Talk:Approval voting/Archive 3

Approval Hybrid
Has there been discussion of Approval/STV Hybrids? By this I mean using Approval-style check boxes, but fractionalizing the vote if the voter selects more than one, and then using STV algorithm by eliminating candidates from the bottom and thus recombining such divided votes on the strongest continuing candidate. One fellow, Bill Baldwin of Kansas describes the concept like this...
 * "Another possibility of using limited STV is the idea of allowing the voters several votes. That is, if N positions are to be filled, and there are M candidates (M > N) to fill the positions, then allow everyone to vote for however many candidates they want in the usual fashion. That means, the candidates voted for are simply checked. Then run STV with all the checked candidates having a rank of one, and all the unchecked candidates as having a rank of two."

I haven't analyzed this to see if it avoids any of the problems I have with Approval (e.g. the election of an inoffensive, but in fact poor candidate who few voters know much about, or actually support, who gets hoisted by the "any-but" strategy of a polarized electorate), but wonder if anyone else has done such an analysis. Tbouricius (talk) 15:50, 12 December 2007 (UTC)


 * I definitely would NOT call this an approval' system at all. I've explored such methods for many years, long supporting the option of tied-ranks counted as an equally split vote, like Equal-and-even Cumulative voting, so as a limit of one ranking allowed and split a vote among choices marked. It's also useful in cases where ballots can't support sufficient ranking slots. Of course there'd be more lost votes in the process over a rank STV. I'd call it maybe unranked-IRV or unranked-STV. My primary intuitive attraction to such a system would be that strategic compromise is expressed as a tied''-preference rather than a betrayal. I've heard suggestions that this approach would allow overvotes of plurality to be counted, better than throwing away overvotes, but a poor solution when you don't know voter intent. Tom Ruen (talk) 17:25, 12 December 2007 (UTC)


 * I'm glad to see Tbouricius exploring alternative election methods other than IRV. The proposal to fractionate the vote, without some recombination, has often been proposed to deal with the alleged violation of one-person, one-vote, but without recombination, it is essentially strategic suicide to cast divided votes. Approval, as it is, is really a form of alternative vote, i.e., voting for two is "I vote for this one if that vote is moot, and for that one if this vote is moot, and if it happens that the election is a face-off between these two, I'm accepting either one, which is an abstention in terms of choosing a winner, *but* a vote for either one if a majority of valid ballots is required to complete the election." In the end, only one vote *at most* actually counts.


 * Note that Range Voting is nothing more than Approval with fractional votes allowed. Thus, in the presence of strong voter polarization or partisan intensity, it reduces to Approval which reduces to Plurality.


 * I have many times proposed, though, a different kind of hybrid. Besides my own idea, one already exists. Bucklin voting. This is a ranked method, of course, which uses simple vote addition to bring in second and lower rank votes. There is a known form called ER-Bucklin which allows multiple votes at all ranks (whereas the Duluth system only allowed multiple votes in the third rank.)


 * Bucklin, of course, clearly satisfies the Majority Criterion, however we might slice it. ER-Bucklin doesn't satisfy the mind-reading version (i.e., the one that depends on a "preference list" that is not how the voters actually vote), but it could be argued that *in substance* it does, since Bucklin does not, under realistic election assumptions, create the alleged motive to bullet-vote "insincerely," that is, the bugaboo of multiple majorities, thus the failure of a majority first preference, can only occur if a significant number of voters vote for both frontrunners. In a three-way race, behavior gets more complex, to be sure, but, still, multiple majorities remain a "problem" that I wish we had.


 * There was a form of Bucklin that used fractional votes. It was ruled unconstitutional in Oklahoma, and I think I agree with that result, but I've been unable to find the actual text of that decision.


 * What I propose, though, is different. The ballot might look like a regular ranked one, but a vote in any rank is, for an Approval result, considered a vote for the candidate. (Though, possibly, there could be additional, non-approval ranks, I have not looked at that.) However, the ballots are analyzed by preference to see if there is a candidate who beats the Approval winner. If not, done (at least if the Approval winner got a majority, or if only a plurality is required, which I dislike, as does Roberts Rules). If so, then there is a runoff between the Approval winner and the preference winner.


 * Election simulations show that with sincere votes, the Approval winner is quite likely to be the Condorcet (pairwise) winner. Of course, "sincere" is difficult to define for Approval. Nevertheless, we can predict that in the vast majority of elections, Approval will choose a Condorcet winner or at least a member of a Condorcet cycle.


 * However, I interpret bullet voting as meaning "I would rather have to vote in a runoff than see any candidate other than my Favorite elected." There is no reason to suppose that it is *ever* insincere. Thus I'm quite put off by criticisms of Approval on the basis that it "encourages tactical voting," when, in fact, with Approval, this means "voting for your favorite." Or, more to the point, voting for a frontrunner in addition to your favorite -- but for most people in a two-party system, it's the former, i.e., bullet voting. With pure ranked methods, equal ranking not allowed, such as ordinary IRV, tactical voting *always* means preference reversal, i.e., ranking a less-preferred candidate above a more-preferred one.


 * Another easy hybrid that has not received the attention it might deserve is IRV with equal ranking allowed, and possibly with Coombs' method elimination, which is far more likely not to pass over a pairwise winner. Coombs method eliminates candidates, like IRV, but it handles the eliminations first with the *lowest* ranked votes.
 * --Abd (talk) 20:44, 12 December 2007 (UTC)


 * I agree Coombs' method is effectively an approval system, counting votes for all except last rank, not that I see any virtue in Coombs' method. Tom Ruen (talk) 20:52, 12 December 2007 (UTC)


 * It seems to me that borda count is an approval method, OR that approval is a specialized borda count [weights: 1,1,1,1...] OR maybe better that approval/borda are both specialized range voting. Well, it seems useful to me to group like-methods together.


 * I got in my first wikiwar with IPs when I started here with voting system page, tried to group methods as one-vote vs multivote, and the approval-lovers jumped in and called plurality/approval as yes/no methods because of the checkbox ballot types! (Rather than putting approval with range voting) Admittingly my groupings were half-baked, and I still don't know if I can call Condorcet a one person, one vote system. Tom Ruen (talk) 19:53, 13 December 2007 (UTC)

Well, the field has come a way since then. Approval is *clearly* a Range method. However, it's also easy to look at it as "Plurality" as well. The whole one-person, one-vote problem can be finessed quite simply. If the ballot of a voter is disqualified, does it shift the election outcome by more than one vote? That is, is the vote count of the winner more than one vote less? If so, we know that the voter did not have more than one vote, in the sense that really counts. Cumulative voting does this, allows voters more than one vote. But all alternative vote systems do allow the voter to cast different *alternative* votes, and Approval is really also an alternative vote system. "I vote for A if B is not going to win, and at the same time I vote for B if A is not going to win, and I am not going to make or break a tie between A and B. Condorcet methods are *similar* without the tie-breaking restriction. --Abd (talk) 04:55, 16 January 2008 (UTC)

Unsourced text in article
[ This section is a continuation of a thread in the POV tag section. DCary (talk) 07:45, 21 January 2008 (UTC)]


 * Well, sort of. It's a consideration de novo, which is why I split it out, beginning with the comment below. Hope this is useful. --Abd (talk) 23:37, 21 January 2008 (UTC)


 * You didn't treat this as a consideration de novo in your initial comments: it would be inconsistent for you respond to or critique my earlier comments in a true consideration de novo.  In any case, I'm not supporting that invention either. DCary (talk) 04:13, 22 January 2008 (UTC)

User:DCary placed some cn tags in the article in pieces of text that I had thought were simple equivalencies, obvious through definitions. I removed the tags, possibly in error in one or more cases, but this -- as cn tags will -- has raised the issues of the accuracy of these claims. My understanding of the principles involved is that Wikipedia text must not state as fact anything which cannot be verified by a reader through reliable sources; however, a limited amount of synthetic analysis, of an obvious nature from sourced facts, may be mentioned. As an example, if there is a list of jurisdictions in the United States which have adopted Instant Runoff Voting, and each item in this list is verifiable through reliable source, and there are, say, twelve jurisdictions on the list, an article could claim that there are at least twelve jurisdictions in the United States which have adopted IRV, even if no reliable source can be found that makes that claim. It's verifiable from present evidence by any reader. How far a reader may be expected to go to do the synthesis involved is an open question, as far as I know. It should not involve, however, expertise or advanced knowledge, that much is in the guidelines. Now, to the issues:

[ The preceding section introduction and the start of the first subsection were added by Abd (talk) at 02:36, 21 January 2008 (UTC) as part of a single signed comment. I am taking the liberty of splitting the two parts and continuing the section introduction as its own thread. If there are objections or suggestions about how to better handle this kind of situation, please say so. DCary (talk) 20:00, 21 January 2008 (UTC) ]


 * I don't see any problem. --Abd (talk) 23:37, 21 January 2008 (UTC)


 * It is perhaps best to review Wikipedia policy on original research at the source.  According to that policy,  synthesis is original research.  Wikipedia policy defines synthesis more specifically and more narrowly than the general sense that Abd apparently uses above.


 * Wikipedia distinquishes synthesis (not allowed) from summarization of ?reliable? sources (allowed and encouraged). The example Abd gives is clearly summarization.  However, in cases where there are no sources, there is no opportunity for summarization.  Flexibility is practiced, but too often at the expense of quality, as this and many other articles on Wikipedia demonstrate. DCary (talk) 20:00, 21 January 2008 (UTC)

Yes. The problem is pretty extensive with the Voting systems articles, because there is widespread development of the field, and expressed expertise, that seems to be happening outside the loop of peer-reviewed publication and other reliable source. That's changing, slowly, and I am, in fact, outside of Wikipedia, working on the problem, by setting up, hopefully, a peer-review system that takes advantage of the distributed expertise of the internet.

Many of the voting systems articles appear to have been written by participants in a project that is described on electowiki. These articles were written by people familiar with the topics, and, as is common with articles written by experts but not for formal publication, they are unsourced. However, *usually*, they represent broad consensus. If we just toss the material out, we lose what is actually pretty solid. Complicating this is that there are political movements which have an agenda, and part of that agenda is discrediting what are seen as "rival" voting systems. There has been a fairly coherent effort, over the last year, to remove from Wikipedia articles, for example, on election criteria, through AfD or merge. These articles, often having been written by one or a very few editors, aren't well watched, and AfDs went through that removed what are actually notable topics, such as Summability criterion. The AfD for Favorite betrayal criterion actually succeeded, but an editor simply recreated it and it survived the next AfD because, at last, someone was watching.

