they've heard of prior to the election? How much useful stuff could get done during that time? Yes, I realize you said "not uniquely rank, just rank", but you also seem to suggest that spending lots of time on this process is a good thing.
I wish I could somehow get more of a sense of which 3 or 4 candidates actually have a good chance of winning. I might spend a little more time thinking about these candidates if I did. As is, I really have no clue who's going to win.
--- "Wikipedia seems extremely inscrutable to outsiders. You're sort of entirely in the world or you're not, and there's very few of us I think on the outside who know where to begin to wade in. There's 85 redundant channels for certain things to get discussed, so that's the challenge it presents to me, but I don't know if I have any good advice on what should be done about it." - Brian Bergstein, Not the Wikipedia Weekly, episode #11
On Sat, 2008-06-07 at 01:57 -0400, Gregory Maxwell wrote:
If the latter, this could happen with the Schulze method as well. People might rank one candidate "1" and leave all the rest blank. There wouldn't be a strategic reason to do this (would there?), but it still might happen.
I'd love to see stats on that. My bet is that a many will rank only a few, because of lacking patience, or a belief that they are helping their primary choices win like on an approval vote. ::shrugs::
One improvement might be to force people to rank all the candidates (not uniquely rank, just rank). I'm sure some would give up voting if they couldn't just put a "1" next to their favorite, but would biasing the election towards the choices of people who are more patient be all that bad? ;)
And what you'll end up with is lots of people putting 2, 2, 2, 2 (or whatever number), or even worse, random numbering into the rest just so they could get out of there.
I'm interested in whether or not we'll even know if this happened. How much data are we going to be given about the votes?
In the past information on how many were approved was released. I don't see why the complete decrypted ballots (which contain no personal information) couldn't be released. Especially since we'll have ranked data it would be interesting to see how different counting systems (IRV, approval of all ranked, approval of all 1s, etc) might have influenced the results... and it's pretty normal for the anonymous ballots in an election to be not treated as secret.
It will be.
On Fri, Jun 6, 2008 at 10:57 PM, Gregory Maxwell gmaxwell@gmail.com wrote:
If the latter, this could happen with the Schulze method as well. People might rank one candidate "1" and leave all the rest blank. There wouldn't be a strategic reason to do this (would there?), but it still might happen.
I'd love to see stats on that. My bet is that a many will rank only a few, because of lacking patience, or a belief that they are helping their primary choices win like on an approval vote. ::shrugs::
One improvement might be to force people to rank all the candidates (not uniquely rank, just rank). I'm sure some would give up voting if they couldn't just put a "1" next to their favorite, but would biasing the election towards the choices of people who are more patient be all that bad? ;)
So I ranked the people I had an opinion about, and then for the remaining candidates that I do not particularly want to be board members, I didn't try to figure out their relative position to one another but just left them unranked, since I thought that would mean they are all ranked as "equally low preference -- do not want" by the software. The directions sort of indicated that is what to do.
However, you're saying that it's better if all choices are ranked, even if I give several people a "99" by hand? Or should I take the time to uniquely rank them? I'm not sure I understand why that's better given the outcome I want, which is for none of those people to become board members.
It will be interesting to see if a bunch of foundation-l readers revote after this thread :) -- phoebe
On Sat, Jun 7, 2008 at 3:03 PM, phoebe ayers phoebe.wiki@gmail.com wrote:
So I ranked the people I had an opinion about, and then for the remaining candidates that I do not particularly want to be board members, I didn't try to figure out their relative position to one another but just left them unranked, since I thought that would mean they are all ranked as "equally low preference -- do not want" by the software. The directions sort of indicated that is what to do.
However, you're saying that it's better if all choices are ranked, even if I give several people a "99" by hand? Or should I take the time to uniquely rank them? I'm not sure I understand why that's better given the outcome I want, which is for none of those people to become board members.
It will be interesting to see if a bunch of foundation-l readers revote after this thread :) -- phoebe
If you rank everyone, giving them a "99" by hand is no different than leaving them unranked (which is, for all intents and purposes, giving them a "100")
Uniquely ranking the candidates you don't like doesn't help or hurt them with respect to the candidates you do like. All that does is say that you prefer "Unlikable Candidate A" over "Unlikable Candidate B", and if your likable candidate is not elected, you're saying that you'd want A before B.
Essentially, the point is that if you have a preference, even a slight one, you should rank your choices separately. If you make no distinction between the unfavorable candidates, you don't have to rank them, but ranking won't hurt the candidates you do like.
