[EM] Let's play Jenga!

[Ross Hyman]

Hi Kristofer,I found Warren Smith's votedesc.pdf document on M. Schulze's site: http://m-schulze.9mail.de/votedesc.pdf
This is a great document.  I think it should be on the Election Methods website, especially since most of the other links to election method descriptions are broken.

The variation I am proposing to Warren's Maxtree method is to constrain the form of the spanning tree to a directed chain (or whatever the official name is) A>B>C>D....  and then maximize the minimum link.  I haven't had time to think about it too much but I am hoping the method will satisfy local independence of irrelevant alternatives. 

What's up with the election methods list?  I have not seen my posting or your response to it on the archive, which is what I read.  I don't get the emails.  The last posting in the archive is from Sept 28. 

Best,Ross

Repeatedly remove the weakest link whose removal leaves at least one
ranking of all of the candidates in which there is a direct win for the
higher candidate over the next lower candidate.  When only one such
ranking exists, elect that ranking of candidates.
This method is different from Tideman Ranked pairs.
Consider the pair ordering B>D, B>A, C>B, D>C, C>A, A>D.
The above method produces: D>C>B>A. The Tideman order is C>B>A>D.  The
Tideman order is better. The Schulze winner is also C.

Warren's Maxtree method is another Ranked-Pairs-like that it might be
interesting to investigate. The method's logic is akin to:

- Ranked Pairs is similar to Kruskal's algorithm for finding a minimum
spanning tree in an undirected graph.
- But the graph induced by the Condorcet matrix is directed.
- So use an MST algorithm for weighted graphs instead.
- This algorithm is Chu-Liu-Edmonds and the method becomes max-tree.
(See Warren's votedesc.pdf for more information)

I've never got around to implementing it, though.

[Kristofer Munsterhjelm]

Post by Ross Hyman
Hi Kristofer,
http://m-schulze.9mail.de/votedesc.pdf
This is a great document. I think it should be on the Election Methods
website, especially since most of the other links to election method
descriptions are broken.
The variation I am proposing to Warren's Maxtree method is to constrain
the form of the spanning tree to a directed chain (or whatever the
official name is) A>B>C>D.... and then maximize the minimum link. I
haven't had time to think about it too much but I am hoping the method
will satisfy local independence of irrelevant alternatives.
What's up with the election methods list? I have not seen my posting or
your response to it on the archive, which is what I read. I don't get
the emails. The last posting in the archive is from Sept 28.
Best,
Ross

electorama.com has had some problems with its mailing list
administration lately. First the election-methods list disappeared from
http://lists.electorama.com/listinfo.cgi some years ago, and then this.

http://lists.electorama.com/pipermail/election-methods-electorama.com/
states

"Note:The archive search index was last rebuilt at Friday, 29 Sep 2017
21:39:45 PDT. Any postings after that will not be found by a search.
Index rebuild is usally done once every 24 hours for this list. You can
use a "View by date" link below to access more recent postings. "

So it seems like some kind of scheduled program has stopped working.

Since the electorama server has had problems in the past as well, it
might be an idea to move it elsewhere... Same with the electorama wiki;
both could use some maintenance, it seems like; although I haven't
recently checked if the TeX problem has been fixed.

On-topic, it could be interesting to have a proof of what kind of
chain/DAG methods satisfy LIIA. LIIA doesn't have to imply clone
independence because Kemeny passes the former and fails the latter, at
least.
----
Election-Methods mailing list - see http://electorama.com/em for list info

[Ross Hyman]

Well, it doesn't satisfy LIIA.A>B, B>C, A>D, B>D, C>A, D>CThe ranking is B>C>A>D.Remove D and the ranking becomes A>B>C.

Post by Ross Hyman
Hi Kristofer,
http://m-schulze.9mail.de/votedesc.pdf
This is a great document.  I think it should be on the Election Methods
website, especially since most of the other links to election method
descriptions are broken.
The variation I am proposing to Warren's Maxtree method is to constrain
the form of the spanning tree to a directed chain (or whatever the
official name is) A>B>C>D....  and then maximize the minimum link.  I
haven't had time to think about it too much but I am hoping the method
will satisfy local independence of irrelevant alternatives.
What's up with the election methods list?  I have not seen my posting or
your response to it on the archive, which is what I read.  I don't get
the emails.  The last posting in the archive is from Sept 28.
Best,
Ross

electorama.com has had some problems with its mailing list
administration lately. First the election-methods list disappeared from
http://lists.electorama.com/listinfo.cgi some years ago, and then this.

http://lists.electorama.com/pipermail/election-methods-electorama.com/
states

"Note:The archive search index was last rebuilt at Friday, 29 Sep 2017
21:39:45 PDT. Any postings after that will not be found by a search.
Index rebuild is usally done once every 24 hours for this list. You can
use a "View by date" link below to access more recent postings. "

So it seems like some kind of scheduled program has stopped working.