My policy in dealing with article text is not to remove it merely because it is not sourced, particularly if it is a matter of common knowledge. Essentially, there is a lot of work to do on the articles, and what is common knowledge and not really controversial shouldn't be the first priority! Further, what is published in the field is actually, generally, biased in a certain way. Access to publications is affected by politics. My favorite example, currently, is the article by Nagel on the Burr dilemma, which is an interesting article in the history it presents; unfortunately, Nagel's application of the historical situation is thoroughly tenuous and speculative, but its having been published in a peer-reviewed publication creates a problem. Nobody in the field that I know, and I know quite a few, accepts Nagel's analysis. But what is published to the contrary? I'm hampered by not having access to the journals -- most of the current articles in the field are being published in journals that don't have free public access. I managed to get a copy of Nagel's article, which is how I know how truly preposterous it is. (This actually is relevant to this article, because the Nagel article gets used periodically to assert Nagel's theory of strategic vulnerability of Approval voting as if it were fact.)

Election of the Secretary General of the United Nations
The article has, in the introduction,


 * Approval Voting is used to select the Secretary-General of the United Nations

and in the body,


 * The selection of the Secretary-General of the United Nations has involved an Approval poll

Now, the citation on the body text was supplied by me.(diff), and I modified that language to reflect what the source indicated, so I'm gratified that Mr. Cary thinks it accurate. What about the introduction? Well, it is not accurate. Problem is, it is also, quite possibly, from a reliable source. Basically, Steven J. Brams has repeated that claim quite a few times, and at least one of these may have been in a peer-reviewed publication. My preference? Take it out. It does not need to be in the intro. It's actually a weak claim, it is not "used to select" the officer, it is used to prepare for the selection. It is *interesting* that Approval is used, and approval polls can be extraordinarily useful compared to "vote for your favorite polls," they can rapidly narrow to the most broadly acceptable candidate(s), making deliberative process more efficient, and I've seen them function that way. So, essentially, my view of the best solution here is not to tag it, but to remove it. *But* someone may well object; we can cross that bridge if we come to it.

This, by the way, is an example of the limitations of Wikipedia reliance on peer-reviewed publication as "reliable source." It's a mild one, and we would have to look at what, exactly, Brams says, his precise language as was peer-reviewed, which I'm not doing, but there is another, far worse example which was cited in this article previously, Nagel's paper on the Burr dilemma, which would never pass muster on Wikipedia if published here. But being published in a publication, supposedly peer-reviewed, which allowed him to make essentially preposterous claims, it can be cited as a source for us. The error he makes has actually been taken up and repeated by some who should know better, such as Warren Smith, who repeats Nagel's assertion that the first Presidents were elected by Approval Voting. Perhaps we should put this in this article? Approval voting was used to elect the first Presidents of the U.S.? I don't think so! Much as I would love that it were so! --Abd (talk) 02:36, 21 January 2008 (UTC)

Is Approval a simple form of Range voting?
I think we are likely to see a book appear any day, some of us have seen prepub copies that will say about exactly this. "Range voting" is a new terminology for an old method, Cardinal ratings. The introduction has:


 * Approval voting is a simple form of range voting, where the range that voters are allowed to express is extremely constrained: accept or not.

To put it in Range terms, the voter may vote 1, equivalent to accept, and 0, equivalent to not accepting the candidate. Mr. Cary raises the issue of various forms of Range voting and special rules that are sometimes proposed as part of the system, but this is sidestepped by the "form of Range voting," and those who promote Range voting universally -- as far as I know -- accept the claim that Approval is a form of Range, and I suspect that Brams does so likewise.

There are two basic forms of Range: sum of votes and average vote. Currently, the Center for Range Voting, much to my chagrin and that of some other Range supporters, promotes average Range. Approval is, of course, not equivalent to average range, but it is equivalent to sum-of-votes Range, and students of Range, again universally, accept sum-of-votes as a form of Range, and all the theoretical work on simulations has been done with *no* study of the effect of abstentions, which is where sum-of-votes and average deviate, otherwise they are totally equivalent. If we have a Range ballot, used for a Range election, everything is set up to count it, and at the last minute there is a change in the rules: voters may only vote 100% or 0%, and we overprint the ballots so that the intermediate ratings can't be used, the Range system could be used to count the ballots, and it would be an Approval election.

Another way to look at this is that Plurality allows voters one vote per office, Approval allows voters one vote per candidate. And Range allows voters fractional votes, at most one full vote per candidate. It is that simple, unless someone tosses in the monkey wrench of using average vote instead of sum of votes, which creates a mess: what about a single voter who writes in a candidate and votes that candidate 100%? Oops! that candidate wins. So then a rule is needed, which CRV calls a "quorum" rule. It's arbitrary, and I've argued for years that if the proposal is average Range, it is certain to be shot down. *Later*, after there is usage of sum-of-votes Range, which is really a very simple progression beyond Approval, there *might* be some reason to consider average vote. I think it's a bad idea, though, for elections. Might be great for polling.

Now, could I find a reliable source that claims Approval is a form of Range? I can find *tons* of email comments and web sites, but I'm not sure that, until the book comes out, I could support that with reliable source. But it is *obvious*. Does anyone here actually claim that it isn't? It seems Mr. Cary did so above, but, frankly, I didn't follow it. He wrote:


 * The non-equivalence of voting for none, 1, or more versus voting for those candidates one approves of is recognized and is a distinction Abd himself has promoted. For example, even when there are just two candidates, I might vote wisely, even sincerely in some sense, for exactly one candidate, even though in another sense I don't approve of either one.

The term "approve" has two different meanings. One is an opinion, the other is an action. In the context of the method, it's an action, and it is exactly equivalent to "vote for." I have raised the distinction Mr. Cary mentions in considering what "sincere voting" means in Approval. I have, in fact, argued that to "approve" a candidate in this method, for an actual election, has little to do with the opinion, but rather with a decision that the voter makes and implements. But the *method* -- which is what this article is about -- cares nothing about whether the voter "approves" of the candidates or not, *unless* "approves" is a merely synonym for "votes for." It is the marks on the ballot that will be counted, they are actions, not sentiments, and if somebody tells Approval voters that they should vote for "all the candidates they approve of," they could be misleading them. Ballot instructions should not place sentimental meaning on votes! Rather, voters may, quite properly, use preferences and preference strengths to determine optimal votes for themselves, but these are irrelevant to the method -- though they do impact the theoretical analysis of how the method handles voter preference patterns and presumed modes of expression.

There are two basic forms of Approval Ballot: one with Yes/No for each candidate, and one where the voter may vote for a candidate or not (i.e, may simply not mark anything for that candidate). In most implementations of Approval, it is really only the Yes votes that are counted for the purposes of determining the winner, No votes are dicta, though they may aid in the prevention or detection of fraud. (In some applications, No votes are subtracted from Yes votes. This is actually a Range 2 method! (i.e., -1, 0, +1, with 0 being the default vote). Approval is Range 1. The Wikipedia Arbitration Committee, thus, is using Range 2 voting.)

Bottom line: the comment of Mr. Cary does not bear on the operation of the method. Take a Range election and prevent all votes except the top rating and the bottom one, and it is precisely an Approval election. (I'm assuming sum-of-votes Range, with abstentions the same as bottom ratings.)

Mr. Cary goes on to discuss Range more, but it seems he has missed something. Yes, there are various forms of Range. But what the article says is that Approval is "a form" of Range, not that it is "equivalent" to Range. However, where there is some variant on Range, there will be, in fact, normally, a corresponding equivalent with Approval. I.e., a form of Approval that is that variation, with the votes simply restricted to min and max. Average Range, for example, can be done with Approval, it is Average Approval. That's not what Brams proposed, certainly. But, as noted, Brams did not actually invent Approval, he studied it. And then Average Approval has exactly the same problem as Average Range: the single voter.... and thus, then, the need to define some minimum support. With polling, it doesn't matter, because polls only advise, and the analyst can take into account the number of voters involved.

I'll look around and see what I can find as a reliable or "semi-reliable" source. (The defacto situation is that *some* sources are routinely allowed that aren't fully qualified as "reliable source." That is unstable, though ... frequently it reflects the fact that the peer-reviewed journals are not available on-line unless one has connections or money). --Abd (talk) 02:36, 21 January 2008 (UTC)


 * Interesting on average of votes approach, apparently to allow votes without an opinion to not affect a candidate's rating. Approval does support this with 2-input ballot yes/no, like Image:Approvalballotchoice.png, so no mark means "no opinion" rather than "no approval". Tom Ruen (talk) 03:22, 21 January 2008 (UTC)


 * The image doesn't say how the votes are counted. That kind of ballot is often proposed, but, in fact, it's only the Yes votes that count. If a voter votes "No," it prevents someone from adding a Yes vote, that's an argument that has been used. (This is one of the arguments used against Approval, that if it uses a standard plurality style ballot, a fraudster may add desired votes. However, it's true currently with the no-overvoting rule, that fraudsters may void undesired votes by adding marks, and there are allegations that something like that happened in Florida 2000. Six of one and a half-dozen of the other. There is no substitute for good ballot security.


 * Some fairly bright people have proposed the average-votes method for Range -- and, of course, it's the same thing with Approval -- but .... some other fairly bright people have claimed that this is an effective way to torpedo the whole thing, noting that it is a *separate* reform, and the *tradition* is essentially sum of votes, and we see this with conflicting initiatives. They are compared with sum of Yes votes, the No votes are only used to determine majority approval of the initiative. Yes/No ballots for candidate elections could also be used like that. Majority yes required to win, but ... winner has the most Yes votes. If the winner doesn't get majority yes (i.e, more Yes than No *for that candidate*), the election fails (i.e., goes to further process). --Abd (talk) 04:00, 21 January 2008 (UTC)


 * Right, a yes/no ballot helps detect undervoting to confirm voter intent and disable fraud. And an average-approval method (whatever you want to call it), could use 3-mark options yes/no/no-opinion. On the otherside, I still can't support a majority requirement in approval since multiple rounds of voting make it too easy to offer false votes, whether or not choices are eliminated, voters ought not to be influenced by unreliable previous approval counts. Tom Ruen (talk) 04:17, 21 January 2008 (UTC)

[ Just a reminder that it is helpful if editors follow Wikipedia's Talk page guidelines. There is a link to those guidelines at the top of each initial editing page of a Talk: page. I'll emphasize three points:  keep on topic,  stay objective,  be concise. DCary (talk) 06:56, 21 January 2008 (UTC) ]


 * Well, Wikipedia would, I think, like to encourage access to editing the encyclopedia by people with disabilities. I have ADHD, confirmed by many specialists. One of the consequences -- besides the ability to hyperfocus, which allows me to see what others can't see, sometimes -- is an inability to be initially concise. I have a good friend who is quite like me, but he is known as a concise writer. How does he do it? He takes about three or four times as long to write a piece. I can be concise, but it is extraordinarily inefficient. So, my suggestion to DCary is, don't read it if you don't want to. There is no obligation on any editor to read what I write. Talk is just Talk, and those who are bored by it may leave the room, they lose nothing but an opportunity to understand what I'm saying, and they will have future opportunities. I'm concise in edits, give summary explanations of edits. And if an actual dispute appears, I can and will retract to concise argument in Talk. I have not considered this a dispute. Yet. So, yes, it would be helpful. Sometimes another user appears who reads what I've written, apparently understanding it, and who summarizes it. If nobody understands and supports what I write.... well, perhaps I'm out on a limb. Or we haven't attracted enough participation yet. Where I am concise, by the way, is when I've developed a clear conclusion. When I'm *not* concise is, in fact, when I'm most objective, because I will include just about every POV that occurs to me, and I think dialectically, which drives some people crazy. --Abd (talk) 15:51, 22 January 2008 (UTC)


 * There is some confusion, perhaps because the title of this subsection and the bulk of Abd's opening comment in this subsection appear to be misdirected at defending a statement that I have not challenged, a statement in the article's first paragraph. What I did challenge were the equivalencies claimed in the first two sentences in the  Procedures section.  Please refer back to my comments in the POV tag section for details. DCary (talk) 20:18, 21 January 2008 (UTC)


 * I was addressing, first, the general situation, and I avoided moving on to the text found deeper in the article. Now that Mr. Cary has specified more exactly what he is challenging, we can turn to that. This is the text of that section:


 * Each voter may vote for as many options as wanted, at most once per option. This is equivalent to saying that each voter may "approve" or "disapprove" each option by voting or not voting for it,[citation needed] and it's also equivalent to voting +1 or 0 in a range voting system.[citation needed] The option with the most votes after all votes are tallied wins.