On Sat, Jun 7, 2008 at 3:03 PM, phoebe ayers phoebe.wiki@gmail.com wrote:
So I ranked the people I had an opinion about, and then for the remaining candidates that I do not particularly want to be board members, I didn't try to figure out their relative position to one another but just left them unranked, since I thought that would mean they are all ranked as "equally low preference -- do not want" by the software. The directions sort of indicated that is what to do.
Unranked choices are effectively treated the same as though they were all ranked with a single value which is one greater than the highest value you selected.
However, you're saying that it's better if all choices are ranked, even if I give several people a "99" by hand?
No, that wasn't my intent. Rather: It's important that you really understand that you *can* rank more than one, and that doing so doesn't really dilute your #1 preference. (I've talked to people who think their vote counts more if they concentrate it on a single #1 and don't rank the rest)
Making ranking all of them mandatory would be one way of getting that message through to voters. It would potentially prevent some forms of confusion such as "you should leave all you do not want unranked", which is only reasonable advice if you do-not-want them all exactly equally since a ballot marked 1, 2, 1, Unranked, 2, Unranked, Unranked is functionally equal to 1, 2, 1, 3, 2, 3, 3.
Or should I take the time to uniquely rank them? I'm not sure I understand why that's better given the outcome I want, which is for none of those people to become board members.
You should uniquely rank them if you think some are worse than others.
So in an election between Batman, Superman, Wonderwoman, Fidel Castro, Frankenstein, and Satan lets assume you prefer Wonderwoman, think Batman and Superman would be okay. You think the rest suck but think Fidel Castro the worst of the ones you dislike.
A reasonable ballot given those preferences might be
Batman 2 Superman 2 Wonderwoman 1 Fidel Castro 4 Frankenstein 3 Satan 3
You could also leave Fidel Castro unranked but that would have exactly the same effect on the election. If you were to leave the three you don't want unranked you would be failing to demonstrate your preference among them.
It will be interesting to see if a bunch of foundation-l readers revote after this thread :)
Indeed.
On Sat, Jun 7, 2008 at 3:23 PM, Ryan wiki.ral315@gmail.com wrote:
If you rank everyone, giving them a "99" by hand is no different than leaving them unranked (which is, for all intents and purposes, giving them a "100")
[snip]
It's important to note that the numbers are there as data entry lubricant. Their absolute values have no meaning, only the relative ranking that they describe. (This isn't a borda count)
That is, if the highest of your ranked values is 3 and you fill in all the unranked with 4 that has the same effect as ranking them all as 99 or leaving them unranked.
Effectively, your ballot is converted to an ordered list of sets of strict preference:
Wonderwoman > Superman, Batman > Frankenstein, Satan > Fidel Castro
Uniquely ranking the candidates you don't like doesn't help or hurt them with respect to the candidates you do like. All that does is say that you prefer "Unlikable Candidate A" over "Unlikable Candidate B", and if your likable candidate is not elected, you're saying that you'd want A before B.
Pretty much.
Schulze method does not not have the "Later-no-harm criterion", so your less preferred choices *can* influence results related to your higher preferred choices. But, unlike approval or borda Schulze's failure to meet later-no-harm doesn't translate into an obvious usable strategy in the general case.
On Sat, Jun 7, 2008 at 9:49 PM, Gregory Maxwell gmaxwell@gmail.com wrote: [...]
This very interesting and understandable explanation should be placed on a page that is clearly linked from the vote page.
Bryan
Bryan Tong Minh bryan.tongminh@gmail.com wrote:
This very interesting and understandable explanation should be placed on a page that is clearly linked from the vote page.
The suggestion box for 2009 is at http://meta.wikimedia.org/wiki/Talk:Board_elections/2009. :)
On Sun, Jun 8, 2008 at 5:10 AM, Jesse Plamondon-Willard pathoschild@gmail.com wrote:
Bryan Tong Minh bryan.tongminh@gmail.com wrote:
This very interesting and understandable explanation should be placed on a page that is clearly linked from the vote page.
The suggestion box for 2009 is at http://meta.wikimedia.org/wiki/Talk:Board_elections/2009. :)
No, I think Bryan meant: for this year. And I second that. The links present on the vote page *today* are more confusing than helping.
Delphine
On Sat, Jun 7, 2008 at 3:49 PM, Gregory Maxwell gmaxwell@gmail.com wrote:
So in an election between Batman, Superman, Wonderwoman, Fidel Castro, Frankenstein, and Satan lets assume you prefer Wonderwoman, think Batman and Superman would be okay. You think the rest suck but think Fidel Castro the worst of the ones you dislike.
A reasonable ballot given those preferences might be
Batman 2 Superman 2 Wonderwoman 1 Fidel Castro 4 Frankenstein 3 Satan 3
You could also leave Fidel Castro unranked but that would have exactly the same effect on the election.
