Re: [EM] Proportional election method needed for the Czech Green party - Council elections
On Mon, May 3, 2010 at 10:16 PM, Kristofer Munsterhjelm km-el...@broadpark.no wrote: Peter Zbornik wrote: Dear all, I am sending a post scriptum to the email below. 1. The conservative method is only interesting if, the unambiguously pre-elected president and vice president(s) are not in the set of proportionally (for instance STV) elected council members. 2. If the unambiguously elected president and vice president(s) is in the set of proportionally (for instance STV) elected council members, then I guess the conservative method would include the optimal method as a special case (the optimal method was where the president and vice presidents are elected from the proportionally elected council members). 3. The number of pre-elected vice presidents in point 1 above can be zero. The president is always unambiguously pre-elected. 4. For completeness, I would like to add one additional requirement, which I think can be resolved after the seletion of a good voting procedure. Requirement: The selected council must contain at least X members of each sex (gender-equality rule). X is specified before each election. This gender rule is used in our organization today. A simple way of doing this, if the council size (after president and VPs have been elected) is even, is to have two elections, each of a council size equal to half the assembly. Then, for the first, only elect women, and for the second, only elect men. Use the same ballots, but remove candidates of the sex you don't want. I am affraid that this is not possible. First we have mostly odd-numbered council sizes, and secondly the gender rule does not require that half of the men should be men and the other half women. Our current gender rule goes as following: for every three members of the body, there has to be one person of each sex. A five member council thus has to have one woman and one man. For seven members it is two men and two women. Methods like Schulze STV work by comparing possible councils to determine which are best. Thus, it may be possible to limit them to only consider balanced councils. I'm not sure how to do this in ordinary STV, however, since it doesn't work that way, and in any case, this would be untested. Election-Methods mailing list - see http://electorama.com/em for list info
Re: [EM] VoteFair representation ranking recommended for Czech Green Party
Markus Schulze wrote: Richard Fobes wrote (2 May 2010): Once again Markus Schulze is trying to discredit the Condorcet-Kemeny method. If I really wanted to discredit this method, then I would mention ... Thank you for giving me the opportunity to put these issues into perspective. ... that this method violates independence of clones. A violation of the independence-of-clones criteria would require both of these conditions to occur in the same election: * There is circular ambiguity, which means there is no Condorcet winner. * Two (or more) of the candidates are recognizable as clones of one another. Each of these is uncommon (but not rare), but to have both in the same election would be rare. Yes, the Condorcet-Kemeny method fails to meet the independence-of-clones criteria, whereas the Condorcet-Schulze method meets this criteria. This is a small difference. For perspective, most of the currently recognized fairness criteria apply to both the Condorcet-Kemeny and Condorcet-Schulze methods. In other words, they meet and fail most of the same criteria. ... that this method has a prohibitive runtime so that it is illusory that VoteFair representation ranking could ever be used e.g. to fill 7 seats out of 30 candidates. The computer-calculation runtime for getting Condorcet-Kemeny results is long (factorial according to the number of candidates) if (!) all the Kemeny scores are calculated. However, not all the scores need to be calculated just to find the sequence with the largest Kemeny score. The wording we agreed on in Wikipedia, with the involvement of a neutral election-method expert, is that calculating the results for 40 candidates only takes a few seconds if well-known mathematical techniques are used. That's not a prohibitive runtime. My VoteFair ranking software calculates the results even faster, using an algorithm that I have not yet revealed. I'm still looking for a forum in which to share the algorithm. (Unlike you, I do not have academic connections that make it easy to publish papers in academic publications.) Yes, it takes the VoteFair ranking software a few seconds longer to calculate Condorcet-Kemeny results compared to software that calculates Condorcet-Schulze results. But even if the calculation time were a few minutes (for a particularly convoluted case), such a wait is not a deterrent for use in real elections. When the number of candidates reaches 30, the bigger challenge is for voters to meaningfully rank that many choices. That's why I recommend using approval (yes/no) voting to narrow the candidates to a reasonable number for ranking. The Condorcet-Schulze method has this same issue of a ballot with 30 candidates being difficult to meaningfully rank. ... that, although this method has been proposed more than 30 years ago, it has never been used by a larger organization. The Condorcet-Kemeny method is impractical to calculate without a computer, and the Kemeny method was proposed before computers became widely available, so it's lack of use prior to a decade ago is not significant. The Condorcet-Schulze method was