 * There are two parts to this. Is Approval voting, the referent of the article, equivalent to each voter "approving" or "disapproving" of each option by voting or not voting for it. I don't know if I can find a text that requires no synthetic judgement, but the basic fact I noted above is that, if I'm correct, Brams applies the "Approval voting" label to elections where the voters use a standard ballot, without the No option, and also to elections where No is explicit. The example for the article, the image, shows a standard ballot, without the No. Which of these, is Mr. Cary objecting to? The second sentence is an *explanation* of the procedure, explaining the *name* of the method. I've argued many times that it is dangerous to consider the votes as "approval" of candidates, but this description is quite common, and it is, of course, implied in the name of the method. I don't call the method that, when I'm writing advocacy pieces -- or sometimes just for fun -- I call it "Count All the Votes." Should make a nice bumper sticker, don't you think? The idea that these votes are "approvals," based on some supposed internal state of the voter, is exactly where critics get off claiming strategic voting. But there is another meaning to approval. It can be an action rather than an opinion. And that is precisely what a vote for a candidate is. It is an action whereby the voter adds weight to the election of that candidate, for whatever reason.


 * As to the claim about Range equivalency. Range is a class of methods, with variable resolution, proposals span the "range" from (0,1), (-1,0,+1), (0,9), (0,100), etc. and Warren Smith would allow more resolution than that if he could get away with it. As I explained above, there is also average Range and sum-of-votes Range, and there can also be, in theory, other aspects to the method that alter its characteristics. Range Voting with voters voting absolute commensurable utilities is actually ideal for the purpose of maximizing overall satisfaction with the result, but, of course ... how do we do that? In some situations, it can happen. But normally, we expect that voters will vote according to various strategies, and, it turns out, the most successful Range method, better than ordinary Range -- which is pretty far ahead of all other methods according to the simulations -- is Range with top-two runoff. The point, though, is that there are all kinds of Range methods, but there is a basic concept of Range, it is what was known previously as Cardinal ratings. What it *boils* down to is allowing fractional votes. Approval is Range with no fractional votes allowed, that is really the most significant difference, and every other Range variation I've seen would have a corresponding Approval variation, and, again, I mentioned this above.

Range with ratings of 0 and 1 is Approval, Approval with fractional votes allowed is no longer called Approval, it's called Range. (A key point is that the vote for each candidate is *independent* from the vote for every other candidate. This is why both Range and Approval are, according to the older definitions with strategic voting meaning preference reversal, strategy proof. Borda count is "range-like," but with strict assignment of the "ratings." Range supporters *uniformly* -- actually I can think of one sort-of exception, someone who was promoting, as I recall (-1, 0, +1) Range and giving it a new name, consider Approval to be a Range method, which should explain a bit why the Center for Range Voting, in the actual political activism it supports, is promoting Approval Voting. It's the basic Range method. Approval advocates, as well, accept the definition of Approval as the simplest Range method. (Some of them think that basic approval is better than higher resolution Range because it forces the voter to consider and make the necessary compromises with political reality, though, in fact, in my opinion, the higher resolution forms really do the same. *Many* think that higher resolution Range is better than Approval, but also consider Approval to be the reform du jour because of its utter simplicity. Take a plurality ballot and simply stop tossing overvotes.

However, these communities are largely without peer-reviewed journals supporting them. The opinions are expressed on web pages around the internet. It's pretty easy to find out, personally, what the consensus is on a topic: just post to, say, the Election Methods mailing list with something contrary to it, you will see very quickly! (Post a question, "is this the consensus, you might get some answers, but for many readers it will be ho-hum.... People are more likely to respond to error than to what is correct.) So, problem is, there is nothing in this that satisfies WP:RS, *even though, in fact, there is more effective peer review of claims through the mailing list than there is in the editorial process for peer-reviewed publications -- at least some of them.* The true problem is that there is no mechanism for formal peer review, no "publication decision" by a publisher, whose reputation is at stake, which is why we consider peer-reviewed publication the gold standard for reliable source. (Well, actually, it's the silver standard. The gold standard is certain peer-reviewed or other responsibly-published review of the field that considers all the points of view. Secondary or tertiary source. Wikipedia *plus* peer-review would be gold of a purity we have never seen before....)

Can I find reliable source for those claims? Short of that, can I find attributable opinion, which is clumsier but which *can* be used. The claim about strategy vulnerability of Approval is actually an opinion, not a fact, and it should be attributed. I'll fix it. (Mr. Cary was correct to insist that the claim based on Nagel et al stay in the article, less correct to remove the weasel word that made it true without contest, but, in the end, the correct response is what he suggested: balance it.

It's nice to see some actual work on the article taking place. It's been insufficient and stagnant for quite a while. --Abd (talk) 23:37, 21 January 2008 (UTC)


 * Abd: I'm checking where we have some agreement. I'd agree to the following, would you? As long as we are talking about range voting where the winner is the candidate with the most total points, as opposed to some version of average points range voting, approval voting is equivalent, at least in some formal sense and in a sense of tallying the votes, to any two-valued range voting system, whether the values are, for example, 0 and 1, or 0 and 10, or -1 and +1.  Also, that any of the varieties of approval voting are equivalent to each other, again at least in some formal sense and in a sense of tallying the votes, as long as there are exactly two ways for the voter to express the vote, independently for each candidate, and the winner is the candidate with the most "affirmative" votes, regardless of whether the two ways for the voter to express themselves are, for example, vote for and don't vote for (abstain from voting for that candidate), vote for and vote against, vote yes and vote no, vote approval and don't vote (abstain from voting for that candidate), or vote approval and vote disapproval.  On the other hand, approval voting is not equivalent, not even in some formal sense nor in a sense of tallying the votes, to a voting system that allows each voter to express a vote in more than two ways, independently for each candidate.  For example, if the voter is allowed to vote independently for each candidate with a 0,1,or  2, with a -1, 0, or +1, or with a 0, 1, 2, or 3, and the winner is the candidate with the most total points, that would not be equivalent to any approval voting system.  (I apologize for the long, complicated sentences.  I'm not trying to be tricky with phrasing.  Any agreement is subject to revision if a rabbit gets pulled out the hat later on.) DCary (talk) 09:12, 22 January 2008 (UTC)


 * Short answer: Yes, written after I wrote what is below, which explores it. The only quibble: it is not the number of different ways a voter may vote, exactly, but the number of different ways that the system, for purposes of determining a winner, counts the votes. I give below an example of an Approval system where the voter may vote in three ways, Favorite, Accepted, and (no vote), Favorite is counted the same as Approved, the Favorite vote is used for other purposes. If the Favorite vote is used in any way to determine the outcome, it is no longer clearly an Approval method. Now, to the exploration:


 * I *generally* agree, and, to match Mr. Cary's concern for precision, I will look for the possible exceptions, none of which should detract from the general agreement. Mr. Cary has objected to my lack of brevity; however, the primary cause is writing without having a fixed conclusion in mind. When I have a fixed conclusion, I can be brief like everyone else. I think this should be understood, or else I wouldn't bother saying it: brevity is commonly a product of a POV. (But people with a POV may also lack brevity, that's another story. Sometimes, rarely, NPOV is also concise. Notice how many times making the text of an article NPOV requires adding text to it. The POV version was brief, and then editors with the POV will complain that the addition is "too much," "unnecessarily detailed," "nitpicky," etc.)


 * The promoters of Range voting initially proposed, in fact, sum-of-votes, and Cardinal ratings is sum-of-votes. Whether or not to promote average Range was heavily debated, and the only reason, in my opinion, that the Center for Range Voting promotes average range is that the primary founder, Warren Smith, has an opinion about it. The other cofounder doesn't actually support that opinion, but isn't forceful about it. Most writers about Range simply assume sum-of-votes, or, more accurately, assume that sum-of-votes and average Range are the same. The simulations done with Range assume, in fact, that it is the same as Approval with fractional votes, and they assume no abstentions on individual candidates. (That is, when the simulations consider so-called strategic voting, they assume that voters give a minimum rating to undesired candidates, typically all of them, even if they actually have preferences among them that might otherwise be significant.


 * Bottom line: When we say that Approval is equivalent to Range, we mean that Approval is a Range method, that the Approval rules correspond to a specific Range implementation, which would be, yes, sum of votes, and an assumption of zero rating if the voter does not vote for the candidate (which is what sum of votes does. Specifically, Approval is Range 1, where the 1 refers to the preference strength expressable, it is equivalent to the number of ratings minus 1. If blanks are considered midrange, this is actually an additional vote option, thus it is Range 3. (If Nos are subtracted from Yes, as ArbComm does, there are then three expressable votes, each with a different effect, so it is no longer Approval.)


 * What is critical is the number of vote options that actually affect the result in distinct ways. A method has been proposed which I called A+. It is simply approval voting with an additional position which voters can mark, called "Preferred." A preferred vote is counted the same as an ordinary vote, for the purposes of determining the winner, hence A+, even though there is an additional option on the ballot, is identical to Approval in behavior *for the purposes of determining the winner.* It is an Approval method. The additional mark is used (1) for public campaign finance, (2) for assigning future ballot position to parties, perhaps, (3) to satisfy the desire of voters to be able to indicate their favorite, and (4) to measure election performance, since this can detect a possible Majority criterion failure in the event of more than one candidate gaining a majority. There is, then, another method, which is *not* pure Approval, but which is, in fact, a form of top-two runoff Approval, with the additional requirement that multiple majorities also trigger a runoff. Or, alternatively, if there is more than one majority, the preference markers prevail. This is moot here, it is only background to explain why, for the purpose of classifying and comparing methods, it is the number of distinct expressable votes for each candidate -- instead of marks on the ballot -- which determine the variety of method. Approval, like Plurality, has two.


