Talk:Independence of clones criterion

2007 Merge comment

mergefrom Clone (voting) - I like the more-straightforward presentation of the other page, perhaps we could re-organize this one to carry the same information 24.19.11.19 09:24, 31 August 2007 (UTC)

Number shifting and second example

I shifted the numbers in the first example. I may be completely ignorant here, so feel free to revert me, but it seems that when alloting 55 votes versus 45 votes, it's much harder to make clear that candidate A(clone) is the same, in every way, as candidate A (since 55 cannot be split evenly, and in fact wasn't split evenly in the example). Switching the votes from 55-45 to 100-75 keeps the principle intact, while making it more obvious that A(clone) isn't just a third candidate, but is an absolute clone of A in every way. Of course, 100+75 does not equal 100, which makes "my" set less pretty than the old one. It also means that the numbers in the first section are different than those in the second section - which, of course, can be easily fixed (or ignored, if my numbers aren't ideal) --Badger Drink 19:58, 18 October 2007 (UTC)

And, as a third-party possibility, we could consider going back to the 55-45 numbers, and changing them from a vote-count to a percentage, thus allowing A and A(clone) to end up with 27.5% of the vote. --Badger Drink 20:00, 18 October 2007 (UTC)

I rewrote the article (including the examples) to match the true definition of independence of clones, from Tideman's 1987 paper. Clones do not need to be identical. They just need to be similar, so that voters can be expected not to rank a dissimilar alternative between (or equal to) similar alternatives. Typically it should be easy to find similar alternatives that are inferior, as in the new example for Borda. From my point of view, this easy way to manipulate Borda should eliminate Borda from consideration for use in public elections. SEppley (talk) 22:08, 11 March 2012 (UTC)

This article should be either rewritten or deleted

This article is totally unsourced, is pure speculation and is completely self-referential. It asserts that "The independence of clones criterion states that the addition of a candidate identical to one already present in an election will not cause the winner of the election to change", and then sets out to prove that assertion mathematically. Providing proofs for esoteric statistical theories is not the function of an encyclopaedia. Either this argument is of the author's own invention, in which case it is original research, or else it is copied from somewhere else, in which case it is plagiarism. I happen to know that it derives from the work of Nicolaus Tideman, but the reader is not informed of this. Either this article must be rewritten as a properly referenced discussion of the issues, or I will nominate it for deletion. Intelligent Mr Toad (talk) 08:36, 5 May 2008 (UTC)

Tideman's article has been cited, and the criterion is of immense practical importance due to the ease of strategic nomination. (That is, a small minority typically has the power to nominate additional alternatives, and it is typically easy to find similar alternatives.) So, no need to delete the article! However, the definition of clones was not Tideman's definition, so I've corrected it: Clones do not need to be identical; they just need to be similar. (Voters can be expected not to rank a dissimilar alternative between similar ones.) SEppley (talk) 21:54, 11 March 2012 (UTC)

Woodall's terminology

As far as I remember correctly, Woodall's "clone-winner" criterion says that replacing a candidate A, who was elected with a positive probability, by a set of clones A(1),...,A(m) must not decrease the probability that the winner is chosen from this set. As far as I remember correctly, Woodall's "clone-loser" criterion says that replacing a candidate A, who was elected with zero probability, by a set of clones A(1),...,A(m) must not change the result of the elections at all. In any case, as far as I remember correctly, those papers, where Woodall introduces these terms, have not yet been published. Markus Schulze 16:38, 26 July 2008 (UTC)

A Head in the Polls

Does the A Head in the Polls episode present or demonstrate (in the ideal case) some of the kinds of of problems that arise with this, and if so, which ones? —Preceding unsigned comment added by 63.194.187.135 (talk) 21:49, 5 November 2008 (UTC)

Borda Count Example

In this example I would not call B(clone) a true clone of B, since B is universally preferred to B(clone) (unlike in the previous example, where A(clone) was preferred to A by half of voters). If I go through the calculations again but giving B(clone) preference over B half of the time, I don't see any change in winner. Am I missing something? 131.215.169.186 (talk) 05:29, 7 December 2009 (UTC)

You were correct, given the incorrect definition of clones that the article used until I corrected it today. Now it matches Tideman's definition, where clones can be similar; they do not need to be identical. This makes the criterion stronger--any voting method that satisfies independence of clones also satisfies independence of identical clones--and much more practical too, since it is often easy to find and nominate similar alternatives. I've also rewritten the examples. SEppley (talk) 22:00, 11 March 2012 (UTC)

Common variation of Instant Runoff fails independence of clones

The article claims Instant Runoff is independent of clones, but in many of the communities that have adopted it each voter may rank only a small number of candidates. (Three seems to be a common limit.) Instant Runoff fails independence of clones under this condition. (So would any voting method that limits how many candidates each voter can express preferences for.) SEppley (talk) 17:36, 13 March 2012 (UTC)

You're correct that the Limited Preferential Voting variant isn't immune to clones. The variant is briefly discussed on the IRV page as it's used in Papua New Guinea, but I don't see any discussion of its vulnerability to clones there, which if you can cite a source, is probably worth adding.

That said, IRV is still vulnerable to strategic nomination, just not of clones. Imagine a semi-clone that beats the original winner candidate on first preferences, but isn't as strong on second preferences as the original is, and therefore loses. This was actually the argument some people were making against nominating Hillary Clinton in the recent US Democratic primary, (which, as it's a primary, is a runoff election, so has somewhat similar strategic implications, with the proviso that not being instant, Some Things May Differ) as she wasn't polling as strongly against the eventual Republican nominee at that stage. I imagine this vulnerability is why it fails independence of clones when you add the limiting rule to it. 54x (talk) 16:36, 22 October 2016 (UTC)

Susceptibility to clones

While Borda may be susceptible to clones, it's no less susceptible to it than first past the post. Just take the 1992 United States Presidential Election as an oft-cited example. Many feel GHWB would have won handily over Bill Clinton were it not for the effective clone candidacy of H. Ross Perot. - KeithTyler 10:29, 30 November 2014 (UTC)