the first (of these two methods) to be implemented in software, and the Condorcet criteria is so important that it is natural for early adopters to choose what's available. But the first Condorcet method to be adopted in this new digital era is not necessarily the best. The benefits of the Kemeny method -- including the fact that it is a Condorcet method -- are becoming known only slowly. The popularity of your Condorcet-Schulze method reflects the popularity of Condorcet methods, not necessarily the popularity of the Schulze-versus-Kemeny choice. Most of the people and organizations that use the Condorcet-Schulze method would not notice any difference in the results if the Condorcet-Kemeny method were used instead. Surely you have noticed that I have not made changes to your Schulze method page in Wikipedia, whereas you have repeatedly attempted to remove every mention of the word VoteFair from the Kemeny-Young method page, and to remove the link that reveals that there is a place where Condorcet-Kemeny calculations are available (for free). If I were more aggressive about promoting the Condorcet-Kemeny method, or if you were less active about trying to suppress it, it would be more popular. It takes time for wise people to make wise decisions. And fairness is very important to me. I'll continue to be patient as I wait for more people to recognize the advantages of the Condorcet-Kemeny method (which is a topic I'll explain in another post, in reply to a fan of your method). By the way, I don't keep track of all the groups that use VoteFair ranking. They find out about it online somewhere, they use it to elect their organization's officers, and I never hear about it. Occasionally I peek at the file contents to see how my free VoteFair ranking service is being used, make sure it's not being abused, verify there is
Re: [EM] VoteFair representation ranking recommended for Czech Green Party
Dear Richard Fobes, you wrote (4 May 2010): The book does not refer to the independence of irrelevant alternatives criteria, so where did you get the idea that it claims to satisfy that criteria? For example, on page 256 you claim: When VoteFair ranking is used, adding or withdrawing non-winning candidates cannot increase or decrease the chances of a particular candidate winning. Markus Schulze Election-Methods mailing list - see http://electorama.com/em for list info
Re: [EM] Proportional election method needed for the Czech Green party - Council elections
This is a good approach in the category of simple (only one method used) proportional ranking based methods. Use of proportional ranking reduces the proportionality of the council and the set of n presidents a bit but not much. The election of the president can be seen to happen before the election of the council. Same ballots are used for all elections. = Good for simplicity. Some small restrictions if the election criteria for P are different from the criteria of VPs and those of the council members. The last vice president positions are probably not needed. Their order will probably become public but should maybe not be emphasized. Markus Schulze of course recommends a Schulze method based approach but also any other good Condorcet method could be used as the basis. The Schulze family of methods has the benefit that it is quite well documented and the basic single winner Schulze method is also already used in some organizations. Probably Markus Schulze will also provide assistance in the promotion of the methods and related software. All these variants are however very similar so the argumentation and software is pretty similar in all cases. I support this approach as one proposal in the category of simple proportional ranking based methods. No need to limit to the Schulze method based approach only but to allow also other base methods to be used (e.g. Ranked Pairs, minmax(margins)). Also other categories or maybe variants of this one should/could be discussed and proposed as alternative approaches. Juho On May 4, 2010, at 3:42 PM, Markus Schulze wrote: Dear Peter Zbornik, this is my proposal: --Use the Schulze proportional ranking method. --The top-ranked candidate becomes the president. --The second-ranked candidate becomes the vice president. --If the first two candidates happen to be male, then, when you calculate the third-ranked candidate, restrict your considerations to female candidates. If the first two candidates happen to be female, then, when you calculate the third-ranked candidate, restrict your considerations to male candidates. The third-ranked candidate becomes the 2nd vice president. --The fourth-ranked candidate becomes the 3rd vice president. --The fifth-ranked candidate becomes the 4th vice president. --If 4 of the already elected candidates happen to be male, then, when you calculate the sixth-ranked candidate, restrict your considerations to female candidates. If 4 of the already elected candidates happen to be female, then, when you calculate the sixth-ranked candidate, restrict your considerations to male candidates. The sixth-ranked candidate becomes the 5th vice president. --The seventh-ranked candidate becomes the 6th vice president. Markus Schulze Election-Methods mailing list - see http://electorama.com/em for list info Election-Methods mailing list - see http://electorama.com/em for list info
Re: [EM] Proportional election method needed for the Czech Green party - Council elections