 * Concise way of saying this, that actually covers it all: "Approval is a form of Range Voting." the use of the indefinite article implies that there are other forms of Range that are not Approval, but that all Approval methods are also Range methods. It follows from the definitions. And I can adduce, if necessary, many sources for this. But I won't do that work unless it is necessary, and I don't consider it necessary. In the Range article, I think that there is a corresponding statement, something like "Range voting, when the number of ratings used is two, reduces to Approval voting." --Abd (talk) 16:22, 22 January 2008 (UTC)


 * Thanks. Your clarification about the way multiple voting options are tallied is on target.  That was a loose end that for the sake of preciseness needed to be addressed. DCary (talk) 23:47, 22 January 2008 (UTC)


 * So two more points, the first, I think is another point of agreement. Given the equivalency of methods for tallying approval votes, the way in which voting is presented to the voter can influence how the voter votes.  It can make a difference whether the ballot instructions simply tell the voter to vote for as many candidates as the voter chooses, or whether the ballot instructions tell the voter to vote for the candidates that the voter approves of.  It makes a difference whether the voter is asked to vote 0 or 1, or whether to vote -1 or +1.  There won't be differences for all voters.  There may be differences for only a few out of a large group, but in a large group, say 1,000 people, we would expect there will be differences.  The cause may be lack of mathematical literacy, perceived nuances in meaning, or other psychological influences that the voter is not aware of.


 * The second point is not necessarily a point of agreement, but my attempt to confirm my understanding of what you have presented. In the context of the Procedures section, the terms 'approve' and 'disapprove' are equivalent in meaning to 'vote for' and 'not vote for'.  (Please ignore for the moment that, as used in that section, the terms should perhaps not be quoted.)


 * Do you agree with the first point and am I on target with the second? DCary (talk) 23:47, 22 January 2008 (UTC)


 * Yes as to the first point, first. The ballot instructions could affect how voters vote. While, certainly, a jurisdiction could decide to give any instructions it chooses (How about "Vote for your favorite Democrat. Other votes will be foolish) but such instructions might well be seen as attempts to improperly influence how people vote, and they could indeed be so. I'll give an example. In Ann Arbor, Michigan, a referendum implemented preferential voting for Mayor. Apparently, voting was along partisan lines, Democrats for, Republicans against. Why? Well, the Republicans were accustomed to winning the mayoral election because the Democrats and the Human Rights Party would split the vote. The result was a mayoral election won by the Democrat, sometimes claimed to be the "first black mayor." The Republicans then promoted an initiative to rescind it, which passed. Now, suppose a similar jurisdiction were to implement Approval voting. Perhaps the Republicans aren't able to get rid of approval, but they can, perhaps, control the ballot instructions. "Vote for all candidate you approve of" could cause a shift toward some of the HRP supporters not voting for the Democrat, thus shifting the election toward the Republican. Current ballot instructions do not tell voters how to vote. They do not say "vote for your favorite," nor do they say, "vote for your preferred candidate among those whom you think has a chance to win" -- which is what most voters actually do. Comes Approval, the situation is the same, and, indeed, the likely sensible voting pattern is the same, only with, now, an additional freedom. You can add other votes if it serves you. These votes are all "approvals" but only in the active sense, i.e., as an agency approves a permit or a government approves a visa. They are actions which give enabling weight to the candidate. They are not sentiments.


 * As to the second, again, yes. "Vote for" and "don't vote for" are equivalent, in analysis of Approval voting, to "approve" and "disapprove;" however, in consideration of the psychology of voting, there might be some difference. "Approve" is sufficiently undefined, however, when we try to use it in the psychological sense -- as distinct from the active sense -- that it is practically useless, to use it one would have to come up with some very specific definition for it. Perhaps "would be pleased to see this candidate elected"? That, however, is not a quality of the candidate, because what we are pleased with depends on context. We are "pleased" compared to something. If a stock corporation issues a dividend, whether the shareholders are pleased or not depends on what they were expecting. Whether or not a candidate will be "approved" depends on the alternatives, including expectations of what is possible. I might prefer Al Gore for President, but ... if he is not on the ballot, do I write him in? What if there is a Ranked Choice election, like in San Francisco? Do I use up a rank to cast a write-in vote for Al Gore? With Approval, actually, it's easy. I can write in anyone I like, subject to the space provided on the ballot, and it is *harmless* at worst. At best, what if everyone else was thinking the same way?
 * --Abd (talk) 04:41, 23 January 2008 (UTC)


 * Thanks for the feedback and confirmations. DCary (talk) 08:27, 23 January 2008 (UTC)

Here are my assessments of where we stand, building on the agreements we appear to have and taking note of some of the differences. "Approve" can indeed mean either to have a kind of thought, a favorable opinion or to express something, to take a kind of action. Regarding the first claimed equivalency, as a first case, if "approve" refers to having the thought, then consistent with our second point of agreement, there is not a full, unrestricted equivalency. There are situations where the two descriptions are not interchangeable in meaning or effect. The lack of equivalency is not necessarily a statement about approval voting as a voting system, per se, depending on where you draw the boundaries for the system in a system/user dichotomy. The claimed equivalency is about how voters vote.

If on the other hand, "approve" means the action, then there are two possibilities. As I read several dictionaries and their definitions (Merriam-Webster definitions) for "approve", they do not divorce the action from the favorable opinion as Abd does. As a second case, consider that "approve" referring to an action means to express a favorable opinion. Then this is really the same case as the first case, and there is no full, unrestricted equivalency. If instead, as a third case, we accept Abd's distinction and interpret "approve" and "disapprove" to mean in this context "vote for" and "not vote for" respectively, then the first equivalency (inappropriate quotation of terms removed):
 * "Each voter may vote for as many options as wanted, at most once per option. This is equivalent to saying that each voter may approve or disapprove each option by voting or not voting for it ..."
 * really just means:
 * "Each voter may vote for as many options as wanted, at most once per option. This is equivalent to saying that each voter may vote for or not vote for each option by voting or not voting for it ..."
 * which is really only a very badly garbled construct. With a stretch it might be construed in a meaningful way, but the easiest way to construe it is as something that is clearly false.

Now with the first claim of equivalency being false or garbled nonsense, what is being said in the second claim of equivalency? What exactly does "it's" refer to? Certainly not approval voting, which hasn't yet been mentioned in the section. The second claimed equivalency is not a claimed equivalency between two voting systems. "It's" reasonably has to refer to one of the parts of the first equivalency. Regardless of which part "it's" refers to, the second equivalency is also about voters voting. So again, as in the first case, based on our second point of agreement, the second equivalency is not a full, unrestricted equivalency. DCary (talk) 08:27, 23 January 2008 (UTC)

So what do we do? While we have reliable sources for describing approval voting procedures, no sources have been provided for the claimed equivalencies. So it is difficult to justify a particular way of trying to rehabilitate them.

I recommend that we drop the claims of equivalencies from this section and stick to language describing approval voting procedures that closely reflect our reliable sources, Brams and Fishburn, 1983, for example. That description mentions voting as approving and just voting for candidates, leaving the relationship between the two unspecified/ambiguous, but certainly not asserting an equivalency. This approach will also have the significant added benefits of simplifying the procedure description and making it more understandable to readers with a limited background on the subject. I'll even suggest that the whole description be moved to the first introductory paragraph, perhaps leaving the first section, retitled, to the remaining example.

The relationship of approval voting to range voting is already mentioned in the introduction. There could be some value to giving a more complete description of that elsewhere in the article, sourced eventually perhaps by the book Abd mentions. There is already an "Other issues and comparisons" section where a more complete description might readily fit.

The whole user interface influence/effects issue is something that is worth discussing in the article, but only after the we have reliable sources for the material. The current section on ballot types is related material tha is emblematic of the problems with unsourced content. DCary (talk) 08:27, 23 January 2008 (UTC)

My concern about a recent series of edits (|diff) (|diff) (|diff) on the introduction focuses on what is now the second sentence and how "approve" is used in relation to voting for a candidate. As described earlier in this Talk: subsection, "approve" can be understood with more than one meaning. The different meanings can reflect different points of view about what an approval vote may or may not mean. As a result, using "approve" can be ambiguous and removing ambiguity can have the effect of stressing one POV at the expense of another. Similarly, "or" can have an inclusive or an exclusive connotation, it may or may not signify an alternative. The current language tends to emphasize an interpretation that "voting for" a candidate is equivalent to or is a (technical) definition of "approving of" a candidate. The recommendation of the previous comment, which received no objection, was to use language that closely reflected the language in a reliable source. On the use of "approve", the language before this change (|diff) more closely reflected the cited source than the language after the change. The current version is closer than some of the intermediate changes. Some explanation and justification for the change would have been helpful, especially given the recent discussion that preceded the change. DCary (talk) 04:15, 6 February 2008 (UTC)

I will also note here the singular/plural grammatical mismatch of "Each voter may ... as they wish.", although it is not a NPOV concern. DCary (talk) 04:15, 6 February 2008 (UTC)

Multiple Ballot Questions
The "Historical use" section opens with an unsourced claim that approval voting has been used in deciding (presumably in Nevada) conflicting ballot questions because voting on conflicting ballot questions is equivalent to approval voting with a majority required to win. I tagged this with a fact tag both because it was not supported by a source and because I seriously doubted its veracity. At the time I left a discussion note indicating that claims of equivalency needed to be supported. Within a few days, Abd deleted the tag, claiming in the edit summary that the equivalence was obvious diff. However, I don't know what such an equivalence is, let alone how to prove it. It certainly isn't obvious to me. So as indicated earlier, I'm following up on this issue and inviting Abd to provide a reliable source for the claim and/or, as best he can, 1) identify and prove the equivalence and 2) show that equivalence and its proof is obvious. DCary (talk) 04:13, 22 January 2008 (UTC)


 * The claim that multiple ballot question process is a the same process as Approval voting is, again, based on definitions. This actually has not been widely noticed, I'm not sure why. Remember, what we had was the claim that Approval voting is not used for public elections, so this is a counterexample. It's been mentioned in quite a few very public places, places where erroneous arguments can meet a firestorm of objections, but none of them, to my knowledge, meeting RS. However, it has also never been challenged in those places, that is, I have seen no argument at all contrary to what this article says: the process for choosing among multiple conflicting ballot options is equivalent to the process for choosing among multiple conflicting candidate options in an Approval system with majority Yes vote required for election.


 * Now, having written what is below, I'll now say that everyone may have overlooked a detail.


 * The reference provided with the claim is to Nevada law. There are other states with the same provision, I think it is common for states that allow initiative by petition. I just edited the text to, hopefully, make this clearer, this is what it is now:


 * To resolve multiple conflicting Ballot Questions, where the vote is Yes or No on each, and some voters abstain on some Questions, some jurisdictions provide that, if more than one question gains a majority, the question receiving the most Yes votes prevails over the other, which is the equivalent of approval voting with a minimum majority required to win.