Would it? Due to the failure of later-no-harm, isn't it possible that ranking Satan over Castro might actually cause Batman to lose and Satan to win?
Schulze method does not not have the "Later-no-harm criterion", so your less preferred choices *can* influence results related to your higher preferred choices. But, unlike approval or borda Schulze's failure to meet later-no-harm doesn't translate into an obvious usable strategy in the general case.
http://en.wikipedia.org/wiki/Later-no-harm_criterion gives a much needed example of this.
I think I can derive at least one general strategy for this situation, and it seems to resolve my problem with the condercet criterion promoting compromise candidates. The strategy is, if there are multiple candidates that are *unacceptable*, you shouldn't rank any of them.
Another way of looking at this is that you really are indifferent between all unacceptable candidates. You may hate Castro more than Satan, but either way if one of them wins and the board ratifies the decision you should leave the project and work actively against it. And from the other side of the coin, if a voter doesn't think either of two candidates are acceptable, their vote shouldn't count anyway.
Cool, I guess I can support the Schulze method once again, with the hope that those Memphis voters don't rank Nashville as their second choice. (http://en.wikipedia.org/wiki/Image:Tennessee_map_for_voting_example.svg)
On Sun, Jun 8, 2008 at 8:17 AM, Anthony wikimail@inbox.org wrote:
You could also leave Fidel Castro unranked but that would have exactly the same effect on the election.
Would it? Due to the failure of later-no-harm, isn't it possible that ranking Satan over Castro might actually cause Batman to lose and Satan to win?
Leaving candidates *unranked* is the same as ranking them with the lowest rank.
However, ranking Satan over Castro might help Satan win vs Batman but only if Satan and Batman are among the set of winners, see below.
Schulze method does not not have the "Later-no-harm criterion", so your less preferred choices *can* influence results related to your higher preferred choices. But, unlike approval or borda Schulze's failure to meet later-no-harm doesn't translate into an obvious usable strategy in the general case.
http://en.wikipedia.org/wiki/Later-no-harm_criterion gives a much needed example of this.
I think I can derive at least one general strategy for this situation, and it seems to resolve my problem with the condercet criterion promoting compromise candidates. The strategy is, if there are multiple candidates that are *unacceptable*, you shouldn't rank any of them.
As above, Leaving them unranked is absolutely equal to giving them all the next lowest rank.
My understanding is that Schulze fails later no harm only in a very limited way: The winner is guaranteed to be a member of the Schwartz set. The Schwartz set is the smallest set who's members are all unbeaten by non-members. It wouldn't be unreasonable to call all members of the Schwartz set ties.
Your ranking of Satan over Castro can't cause Satan to be unbeaten by Batman. But if all three of these were members of the Schwartz set then your preference of Batman > Satan > Castro might end up helping Satan win over Batman (and Castro) by influencing the tie-breaking.
But if your hated candidate is already in a form of a tie with your favored candidate.. you're already in a pretty bad situation. :)
Incidentally I think the only way to remove this risk entirely is not to vote at all. Which, I guess, is why Schulze fails the Participation criterion (http://en.wikipedia.org/wiki/Participation_criterion).
Another way of looking at this is that you really are indifferent between all unacceptable candidates. You may hate Castro more than Satan, but either way if one of them wins and the board ratifies the decision you should leave the project and work actively against it. And from the other side of the coin, if a voter doesn't think either of two candidates are acceptable, their vote shouldn't count anyway.
Sounds like am argument for IRV. ;)
Or really, under your line of thinking, plurality with voters who are always honest about their preferences would probably be ideal. Dunno where you find voters who are always honest. ;)
On Sun, Jun 8, 2008 at 12:08 PM, Gregory Maxwell gmaxwell@gmail.com wrote:
On Sun, Jun 8, 2008 at 8:17 AM, Anthony wikimail@inbox.org wrote:
Another way of looking at this is that you really are indifferent between all unacceptable candidates. You may hate Castro more than Satan, but either way if one of them wins and the board ratifies the decision you should leave the project and work actively against it. And from the other side of the coin, if a voter doesn't think either of two candidates are acceptable, their vote shouldn't count anyway.
Sounds like am argument for IRV. ;)
Or really, under your line of thinking, plurality with voters who are always honest about their preferences would probably be ideal. Dunno where you find voters who are always honest. ;)
No, I don't think plurality voting would be ideal because I think independence of clones is important. Unless you can convince (and allow) the clones to form political parties and hold primaries, anyway.
wikimedia-l@lists.wikimedia.org