If you are looking for a proportional Condorcet method, I will also recommend the proportional election method that I developed. It is not STV-like, but it achieves proportionality when there are blocs of voters. It has the added advantage that it is already built into a running Internet voting system, CIVS. This algorithm has been used for many online polls and has been a success. The code of CIVS is publicly available. For more information about the method, see: http://www.cs.cornell.edu/w8/~andru/civs/proportional.html By the way, CIVS has recently acquired support for internationalization. It would be easy to construct a Czech instance if someone were willing to translate approximately 250 sentences from English to Czech. There is, for example, a Hungarian version (see http://www.cs.cornell.edu/w8/~andru/civs-test/index.html.hu, translated by Árpád Magosányi). I am in the market for help translating to other languages. Cheers, -- Andrew Election-Methods mailing list - see http://electorama.com/em for list info
Re: [EM] Proportional election method needed for the Czech Green party - Council elections
On Tue, May 4, 2010 at 8:12 AM, Peter Zbornik pzbor...@gmail.com wrote: I am affraid that this is not possible. First we have mostly odd-numbered council sizes, and secondly the gender rule does not require that half of the men should be men and the other half women. Our current gender rule goes as following: for every three members of the body, there has to be one person of each sex. A five member council thus has to have one woman and one man. For seven members it is two men and two women. For elimination based PR-STV, I think my suggestion would be the most reasonable. - Set the threshold at one larger than the required number - protect from elimination members of a gender if elimination would reduce their number below the threshold - prohibit from election members of a gender if that election would leave less than a threshold for the other gender - on the last round, remove remove the restrictions In a 5 person council, that means that there must be at least 1 man and 1 woman. If the candidates were Men: M1 M2 M3 M4 M5 and Women W1 W2 W3 then an election might go something like Round 1 M1: 20 M2: 15 M3: 15 M4: 10 M5: 10 W1: 10 W2: 8 W3: 12 Total: 100 Quota: 17 M1 gets elected + 3 are distributed Round 2 M1: 17* M2: 16 (+1) M3: 15 M4: 10 M5: 11 (+1) W1: 11 (+1) W2: 8 W3: 12 W2 is lowest, so is eliminated, +8 are distributed Round 3 M1: 17* M2: 16 M3: 15 M4: 13 (+3) M5: 14 (+3) W1: 13 (+2) W2: 0 W3: 12 W3 is lowest. However, eliminating W3 would reduce the number of women below 2, so the lowest man is eliminated. M4 is eliminated + 13 are distributed Round 4 M1: 17* M2: 17 (+1) M3: 17 (+2) M4: 0 M5: 16 (+2) W1: 17 (+4) W2: 0 W3: 16 (+4) M2, M3 and W1 all meet the quota, so all are elected, but no surplus is distributed. Round 5 M1: 17* M2: 17* M3: 17* M4: 0 M5: 16 W1: 17* W2: 0 W3: 16 If this had been a previous round, W3 would be protected from elimination, as there are only 2 women left. However, since this is the last round, (only 1 seat left to fill and 2 candidates for the seat), the restriction is lifted. Both W3 and M5 have 16 votes, so a tie break rule (say coin toss), would decide which one is eliminated. If M5 is eliminated, then the results are: M1+M2+M3+W1+W3 if W3 is eliminated, then the results are M1+M2+M3+M5+W1 In both cases, the requirement for at least 1 man and 1 woman is met. Methods like Schulze STV work by comparing possible councils to determine which are best. Thus, it may be possible to limit them to only consider balanced councils. I'm not sure how to do this in ordinary STV, however, since it doesn't work that way, and in any case, this would be untested. Yes, you can. The software would just need to be updated. A council with 5 men and 0 women would be considered to lose to a council of 4 men and 1 woman. Election-Methods mailing list - see http://electorama.com/em for list info
[EM] WMA (It's not monotonic or participation compliant, after all)
Kevin, I'm sure that you are right, but it makes me think that the only reason ordinary Approval complies with Monotonicity is that the information from polls is not an official part of the election: showing (true or alleged) support for a candidate in the polls can change her from a winner to a loser in the actual election. For me there are two reasons for using some kind of DSV: (1) to relieve voters from the burden of converting their sincere rankings or ratings into strategic ballots, and (2) incorporating the information necessary for good strategy into the official method, instead of relying on the unofficial polls that have so much potential for corruption. From this point of view Conditional Approval, WMA, DYN, and Rob LeGrand's sequential DSV based on Approval strategy A are all better than ordinary Approval based on disinformation from the unofficial polls. Here's a way (in the context of WMA) to lower the probability of inadvertently changing the winner when increasing her ballot support: instead of using full random ballot probabilities select a small subset of the ballots at random, and use the first place proportions from that subset in place of the full random ballot probabilities. For example in an election with a million voters, take a random sample of an hundred ballots to approximate the percentage of first place support for each candidate. Then in the vast majority of cases (99.99 percent of the time), raising a candidate (even to first place) would not change the probabilities on which the DSV strategy is based, so we could say the method would be at least 