 * Here is why this is important: it is claimed that Approval does not satisfy the majority criterion, which is debatable, it depends on definitions (as we have seen above, on the definition of "sincere vote"). That's really semantics, but underneath it is a political claim that some (such as FairVote make explicitly, Approval voting allegedly does not satisfy "majority rule." (We ought to have that in the article by the way, we need better presentation of the controversy over Approval, the efforts of Steven Brams, etc. It's notable.) The case presented is precisely the one that these Ballot Question rules deal with, i.e., two candidates both gain a majority. In that case, it is possible that one, the one with a lower Yes vote (or just plain lower vote, which amounts to the same thing *except* for the satisfaction of a majority issue), was preferred by a majority over the other. So if "majority rule" is violated by Approval, it is violated already, in the multiple ballot question issue.


 * Now, thinking this over, I re-examined the question of this "majority requirement," and I may have incorrectly -- still -- presented the situation. I need to examine this rigorously.


 * These are the possible situations. This can be analyzed with more than two initiatives, but, for simplicity, I'm sticking to two. if VY(1) is the Yes for for the first initiative and VN(1) is the No vote, and we have the like of this with the second initiative, we have the following conditions:


 * VY(1) > VN(1) alone, 1 would pass. OR
 * VY(1) <= VN(1) 1 fails.


 * VY(2) > VN(2) alone, 2 would pass. OR
 * VY(2) <= VN(2) 2 fails.


 * so, overall, there are the following possibilities.
 * 1. 1 passes, 2 fails.
 * 2. 1 fails, 2 passes.
 * 3. 1 and 2 fail.
 * 4. 1 and 2 pass, 1 prevails
 * 5. 1 and 2 pass, 2 prevails
 * 6. 1 and 2 tie (I think this contingency is considered, but I'll neglect it)


 * Now, it could occur that 1 passes and 2 fails, as an example. But suppose that the Yes vote for 1 is lower than the Yes vote for 2. Under ordinary Approval (plurality, Yes/No voting) rules, 2 would prevail. This outcome would actually contradict majority rule, since a majority voted against 2. But in most implementations of Approval, in fact, the No vote is assumed, so the precondition is impossible, because an increase in the Yes vote would always represent a decrease in the No vote. Yes/No approval potentially has this problem, but it is easily avoidable: in order to win, a candidate must obtain majority approval.


 * This, of course, could lead to unresolved elections, just like any majority requirement. With initiatives, this is no problem, because there is no necessary desirability that any initiative at all be passed. With officer elections, it's the same as a majority requirement now; a runoff will be triggered.


 * In what is truly significant about the situation with initiatives, the matter of multiple majorities, this exception is moot: the precondition is that both pass. It is only with Yes/No voting that it arises. And, of course, we have opened a whole can of worms. Well, that's the way the cookie crumbles. I've been accused in the past of cherry-picking data and facts and analyses. I don't. I report what I find, and, in fact, detest the habit of too many political activists to reveal what helps their goals and conceal what does not. I place honesty above my personal goals.


 * The point about "majority rule" remains correct, and that is quite possibly where this information should go. It certainly is *almost* equivalent to Approval voting, and it is the same as Approval with a majority requirement to win or the candidate is disqualified from winning, which is what I meant in what is in the intro.... I'd say that *this* claim is now correct, but we have uncovered a subtle difference between Yes/No approval and Plurality ballot Approval. In practice, if Yes/No approval disregards the No votes, they are the same. But, that, of course, raises other issues.--Abd (talk) 20:07, 23 January 2008 (UTC)


 * Do any of the public places this was mentioned without receiving objection happen to be instances of the "peer-review by mailing list" or other such thing that you have touted as a source for reliable information, even if it is not acknowledged by Wikipedia policy? If so, can you provide specifics?  Where? When?  Submitted by whom? DCary (talk) 21:40, 23 January 2008 (UTC)


 * Fairly clearly, the justification for the claim is not obvious. Actually the claim is erroneous.  The conflicting ballot questions election is some form of 3-value voting system and can not be equivalent to approval voting.  Since the claim is about a 3-value voting system, it probably does not belong in the approval voting article.  Without any sources, this claim is clearly original research, apparently Abd's OR.  Abd originally added the claim to the article (|diff).  The recent attempts at rehabilitating the claim still left it as an erroneous and unsourced claim.  As original research, and especially as erroneous original research, it does not belong in any Wikipedia articles, least of all as a Wikipedia-based exercise in self-publishing.  The claim will be removed.  DCary (talk) 02:28, 25 January 2008 (UTC)

Comparing Bucklin and IRV
The "Historical use" section contains an claim about the use of Bucklin voting in the United States. The claim is started with a comparison of Bucklin voting to Instant-runoff voting. I had added a fact-tag to the claim of similarity, and Abd removed that tag (diff), justifying the claim of similarity was obvious on the following points:
 * Both use a ranked preference ballot
 * Both do a sequential analysis in rounds
 * Both stop further rounds when a candidate has a majority of the votes

I am continuing my challenge to this part of the entry and I am expanding the challenge to include the entire comparison between Bucklin and IRV. I agree that Bucklin and IRV share the points of similarity that Abd enumerated. However, there are a number of other voting systems that share those similarities with Bucklin or that can be described in those terms. Even approval voting can be described in a way that matches those points of similarity.

So the fundamental question is: Why mention IRV and only IRV? The Bucklin voting article is subsequently referenced and a brief description of the method is also given. Then the comparison is given that is actually important to the claim, the similarities between Bucklin and approval voting. In addition, the rest of the comparison is inaccurate, especially the uses of the words "only", "approval", and "alternative".

I recommend that the comparison to IRV be removed. That comparison is not relevant, it confuses people, especially those who do not have the requisite prior knowledge, and its use at the start of the item only delays and obscures the core content of the item. Those problems could only be exacerbated by an attempt to fix the current inaccuracies in the comparison. We don't have a reliable source to provide a basis for keeping the comparison. DCary (talk) 21:18, 23 January 2008 (UTC)


 * I agree and boldly removed the reference to IRV. Tom Ruen (talk) 21:54, 23 January 2008 (UTC)

Is Approval "susceptible to strategic voting"
An anonymous editor added some old text back to the Strategy section. I modified it to "allegedly susceptible," and we need more discussion in the article than that. Approval was designed for minimum susceptibility and, indeed, what the critics assert as "strategic voting" requires an inconsistent definition of "strategic voting" than is used with other methods.

Strategic voting with ranked methods, which is all that voting theorists used to consider, requires expressing reversed preference. That is, the voter supports A over B, when, in fact, the voter prefers B to A. The voter does this with an expectation that the outcome will improve from the voter's point of view. The most common and blatant example of this would be a voter who votes for Gore when she prefers Nader, because she knows that the Nader vote would be wasted, and Bush might win. (no political implications intended!)

Approval was designed to avoid this, and it succeeds, completely. So what is the criticism? What is the alleged vulnerability to strategic voting, what does the voter do? Well, the most common one is that a voter votes sincerely! Yes, that is exactly what is claimed, though, of course, it isn't stated that way. A voter allegedly "actually" approves A and B, but wants A to defeat B, so votes for A only. *That is a sincere vote.* The conditions establish that the voter prefers A over B (or else why would the voter vote for A), but, supposedly, approves both.

However, "approves both" is a situational decision that voters make depending on reasonable election outcomes. This is not strategic voting in the sense being claimed. (tactical voting is what Wikipedia calls it, and that is its own can of worms). It is simply voters voting to maximize their expected outcome for the election, which is what voters are *supposed* to do. The articles cited are hit pieces against Approval voting.

By establishing a new criterion, "approval," as if it were an absolute, critics make it appear that Approval is susceptible to strategic voting, i.e., "insincere voting." It is preposterous, really. --Abd (talk) 18:54, 21 January 2008 (UTC)


 * The addition of "allegedly" is not in keeping with Wikipedia's policy on neutrality.  Abd's comments above only confirm that assessment. See also Words to avoid, especially  "allegedly" . As a result, I have reverted the addition of "allegedly".


 * I agree with Abd that there is much more that can be said about strategy and about what the already referenced reliable sources have to say about it. However, where there are differing points of view, Abd is required by Wikipedia policy to neutrally present alternative points of view based on content of  reliable sources.  In his comments above, he only offers alternative perspectives as his own  original research, which can not be the basis for article content.  That is not otherwise passing judgment on the validity of the ideas Abd expresses, just applying the rules of Wikipedia.  Start with reliable sources. DCary (talk) 22:25, 21 January 2008 (UTC)


 * I present my perspective on the issues as background. I don't think you will notice me putting original research into articles, unless it is the minimal OR that is allowed, i.e., research that anyone could have done and which they would now be able to do with the sources provided. Like 2 cat in this place and 2 cats in the other place equals 4 cats in both places. A minor point is also that editors are not *required* to do anything. We are free. If we violate civility, if we edit war, if we are disruptive, we can be blocked. Failing to "neutrally present alternative points of view" is not a punishable offense. Rather, the "penalty" is that the edit might not stand. And, indeed, it should not stand, if inadequately established and properly challenged. Problem is that we know there are notable points of view that are not rooted in reliable source. It's a basic problem for Wikipedia in general, and there is still quite a bit of controversy about it. *In fact*, consensus of editors routinely trumps an absolutist interpretation of WP:RS, the *policy* is WP:V. If I'm correct, editors have been blocked for edit warring to defend a demand for "reliable source." It's worth noting that the former, RS, is only a guideline, and it has been proposed that it be merged into the verifiability policy page. However, the whole policy/guideline is in trouble, because WP:V refers to WP:RS for the definition of "reliable source." What seems to be sufficient is that the facts stated in the article are reliably verifiable, which would be the obvious interpretation of "verifiability" and the insistence that it be a published source is weaker.


 * While starting with reliable sources is nice, it can be highly inefficient. In my view, the best articles are written *as a start* by people who know the subject well and who also understand WP:NPOV and are careful about it; this will often be sparsely sourced. It takes a lot of time to properly source text, far, far less than to simply write it. Articles that get written directly from reliable source often simply copy it, or paraphrase. And, of course, "reliable sources" often have their POV, as I mention above, there is reliable source and there is reliable source. Later, others come along and source it. Meanwhile, this is distracting me from actually finding sources for what's well-known in the field.

--Abd (talk) 01:33, 22 January 2008 (UTC)

One more point. "Allegedly" can be used in summary style. What I did, inserting that word, was to consider that there is a controversy here, and only one side is in the article, and, in fact, the other side is more notable. In any case, the fix is to edit the section, the weasel word was only intended as a stopgap. It will be clear when I'm done, I believe, that this was an notable allegation, not a fact. --Abd (talk) 01:36, 22 January 2008 (UTC)

Approval polling

 * Approval voting can also be used for voting or polling questions which allow a variable number of winners. A clear example is the question of candidate inclusion for debates. An approval poll would be better to ask: "Which candidates do you want to see in the debate?" rather than the usual polling question: "Who would you vote for if the election was today?"