Since the electorama server has had problems in the past as well, it
might be an idea to move it elsewhere... Same with the electorama wiki;
both could use some maintenance, it seems like; although I haven't
recently checked if the TeX problem has been fixed.

On-topic, it could be interesting to have a proof of what kind of
chain/DAG methods satisfy LIIA. LIIA doesn't have to imply clone
independence because Kemeny passes the former and fails the latter, at
least.

[Toby Pereira]

I've never seen an explanation of why LIIA is something you'd want to achieve in a method. It seems to be a criterion for the sake of it.

Sent from Yahoo Mail on Android

On Sun, 8 Oct 2017 at 14:25, Ross Hyman<***@sbcglobal.net> wrote: ----
Election-Methods mailing list - see http://electorama.com/em for list info

[Rob Lanphier]

I've never seen an explanation of why LIIA [(Local Independence from
Irrelevant Alternatives)] is something you'd want to achieve in a method.
It seems to be a criterion for the sake of it.

Do you believe standard IIA[1] is something worth achieving?

Rob

[1]: https://en.wikipedia.org/wiki/Independence_of_irrelevant_alternatives

[Toby Pereira]

I'd certainly say that IIA is worth achieving if it can be done. But while I imagine that LIIA is supposed to go some way towards full IIA, I think that this is an illusion. Let's say we have an A>B>C>A cycle. If A wins, then the result order under an LIIA compliant system must be:
1. A2. B3. C
Adjacent candidates in the order will never swap over if other candidates are removed. If A is removed, B wins. If C is removed, A wins. But the question is: what is so special about adjacent candidates? This simply means that if B is removed, then C wins. This is arguably worse than adjacent candidates swapping because a candidate has gone above a candidate that was originally two places above them.
Failure of IIA means that there is the possibility that a pair of candidates will swap their order with the removal of a third candidate. Passing LIIA simply determines which candidates might swap over, but without any justification of why these swaps are better than other swaps.
Toby

From: Rob Lanphier <***@robla.net>
To: "***@yahoo.co.uk" <***@yahoo.co.uk>
Cc: "***@sbcglobal.net" <***@sbcglobal.net>; Kristofer Munsterhjelm <***@t-online.de>; "election-***@lists.electorama.com" <election-***@lists.electorama.com>
Sent: Sunday, 8 October 2017, 17:41
Subject: Re: [EM] Let's play Jenga!

On Sun, Oct 8, 2017 at 6:44 AM, Toby Pereira <***@yahoo.co.uk> wrote:

I've never seen an explanation of why LIIA [(Local Independence from Irrelevant Alternatives)] is something you'd want to achieve in a method. It seems to be a criterion for the sake of it.

Do you believe standard IIA[1] is something worth achieving?
Rob
[1]: https://en.wikipedia.org/wiki/Independence_of_irrelevant_alternatives
 

[Ross Hyman]

I think it is useful in a situation where you electing a delegate and a first and second alternate.


On Sunday, October 8, 2017 2:07 PM, Toby Pereira <***@yahoo.co.uk> wrote:


I'd certainly say that IIA is worth achieving if it can be done. But while I imagine that LIIA is supposed to go some way towards full IIA, I think that this is an illusion. Let's say we have an A>B>C>A cycle. If A wins, then the result order under an LIIA compliant system must be:
1. A2. B3. C
Adjacent candidates in the order will never swap over if other candidates are removed. If A is removed, B wins. If C is removed, A wins. But the question is: what is so special about adjacent candidates? This simply means that if B is removed, then C wins. This is arguably worse than adjacent candidates swapping because a candidate has gone above a candidate that was originally two places above them.
Failure of IIA means that there is the possibility that a pair of candidates will swap their order with the removal of a third candidate. Passing LIIA simply determines which candidates might swap over, but without any justification of why these swaps are better than other swaps.
Toby

From: Rob Lanphier <***@robla.net>
To: "***@yahoo.co.uk" <***@yahoo.co.uk>
Cc: "***@sbcglobal.net" <***@sbcglobal.net>; Kristofer Munsterhjelm <***@t-online.de>; "election-***@lists.electorama.com" <election-***@lists.electorama.com>
Sent: Sunday, 8 October 2017, 17:41
Subject: Re: [EM] Let's play Jenga!

On Sun, Oct 8, 2017 at 6:44 AM, Toby Pereira <***@yahoo.co.uk> wrote:

I've never seen an explanation of why LIIA [(Local Independence from Irrelevant Alternatives)] is something you'd want to achieve in a method. It seems to be a criterion for the sake of it.

Do you believe standard IIA[1] is something worth achieving?
Rob
[1]: https://en.wikipedia.org/wiki/Independence_of_irrelevant_alternatives