99.99 percent monotonic. Even if we only used a sample size of ten random ballots, I think that the method would be better than ordinary Range or Approval based on informal disinformation. It would be well worth the tiny sacrifice in monotonicity. Here's a related thought. Although Rob LeGrand's sequential approval DSV method (based on strategy A) fails Participation, if we take the following formulation of Participation too literally, his sequential method satisfies it to the max: If one more ballot B is counted after the election winner W has already been determined, the winner can only change to somebody ranked higher than W on that new ballot B. In fact, since Rob's sequential method puts the approval cutoff adjacent to the current winner, the only candidates that can possibly benefit more than W from the new ballot B are the ones ranked strictly ahead of W on B. [Here we assumed that Rob takes the ballots in the sequence that they are submitted, etc., which is not really true, so the method doesn't really satisfy Participation.] Hi Forest, --- En date de?: Lun 3.5.10, fsimm...@pcc.edu a ?crit?: De: fsimm...@pcc.edu Objet: [EM] WMA (It's not monotonic or participation compliant, after all) ?: election-methods@lists.electorama.com Date: Lundi 3 mai 2010, 18h06 Kevin and Chris, You?re both right, my ?proof? only showed that raising the winner W could not hurt her prospects on ballots where she was already approved or on ballots where she was rated at the highest currently unapproved level.? But , as Kevin pointed out, if there were candidates at two or more levels above W, and not all of these levels were above the current cutoff, the cut off could lower without lowering all of the way down to W. Back to the drawing board.? I?m starting to think that it may be impossible to have any kind of monotonic, non-trivial DSV method for automatically choosing approval cutoffs. I definitely think it is impossible. If you let voters choose their cutoff based on X, it's impossible to guarantee anything about how you are using X. And X is probably affected by raising candidates. I think you will have to greatly constrain how much freedom the voters have to place the cutoff. It will have to be based on as little information as possible. For instance you can view the situation in Bucklin as that voters lower their threshold continually until the method ends. If you raise a winner the method may just end sooner, but with the same outcome. Though for a DSV method this leaves voters relatively quite blind. They can't react to threats, they just gradually become impatient and start to compromise. I like my old (non-monotonic) Conditional Approval method where (if we assume a three-slot ballot) voters repeatedly add in their second-slot approval whenever the current leader is a disapproved candidate. No one can retract second-slot approval once granted. The method ends with the round where nothing else changes. The justification for failing LNHarm feels more tangible than is often the case: If we didn't count your lower preference (and those like it) when we did, the winner would've been somebody that you said you didn'tlike. Kevin Election-Methods mailing list - see http://electorama.com/em for list info
Re: [EM] MinMax(AWP)
On May 4, 2010, at 6:17 AM, C.Benham wrote: I think the idea that the CW should always be elected but it is sometimes ok to elect from outside the Smith set is a bit philosophically weird, and not easy to sell. I think electing outside the Smith set is a healthy idea :-). I agree that it is not the easiest to sell (if someone first brings the Smith set argument in). If group opinions would be transitive / linear as we expect the opinions of individual voters to be, then one could argue that the cyclic opinions in the Smith set must be fixed and in the resulting transitive order it would not make sense to elect anyone else but the first in that order. And that candidate could be only someone from the Smith set. However, opinions of groups are not always transitive but may contain sincere cycles. The cycle fixing approach that I described above removes all the cycles from the opinions and when doing so it ignores and hides the defeats within the Smith set. There are rare cases where the defeats of all the members of the Smith set are stronger than the defeats of some candidate outside the Smith set. In such cases it makes sense to elect that candidate outside the Smith set if the intention of the election is to elect a candidate that would have lowest opposition against her (as Condorcet methods typically do). No good method should have a tendency to elect outside the Smith set, but good methods may well be prepared to elect outside the Smith set in the rare cases where some of those candidates is considered to be a better choice (e.g. with less opposition) than any of the Smith set members. Human beings may visualize the defeat graph as a structure where the Smith set can be drawn at the top and other candidates below that set. That drawing / imagining technique is based on the hidden assumption of linear preference order of the candidates. The Smith set members are also generally not clones that could be logically replaced with one big bubble (= a new imaginary candidate that would represent all the clones). The cyclic relationships within the Smith set are hidden or maybe shown as strange / illogical curved or backwards pointing arrows. The world of potentially cyclic world of group preferences has been distorted. There is no natural two dimensional geometric way to express the cyclic preferences. The preference order or values describing the level of opposition of each candidate could be expressed in a one dimensional space, but one might not draw the Smith set members together and in the first positions. The explanation behind electing always the Condorcet winner but not necessarily always from the Smith set is that the Condorcet winner is not defeated by anyone but all the the Smith set members are, and they may be beaten badly when compared to some candidate outside the Smith set. Juho Election-Methods mailing list - see http://electorama.com/em for list info