 * In such a poll, a fixed threshold for inclusion could be made. For example, a debate could include all candidates above 15% approval support. Special rules would be needed to guarantee at least 2 candidates passing, possibly simply including all of the candidates.


 * The advantage of approval polling is that voters have no fear that "overvoting" will hurt their higher choices. Undecided voters will tend to want to hear from more candidates early in the campaign, and will tend to reduce their preferences as voting day approaches.

I remove this unreferenced section above. It might be called approval polling, but it isn't approval voting in the sense of this article's subject, rather represents actually N independent polls for N candidates. Tom Ruen (talk) 03:56, 25 January 2008 (UTC)

Removal of claim "Approval is vulnerable to strategic voting."
I removed this from the introduction. I cannot verify the claim from the sources. I have read the Nagel article, and I don't recall seeing a mention of "approval is vulnerable to strategic voting," and the other article is clearly argumentative, just look at the title. It is one side of a debate, and practically incivil in it at that. This is what I took out:


 * Approval voting is susceptible to strategic voting.





If someone can quote exact text from these articles showing the claim that "Approval voting is vulnerable to strategic voting, as it is defined in the Wikipedia article -- which is what we are saying when we link to that -- then I'd have to change my position. Nagel discusses a problem he sees with Approval, a problem which I think is 99% fantasy (it didn't happen with Burr, it is his imagination that it *might* have happened), but that problem isn't "strategic voting." As to Saari, anyone got access to the text? Did I ask this before? The necessary claim isn't in the abstract. More to come. --Abd (talk) 01:54, 22 January 2008 (UTC)

(I'm going to have a problem, Nagel does say that, I think, but ... the devil is in the details) --Abd (talk)

Okay, from the Wikipedia article on Tactical voting, which is what the text actually pointed to: "In voting systems, tactical voting (or strategic voting or sophisticated voting) occurs when a voter supports a candidate other than his or her sincere preference in order to prevent an undesirable outcome." In the Burr dilemma, the problem is voters voting for their sincere preference! What *exactly* do they claim? You can't just summarize it as "strategic voting" without a definition of strategic voting and without the article actually saying so -- or matching the definition precisely. --Abd (talk) 02:00, 22 January 2008 (UTC)

First of all, what is "tactical voting" or "strategic voting"? We have the Wikipedia definition above; it isn't really kosher as a source, of course. (but if it is incorrect, and we link to it, using the same name, we have a problem and we should really address an incorrect definition.) So I'm asking de novo: what can we find?

[http://www.crest.ox.ac.uk/papers/p94.pdf Extending the Rational Voter Theory of Tactical Voting, Stephen Fisher] "when voters decide it is optimal to abandon their first preference party and vote for another it is said that they voted tactically (or strategically)."

Nagel, it turns out, does discuss "strategic voting." I'm reading it over again, he goes on and on, it's frustrating reading.... so I'll be back with that in more detail. --Abd (talk) 02:18, 22 January 2008 (UTC)

I do not find a place where Nagel defines what he means by "strategic voting." He makes a statement: "The mere fact that approval voting runs into a problem with strategic behavior is not sufficient to reject it. We know from the Gibbard-Sattherthwaite theorem that, when there are three or more choices, all voting procedures are vulnerable–not always, but under some configurations of preferences--to manipulation by strategic voting (Gibbard 1973)."

Now, reading over what was previous to this, the "problem" Nagel refers to is not a problem of the method, per se, but a "problem" that the voter faces: shall the voter approve a candidate other than the favorite? If so, the voter risks the favorite losing to that candidate; if not, the voter risks that both the favorite and the other candidate lose to a third. This is what the article properly describes as "strategic voting" in Approval, i.e., strategy of voting. Here is where Nagel slips in the shift in definition. Voting for a second is not "abandoning" the first preference. Or is it? We will need to examine this.

Most hits I find on "tactical voting" and "strategic voting" are referring to the practice with Plurality. This involves preference reversal. Much of the problem with this issue is that "strategic voting" and "sincere voting" get all tangled up. Some claim that to add a second approval is insincere voting if the voter has any preference at all between the first and second. Other claim that only preference reversal is insincere. And, in many months of going over this topic, I've found no way to clearly resolve it. It is definitions of words, and people extended the definitions without being explicit that they were doing so. James Armytage-Green, at least, was quite aware of the problem, and attempted to address it. In order to examine the Majority Criterion performance of methods, he had to assume that the votes were sincere. What is a sincere vote in Approval? There really is no definition, other than you vote for your favorite and you don't vote for the worst candidate. What you do with candidates in between is what? Sincere? Insincere? *There is no way to avoid failing to fully express preference when there are three or more candidates, because Approval allows only two ranks or ratings. So the voter must equal rank at the top or at the bottom, and, either way, does not express a preference. So are all Approval votes in such elections insincere? Seems preposterous? Does to me, too. I think I just defined a sincere vote in Approval, quite well, just above. Top vote and bottom vote. The rest is optional.

So what is "strategic voting" in this situation? I think Brams got it right. Approval is designed to provide no incentive for "strategic voting," which was intended to mean "insincere voting."

In any case, if we are going to have a *refutation* of Brams in the article, we certainly should have Brams original claims with it. Otherwise it introduces severe imbalance. Indeed, why doesn't the article give the history of Brams introduction of Approval? It's well-known, and it should be there. Approval was *designed* to be strategy-proof, in the sense that strategy was understood. Brams has written a history. When we have this, from Brams, in the article, it will then be important to introduce criticism paired with it.

Well, this is more than I can complete tonight.... —Preceding unsigned comment added by Abd (talk • contribs) 02:55, 22 January 2008 (UTC)

Once again, Abd has taken action that is not in keeping with the Neutrality policy. As before, the comments Abd makes to justify his action only serve mostly to confirm the problems with his actions. First, disagreeing with or not liking a sourced statement is not reason to delete it. Second, it would have been more appropriate for Abd to (re)read the referenced source before finding fault with the source or the statement in the article; doing so would have avoided premature and ill-founded claims about the source and its representation in the article. It certainly is not appropriate for Abd to delete the statement from the article while he thrashes about, spending time educating himself on the issues. As Abd points out, this is not new material for this article. As a result of all this, I have reverted his deletion.

This reversion is not a judgment on my part on the specifics of how the article ultimately should deal with this issue. However, I do have an expectation that where there are different, even (seemingly) conflicting opinions or approaches in various reliable sources, that variety should likely be neutrally presented, not one to the exclusion of others. DCary (talk) 05:10, 22 January 2008 (UTC)


 * I'll point out a distinction between an action "not in keeping with ... neutrality policy," and text that does not conform. NPOV text is *commonly* found through a series of edits that are each, arguably, POV; as such, given sufficient editor participation and common assumptions of good faith, eventually the text settles as NPOV *in the judgement of the participating editors.* NPOV is actually not a characteristic of text, it's a judgement by those participating. Absolutely, a disagreement with a statement is not a reason for it to *ultimately* be excluded. However, exclusion *can* be part of the process of getting to NPOV. It will then be brought back by another editor, hopefully modified to reflect the POV or possibly neutral intention, perhaps poorly realized, in the explanation of the removal. This has actually been done, now, and if I had not removed the text, it might not have become balanced as it is. In other words, I'll claim some measure of success resulting from my removal of what was, in fact, a POV claim. That a claim is in a source, or even in a series of sources, does not make it a fact, particularly when the sources acknowledge the controversy! I'll get to this below. The text had, as a fact, "Approval voting is vulnerable to strategic voting." It's not a fact, it is an argument, and it uses loaded language: "vulnerable," which implies harm, and "strategic voting," which is, as I've shown, unclearly defined. "Strategic behavior" actually has some of the same problematic implications, *but*, expressed as it now is, it is certainly far closer to NPOV than it was, and I just might leave it there. Progress, not Perfection. --Abd (talk) 16:36, 22 January 2008 (UTC)


 * I found an opposing peer-reviewed source which, in it's introduction, specifically refers to the other two sources in contention here. Clearly WP:NPOV demands both sides be presented. MilesAgain (talk) 05:16, 22 January 2008 (UTC)

I found three more on the con side, and no more on the pro side, all from the same google search (["approval voting" strategy]) -- I am sure there are more on both sides. MilesAgain (talk) 05:51, 22 January 2008 (UTC)


 * Feel free to share at least the references to the sources you are finding. Btw, as you describe it, which side is pro and which is con, and to what? 71.139.15.101 (talk) 08:19, 22 January 2008 (UTC)


 * diff MilesAgain (talk) 08:21, 22 January 2008 (UTC)

(unindent)MilesAgain is a POV editor, an acknowledged sock, but he is also quite commonly a useful editor (and his behavior has generally stopped short of what would cause consequences from being a sock). He is pro-IRV, which commonly translates to anti-everything-else. Thus he frames what apparently an effort by a number of editors, to find neutral text, as a POV conflict, and edits are going to be seen as supporting Approval (Pro) or as criticizing it (Con). He is correct, above, however, in noting that we should ensure that "both sides be presented." The framing of this as having only two sides is characteristic of POV, by the way. But it does, in fact, fairly represent this issue: claims of vulnerability to strategic behavior are commonly associated with support for IRV. But not always. In any case, the Niemi citation is particularly interesting. The original claim in the article had Niemi and Nagel as sources. Nagel is explicitly promoting IRV, and is the one who uses "strategic voting" in his language, in the Burr Dilemma article -- which is essentially a mass of speculation, published in a peer-reviewed journal. Niemi, however, seems to be much more neutral and careful in his examination, though his conclusions are certainly debatable. I think it is worth quoting the abstract here.


 * Approval voting is being promoted as "the election reform of the 20th century" (Brams, 1980, p. 105), and indeed if voters' preferences are dichotomous, approval voting has some remarkable qualities: it is uniquely strategy-proof, a candidate wins if and only if he is a Condorcet winner, and voters have simple strategies that are at once sincere and sophisticated. However, all of these results depend on the existence of dichotomous preferences, a contrived and empirically unlikely assumption. Here I show that these virtues of approval voting are replaced by some rather undersirable features under more plausible assumptions. More fundamentally, rather than promoting "honest" behavior, as is sometimes implied, the existence of multiple sincere strategies almost begs voters to behave strategically. I also examine sophisticated approval voting and show that in the general case it need not pick a Condorcet alternative. Ironically, there is a condition under which Condorcet winners may always be picked, but for this to occur, voters sometimes have to vote for candidates of whom they disapprove.

First of all, this does not state "Approval voting is vulnerable to strategic voting." The article might, I haven't seen the article yet, except for some interesting pieces that are on the page visible on JSTOR.

Rather, Niemi talks about "strategic behavior." Is Approval voting vulnerable to "strategic behavior"? Who would put it that way? The article describes strategic voting under Approval, and the description was not controversial. Approval advocates describe Approval voting strategy. But is the optimal strategic behavior "strategic voting," or "tactical voting"? This is presumably undesirable, but that is quite debatable, because "strategic" or "tactical voting" -- strategic optimization -- under Approval does not involve preference reversal, unlike the case with ranked methods such as IRV. So what does Niemi actually say? Above, we see that he acknowledges a central claim of approval advocates: "if voters' preferences are dichotomous, approval voting has some remarkable qualities: it is uniquely strategy-proof, a candidate wins if and only if he is a Condorcet winner, and voters have simple strategies that are at once sincere and sophisticated."

