[EM] Participation axiom re 2nd preference alterations

[Craig Carey]

 >From:  Markus Schulze <markus.schulze at a...>
    X-Mailer: Windows Eudora Light Version 3.0.1 (32)
 >Date:  Thu Sep 13, 2001  11:38 am
 >Subject:  Re: IFPP and Schulze comments (re [EM] Craig's posts
 >
 >Dear Craig,
 >
 >you wrote (13 Sep 2001):

Can the "you" be changed here, MrSchulze?. Eudora 5.1 is nice except
it is atrocious over HTML-izing. Every at the end of most edits, I
do this:  Edit / Text / Clear Formatting.
That is astonishly inconvenient, but what the heck, that is the v. best
Eudora can do. From time to time they make it worse to remind us how
it is that their use HTML-ization and do not test having their clients
emit "plain text". Also the problem is absent from version 3. But I
expect that a good reason for quitting version 3 of Eudora is that it
is more prone to losing all data. Even in version 5.0.2, its search
feature could not find all matching messages.

Eudora can seem a lot worse if these lines are not added to the
eudora.ini file:
-----------------------------------------------------------
[Settings]
AlwaysExcerptResponse=0
ConvertFormatFlowedtoExcerpt=0
InterpretFormatFlowed=0
WordWrapOnScreen=0
WordWrapMax=30000
ReplyAttribution=At %1, %2 wrote:
ReplyAllAttribution=At %1, %2 wrote:
MainLexFiles=british.tlx,british.clx
-----------------------------------------------------------

See that the Reply* lines get rid of the "you wrote".
The WordWrapMax stops wrapping a surprise event of wrapping at
column 78 at the instant of sending. People would prefer lines
82 characters long than lines that got unexpectedly wrapped.

The stupid excerpt bars can are not the result of HTMLizing but the
result of a new special tag made up Eudora. Use of Ctrl-C, Ctrl-V
and a text editor can get rid of that. Astonishingly inconvenient,
but, 'hey, what the heck, it is the best Eudora's programmers can
do'. It is the same with STV. Life stops once it is coded in an
algorithm, which seems to be first done in the Victorian era.


 >> Markus wrote (7 Sep 2001):
 >> > IFPP also violates the participation criterion. If 5 BCA voters hadn't
 >> > gone to the polls, then C would have been elected. Thus these 5 BCA
 >> > voters changed the winner from C to A.
 >>
 >> Where is the example?
 >
 >I am talking about Rob's example
 >(http://groups.yahoo.com/group/election-methods-list/message/8046):
 >
 >   45:A>B>C
 >   22:B>C>A
 >   33:C>B>A
 >
 >The IFPP winner is A. However, if 5 BCA voters hadn't voted, then this
 >example would have looked as follows:
 >
 >   45:A>B>C
 >   17:B>C>A
 >   33:C>B>A
 >
 >Now the IFPP winner is C. Thus, by going to the polls these 5 BCA voters
 >changed the winner from candidate C to candidate A.
 >


I mentioned this privately then resubscribed so I could post this.

Under my axioms, both P1 and the stronger P4, this is prohibited:

    [ 22:(BCA), A wins ] <--> [ 17:(BC*) + 5:(B), C wins ]

But the example you gave is allowed. That is an instance of this:

    [ 22:(BCA), A wins ] <--> [ 17:(BC*), C wins ]

Suppose the extra 5 votes for B were 'going off to D' and making
D win, and thereby stopping votes leaving D and going towards C.
Then it is entirely appropriate that the IFPP behaviour show that
did occur can. If the rule is rejected when >=4 candidates, it
is a test that applies when there are 3 candidates.

For what its worth, the preceding list that is (B), could instead
be 300,000 preferences long. While it is still only 5 votes that
is leaving B, it could have been that A and C were almost tied
and it would only take 10**(-3) of a vote to shift from A to C,
to cause A to beat C.

Under such a view it looks a lot more obvious that you have a
wrong principle and it can be rejected by just using that STV-ish
idea of transfers where each use of the paper of positive weight
has to not exceed the original weight of the paper.

The participation axioms is not questioned when it is making
statements about just the first candidate. In that case I say
that it is too weak, but such a rule extended to allow vote
shifting rejects the Alternative Vote (i.e. the line not only
passes through the vertex for paper (C), but it can scribe through
a simplex imposing the same condition ('reducing support for C
can't turn it into a winner').

A multiwinner extension would perhaps involve my binary number
satisfaction function.

Can subscribers post up defences for me so I do not need to
subscribe. If there was more success amongst the subscribers then
I could leave all the wrong criticisms of my IFPP theory of voting
unanswered. Is Schulze relying on a paper of Mr Ron Holzman for the
undesirable definition of the participation axiom/criteria?.

This list seems to be one of the foremost places to collect errors,
e.g. this new "interpretation of 5 votes removed from Rob's example"
error.


G. A. Craig Carey
Auckland
http://groups.yahoo.co/group/politicians-and-polytopes