Re: [EM] Proportional election method needed for the Czech Green party - Council elections
Some more comments on how the male/female requirements could be handled. In the description of Markus Schulze (see below) there were two steps where the male/female proportionality was handled. That approach works if there are separate requirements for the set of three first (vice)presidents and the rest of the council members. My understanding is that in the Czech Green Party there are no such requirements on the presidents. In that situation it may be better to push the forced male/ female election to the end of the list. It may be better to allocate the resulting problems in the last seats and elect the first seats in a more optimal way. It could also be a problem if we for example know what the three largest groupings that are likely to get the three first seats are. In that situation the idea of forcing the third grouping to always be the one that will be forced (if needed) not to elect their best candidate doesn't sound fair. Towards the end of the list the level of randomness is higher and the groupings that get those last seats may be happier to get them and never mind if the representative is male or female. This style of ensuring that appropriate number of male/female candidates will be elected is not optimal. It is for example possible that the fourth elected representative has an alternative of other sex that is about as popular as the elected president. In that case it could make sense to elect that alternative and in that way avoid the need to do some more violent changes later on the list. This approach of pushing the forced decisions towards the end of the list is however a working although somewhat ad hoc solution. More accurate solutions may be much more complex, e.g. ones that compare all possible sets of representatives and then pick the one that distorts proportionality with respect to voter preferences and sex related proportionality as little as possible. What would be a better but still simple approach? If one pushes the forced elections towards the end of the list the method could look as follows. --Use a Condorcet based proportional ranking method. --The top-ranked candidate becomes the president. --The second-ranked candidate becomes the first vice president. (optional step) --The third-ranked candidate becomes the second vice president. (optional step) --Also the following n candidates will become members of the council. --If at some point in the process all the remaining representatives must be male or female to make sure that the number of male/female candidates will meet the requirements, then restrict the consideration to male or female candidates only. This approach is thus not an optimal way to handle the sex requirements but maybe good enough and at least a simple one. (I note that Raph Frank proposed also an approach where the election of the last representative would be free of these sex related requirements. That is one way of relieving the proportionality related problems since at least the last choice that often distorts proportionality the most can be done quite freely. I'm not sure how big the improvement would be. There may be also other more sophisticated approaches as noted above.) Juho On May 4, 2010, at 5:03 PM, Juho wrote: This is a good approach in the category of simple (only one method used) proportional ranking based methods. Use of proportional ranking reduces the proportionality of the council and the set of n presidents a bit but not much. The election of the president can be seen to happen before the election of the council. Same ballots are used for all elections. = Good for simplicity. Some small restrictions if the election criteria for P are different from the criteria of VPs and those of the council members. The last vice president positions are probably not needed. Their order will probably become public but should maybe not be emphasized. Markus Schulze of course recommends a Schulze method based approach but also any other good Condorcet method could be used as the basis. The Schulze family of methods has the benefit that it is quite well documented and the basic single winner Schulze method is also already used in some organizations. Probably Markus Schulze will also provide assistance in the promotion of the methods and related software. All these variants are however very similar so the argumentation and software is pretty similar in all cases. I support this approach as one proposal in the category of simple proportional ranking based methods. No need to limit to the Schulze method based approach only but to allow also other base methods to be used (e.g. Ranked Pairs, minmax(margins)). Also other categories or maybe variants of this one should/could be discussed and proposed as alternative approaches. Juho On May 4, 2010, at 3:42 PM, Markus Schulze wrote: Dear Peter Zbornik,