"Sophisticated" is a stretch in a dichotomous situation. The sophistication arrives in *determining* the dichotomy. In any case, Niemi is here acknowledging that an equal vote in the presence of a preference is sincere. "Tactical voting" implies insincerity, and it and the equivalent term, "strategic voting," excepting rare contentious references to Approval voting, are *always* used to indicate insincere voting, as I found in my searches underneath what I wrote above. I did not, in fact, *find* any references to Approval in those searches unless I included the term in my search (they would be there, but buried in massive numbers of other hits; almost all hits are about Plurality elections). So if he later calls the strategic behavior he is going to examine, "strategic voting," he will be, to some degree, contradicting himself. I'm going to suspect that he used the word "behavior" here because he understood the implications.

Niemi goes on, "However, all of these results depend on the existence of dichotomous preferences, a contrived and empirically unlikely assumption." Now, there is a standard dichotomous preference that is what is commonly, practically universally, described as optimal Approval strategy, it is simple, and it is very effective. In practically every partisan election in the U.S., and in most nonpartisan ones, this simple strategy would work; it results in dichotomous preferences, and it is likely, I'd suggest, that most voters would understand and use it. For most voters, in a two-party or two-front-runner situation, *by definition*, this strategy would suggest bullet voting. So I don't understand why Niemi considers this "empirically unlikely." Perhaps he explains this elsewhere in the article. The strategy, of course, is (1) vote for the favorite front-runner. This is, in nearly all plurality elections actually taking place, already the most common voter behavior. (2) Then, vote for any candidate you prefer to the favored front-runner. That's it. As Niemi says, simple. Sophisticated? Well, the whole point of considering Approval strategy-proof is that this is actually the optimal strategy in the situation where there are only two front-runners, no more consideration is required. You can't do better by being clever or "sophisticated." The *only* exception is when there are three or more front-runners, which is actually, even with nonpartisan elections, quite rare. Then strategy becomes more complex, though, in fact, the simple strategy still works quite well: make your best guess as to whom the top two will be, then follow the simple strategy. *However*, to be even more effective, the voter must start to consider preference strength and odds.

But none of this involves "insincere voting." Thus using the term "strategic voting" is problematic. Niemi then calls the strategic behavior, "rather undesirable." And, he continues, "More fundamentally, rather than promoting "honest" behavior, as is sometimes implied, the existence of multiple sincere strategies almost begs voters to behave strategically."

There is, it seems to me, a strange confusion here. Approval voting allows voters to simply vote for their favorite, the simplest possible strategy. If every voter does this, Approval reduces to Plurality, which is often considered a criticism of Approval. I and others claim, though, that *if we are comparing Approval with Plurality* -- which is almost always the actual comparison being made, with the exception of comparisons, sometimes, with IRV -- this is no criticism at all. It's saying that if the features of Approval aren't used, the "failure" is the status quo, and, since there was no cost involved, simply a minor shift in rules, requiring no reprogramming of voting machines (all must already be able to handle multiple votes or Yes/No votes), and voter education is quite simple, and voter effort is minimal under most conditions, this isn't a harm, it is a lack of benefit. But that behavior, all bullet voting, in most election situations with more than three candidates, is actually extraordinarily unlikely.

In the three candidate situation, Approval allows voters to behave strategically, that is, to vote for a candidate whom they do not, in some sense, "approve of." It is only by categorizing votes as approvals -- i.e., as sentiments about the candidates-- that we can claim that the vote is insincere. And Niemi notes that this came up in the review of his article, in a note on the page I can see on JSTOR.

"As an anonymous referee pointed out, one useful perspective is to ignore any implications of the word approval and simply to think of the multiple votes as simply one way of aggregating preferences. Then my results can be viewed in part as an application of the Gibbard (1973) and Satterthwaite (1975) theorem showing that all such voting schemes are manipulable. If one does lend credence to the notion of approval, then it is worth pointing out some of the consequences of that view."

This is the crux of it. The referee pointed out what I've pointed out many times, though I may place emphasis differently. I argue that votes are votes, they are *decisions*, not sentiments, though sentiment is involved in how voters make their decisions. It is possible, indeed common, to think of Approval voting as involving "approval" of candidates, but there is a semantic ambiguity there. Niemi takes "approve," it appears from what he wrote, to refer to an absolute sentiment, i.e, a quality of candidates, divorced from context, which the voter then expresses in a sincere vote. It is *this*, in fact, that is divorced from reality. Voters rarely vote their fully sincere preferences, period. Take any plurality election that allows write-ins. For a voter to vote sincerely by the absolute preference standard, the voter *must* write in the number one person among all eligible persons -- or even among all persons, period, depending on how strict we want to make the rules for sincere vote. In plurality elections with write-ins, in fact, the election is totally open -- in theory. (Typically, in top-two runoff, write-ins are still allowed, so, if voters really do prefer another candidate not included in the top two, they could -- and perhaps should, if they think they are in the majority or close -- write in the name of their favorite, and they could, in fact, thus frustrate the runoff and maybe even win it. From this perspective, in fact, most voting methods are Condorcet-compliant *if* voters are aware of their position and vote it. Indeed, this is why direct democracies, using parliamentary process, don't need advanced voting methods. Advanced methods are needed for uninformed electorates, voting by secret ballot election without good information about the real context.)

Niemi, considering "approval" to be an absolute sentiment, then, is concerned about "strategy." This definition allows him to think of a decision of the voter to divide the candidates into dichotomous preferences, i.e., an "approved" set and a "not approved set," to be "strategic behavior" -- he is correct in any case, -- and for it to be a "problem." But *nobody* on the other side, i.e., claiming Approval to be "strategy-free" pretends that the voter should "approve" -- in the absolute sentiment sense -- of all "approved" candidates. Rather, in fact, as Approval was proposed, "approve," for these, means an act of voting for, a decision to support, which may even involve clothespins on the nose. That is, if the voter has a large preference for C, but A and B are the frontrunners, the voter, in Approval, can vote for C and A. Think of a runoff election from this, if multiple votes are allowed. Suddenly, C actually has a chance! All it takes is for the voters who actually prefer C to write in the name and vote for C. The voters do *not* need to vote tactically, they simply vote for both A (the preferred "frontrunner") and C, whom they write in. Or, depending on their perception of the realities, just for C. Without Approval, they must take quite a chance! And if the voters are not sufficiently exercised to write in the name of C, we might infer that they don't have a significant preference. As is characteristic of Range methods, preference strength matters!

Niemi delivered his paper as a speech, originally, so the original paper may be available, that sometimes happens. I'll look for it. Bottom line, from the abstract, Niemi is *not* a source for a claim that Approval is "vulnerable to strategic voting," that is an interpretation, and one which Niemi may very well be avoiding, and, as well, may depend on a variant meaning for "strategic voting," if he does make the claim. Not all "strategic behavior" is "strategic voting."

This is currently moot (thanks to MilesAgain's thoughtful edit) except as background for further edits, and particularly the "sincerity" section which has been created since I started to write this. --Abd (talk) 21:06, 22 January 2008 (UTC)


 * MilesAgain is apparently not just a sock, but a super sock. Who/what else could have taken a claim that a large electorate of rational voters will vote sincerely using Instant-runoff voting, add it to this article, and have Abd not just praise him for his accomplishments, but praise him in the name of NPOV?  And for the record, I'm not trying to be a trouble maker by pointing this out.  ;)  DCary (talk) 04:34, 23 January 2008 (UTC)


 * Well, definitely reinforced with nylon, polyester, or something, knowledgeable about the political spin of the situation, as I've seen with the Instant-runoff voting article. However, that's his privilege, and, as long as he remains civil and doesn't edit war, I'm not complaining. I just name it from time to time. The "claim" isn't in his edit here, unless I missed something, it is found in Nagel, which is truly a sorry article, I will someday, somewhere, deconstruct it in detail. However, I'm confining myself, here, to what bears on actual edits. Niemi may be another story. I'm still looking for a copy of his paper. I now have Nagel and Saari. Saari is interesting. --Abd (talk) 03:04, 24 January 2008 (UTC)

Sincere Vote
I have added a "Sincere Vote" section to the article because the terminology had been introduced in the strategy section. I've initially populated the section with material from Brams and Fishburn. I think that gives decent explanation for how the term is used in the strategy section. As we find reliable sources for alternative definitions or for commentary and critiques of the Brams-Fishburn definition, those could be added.

For the sake of clarity, I glossed over two features in Brams and Fishburn. First, they use the term "strategy" to mean approval vote. Second, they do not define a sincere vote for a voter who is indifferent to all candidates. The latter feature, I suspect, is mostly to simplify the statement of some of their theorems. DCary (talk) 19:28, 22 January 2008 (UTC)


 * The section title and placement seems wrong to me. Since it pertains to the Strategy section, I wonder if it should be there, entitled "Definition of strategic approval voting" or similar. MilesAgain (talk) 19:40, 22 January 2008 (UTC)


 * Well, underneath a whole series of problems in Voting systems articles is the definition of "sincere voting." It crops up with the definition of the Majority criterion. Just the other day, I reread an extended discussion, mostly between myself and Terrill Bouricius, about the definition of the Majority criterion, on the mailing list EMIG-Wikipedia@yahoogroups.com (which is a subcommittee of the Election Methods Interest Group), and it revolves about the definition of "sincere vote." With ranked methods it is simple: it is a vote for any candidate that does not reverse preference. How then to consider Approval? Basically, the early uses of "Majority" as a criterion don't even consider the question. Frequently, the Majority criterion is stated in such a way as to be unspecific about whether the "ranking" involved is actually expressed on the ballot. No method can consider unexpressed preferences! If the Majority criterion is about expressed votes, no problem. *Approval passes.* (And there is a contingent that think this way). If it is about voter preferences, *not necessarily expressed*, then, to avoid the unexpressed problem -- every method would fail -- we have to define how the voter translates the preference onto the ballot. The classic answer is to require that the voter vote "sincerely." But what is that? James Armytage-Green, in a note, struggles with the problem, as I've noted many times. I ended up considering that there were two majority criteria, one involving actual votes, and the other involving mental states, voter preferences, and I did develop a means of transferring the preferences to the ballot that was, in fact, James Armytage Green's, only more directly (and accurately) stated, and it is convoluted enough, involving a double negative, that I really have trouble remembering it! The only way that writers get away with claiming that Approval fails the Majority Criterion, as far as I can see, is that they leave the question of "sincere vote" undefined; if they were to define it, the complexity would be revealed. (But I assume they are sincere, it is easy, particularly for someone very accustomed to considering only ranked methods, to think that it's obvious). Remember, the early writers on this topic were only considering ranked methods, where the translation was simple and did not even need to be stated, it was so obvious.


 * So, in summary, this is actually a very good topic to have a section on, we now have one, thanks to Mr. Cary, and there is actually some peer-reviewed source for it. It is *connected* with strategic considerations for Approval voting, but isn't the same topic. What is a "sincere Approval vote."? It's about time!
 * --Abd (talk) 21:35, 22 January 2008 (UTC)


 * Good point. I put the Sincere vote section first as a lead-in for the strategy section.  I kept them separate partly because I thought they were each already big enough, because each has potential for growing substantially in size with their own subsections (not including examples), and because each is its own interesting issue.   Also, this way a Definition subsection of strategy, if we use one, can focus on just that, defining (multiple ways??) what strategy is, referring as needed to the (various) definition(s), explanations, and examples for sincere voting. That was my reasoning for what I did, but if there is a better way, we should do it. DCary (talk) 22:17, 22 January 2008 (UTC)

I find the paragraphs about a second definition of sincere voting to be confusing. From what I can tell after reading the two referenced sources, is that what is being talked about is a definition of sincere vote that assumes the voter has an approval threshold for the candidates. Having an approval threshold would allow the voter to answer the question "Do you approve of this candidate for [whatever the election is about]?" independently of any decision about how to vote and what other candidates there may be. Given such an approval threshold, the sincere vote would be defined to be the vote that votes for all of the approved candidates. Any other vote would not be sincere by that definition. Voting for an unapproved candidate or not voting for an approved candidate would each be insincere. A similar way of expressing this is to talk about the candidates that the voter finds acceptable, rather than those the voter approves. If that is what is being discussed, then any concerns about stronger or weaker candidates, presumably stronger or weaker in terms of which is more likely to win or be in a tie for winning, is irrelevant to the definition. Calling something both an insincere vote and a sincere strategy is perhaps mixing the two definitions. It becomes particularly confusing to a reader who thinks insincere and sincere are both being used according to the second definition. DCary (talk) 01:23, 8 February 2008 (UTC)


 * That is correct; voting systems analysis, with ranked methods, considers only relative preference, and with relative preference, sincerity is simple to define, even if we use the mental state definition. No absolutes are involved. However, when considering Approval voting, the name of the method has apparently caused many, including experts, to think of "approval" as an absolute. But whether I approve of something isn't an absolute, ever. If you give me a ten dollars, no strings attached, would I approve of this? Wouldn't that depend on the options? Ten dollars, on it's own, Great, I approve. Now, make it an election, and the choices are worth, to me, $10, $20, and $100. Any vote in this Approval election that I cast which is "sincere" by the definition considered is abstaining from the election. No, our approval of an outcome depends upon our perception of the opportunities and probabilities. It is always relative. This is why I actually prefer to avoid the term "approval" when discussing the method. These are votes, and the simple definition of a sincere vote in approval is definitely sincere: voting for a candidate and also for every candidate one prefers to that candidate. This says nothing about where one sets the approval cutoff, and the choice of approval cutoff is totally at the option of the voter, and no choice is insincere. Indeed, this was the intention behind the proposals for Approval Voting by Brams et al, for the method to be strategy free. They were correct, but, I think, this offended the proponents of other methods, and they counterattacked, using this slipperiness of definitions and claiming great "vulnerability" of Approval voting to strategic manipulation. It's quite a story, actually, might make a good article for publication.


 * To repeat this, there is no definition of a sincere vote for a candidate considered in isolation. Generally, when considering the application of the majority criterion to non-ranked methods, writers have had to add to the criterion an assumption that the vote is "sincere," which could be interpreted to mean that it does not conceal a relevant preference. However, instead, what has been done is to define a clearly insincere vote -- which is easy. It is any vote which reverses preference. Then, sincere is defined as a vote which is not insincere. Through this, a vote which conceals the preference involved in the majority criterion is "sincere" if it does not reverse preference. Sure. In a way.


 * Basically, to apply the majority criterion, one must specify how the internal preference is expressed as a vote. And, it turns out, Woodall, who considers that Approval fails Majority, also considers Plurality to fail Majority (as I recall). This is consistent. However, there is a definition that is fairly convoluted that does successfully cause Plurality to pass and Approval to fail... but the whole point of election criteria was to have some objective means of comparing election methods, and if you can manipulate the definition to make a method pass or fail ... it's pretty dangerous.


 * But with strategy, a voter who votes a single preference, accurately expressing that the voter prefers this voter over all others, a clearly sincere vote by the definitions used for judging if Approval passes Majority, is considered to have cast an insincere vote if, supposedly, this voter "actually" approves of another as well. This is the nasty strategic bullet voting. I.e., sincere voting, fully expressing a preference. But what if the voter has no preference? Then, certainly, this bullet voting is mere laziness, and isn't, by definition, strategic voting, it does not improve the outcome for the voter. There is some really, really bad thinking going on.
 * --Abd (talk) 07:20, 9 February 2008 (UTC)


 * As I carefully read the references for the second definition, neither one advocates the second definition. Instead, both rely on the first definition.  Niemi briefly considers the second definition on a what-if basis, and then rejects it.  Both sources discuss the underlying concepts and their implications without redefining "sincere voting".  I think it is appropriate for the article to take a similar approach, unless and until we have reliable sources to support the presentation of an alternative definition. I'll make the changes as part of a larger update to the "Strategic voting" section. DCary (talk) 03:41, 12 February 2008 (UTC)

Strategic voting
I have expanded the section on strategic voting, especially going into more detail on the phenomena previously referenced in the first paragraph in the subsection "Strategy under approval". As part of the changes, I have: DCary (talk) 05:16, 12 February 2008 (UTC)
 * Renamed and added subsection titles.
 * Deleted the second definition of sincere voting, As explained in the Talk:Sincere_voting section, but discussed the underlying concepts of that definition in the subsection "Approval threshold".
 * Deleted the sentence about what is strategy under plurality, in part because that sentence was not particularly relevant in this article, and in part because what was really needed was an explanation of the relationship between strategic voting and sincere voting under approval voting.
 * Moved the assertion about electing Condorcet winners in practice, for the moment largely unchanged, to the section on "Effects on elections", since that assertion involves little regarding strategy.
 * Moved the assertions and refererences about the Burr dilemma, to the section on "Effects on elections", in part because it can be mentioned without specifically discussing strategies. A more detailed discussion of the Burr dilemma and approval voting should delineate that the strategic context for the Burr dilemma is diffferent from what has been detailed so far in the "Strategic voting" section.
 * Provided in the examples subsection more and different scenarios that illustrate some of the items that are previously discussed in the section.

Robert's Rules of Order
The recent addition in the introduction about how RRO disallows approval voting (diff)seems rather questionable for several reasons: DCary (talk) 18:18, 20 February 2008 (UTC)
 * It is vague and perhaps misleading. Just because RRO prescribes something else at the quoted location does not mean RRO in general disallows approval voting.  Just because RRO does not specifically prescribe approval voting anywhere, does not mean RRO in general disallows approval voting.  My understanding is that RRO allows a committee to choose at the committee's discretion whether, when, and how approval voting is used.  In that sense, RRO allows approval voting.  These distinctions should be clarified if the item is kept in the article.
 * Is it OR synthesis? As the above suggests, a broad interpretation of the claim requires an analysis of all of RRO.  As such, the claim as currently worded may require reference to a reliable source that makes the claim.  Simply quoting a snippet of RRO is just bad OR.
 * Is it really important? Is there a reason why this alleged disapproval is worth mentioning?    We don't fill every voting system article with particular descriptions of the places where the system is not prescribed.
 * It doesn't belong in the introduction. If the item in some form is kept, I don't see a justification for including it in the introduction.  The "Other issues and comparisons" section would be a better place.  I don't recommend starting a "Situations where approval voting is not used" section, or even a "Situations where approval voting is allowed but not prescribed" section.


 * Right, doesn't belong. The quote seems to be about overvotes, same as any existing one vote election system. It should be removed. Tom Ruen (talk) 18:23, 20 February 2008 (UTC)
 * A new article is created as well for Overvote. Tom Ruen (talk) 18:25, 20 February 2008 (UTC)

Two to tango.
Excuse me if this is either off-topic, or irrelevant. I appreciate the comments that have been made here, but (and I have not read them all, nor claim to understand all that I've read) isn't this discussion missing the role of the parties themselves? Their 'strategic' or tactical choices have to be taken into account, in my opinion. In the example of the city and rural republicans, if the schism between them is so great that voters will not consider supporting them, then it is perhaps 'correct' that a voting system disfavor them. —Preceding unsigned comment added by 72.70.254.248 (talk) 18:03, 24 February 2008 (UTC)

It would be great to increase the presentation in the article about candidate or party behavior under approval voting, as long as the information is based on reliable sources and not  original research, especially not personal or speculative OR. The only candidate strategy issues that are currently explicitly addressed in the article are about the generalized Burr dilemma. The example given at The dilemma with Approval Voting could be interpreted as illustrating the generalized Burr dilemma, but the presentation of the example devolves into an off-topic OR discussion. This talk page is not a general forum for discussing approval voting, but a place for editors to discuss how to improve the article. See WP:talk for additional guidelines. DCary (talk) 02:45, 4 March 2008 (UTC)

Effect on Elections
This talk section is created for discussing changes to the corresponding section in the article.

I've changed the description of the referenced Science 2001 article by Brams and Herschbach in order to improve accuracy and keep the focus on the the effect of approval voting on elections. The article is an editorial, not the presentation of a study. No study is even referenced. The editorial does not use the term "fairer" or "fair". While the article discusses weaknesses of plurality, Borda, and instant run-off voting, it does not offer an explicit assessment of preference voting in general. I'm not sure exactly what the second sentence is saying, but it does not appear to have any clear basis in the referenced article. DCary (talk) 02:03, 4 March 2008 (UTC)

Criteria
This is a section for posting arguments about the criteria table, that is, whether approval voting satisfies the criteria under different voter models. Obviously, if anybody wanted to make a WP:POINT, they could insist that all cells of the table which do not have an external WP:RS be reset to white. However, I hope that we can all agree that the goal of having a filled-in table is superior to the goal of requiring peer-reviewed sources for mathematical facts. So let's argue the mathematics AND/OR the sources here, and only white out squares when neither of those give us a consensus answer.

It would also be OK to argue that we should tighten the Nash view to include only certain kinds of Nash equilibria. For one thing, I think it would be OK to iteratively remove weakly dominated strategies; since we're working in perfect information, that simple step ensures semi-honest ballots.

First question: does the Nash Equilibrium model of approval satisfy the Consistency and Participation criteria? Homunq (talk) 16:54, 22 June 2009 (UTC)


 * The Condorcet criterion and the participation criterion are incompatible. So when the voters have preferences and the used election method satisfies the Condorcet criterion, then this method cannot satisfy the participation criterion anymore. Markus Schulze 18:03, 22 June 2009 (UTC)