[EM] Re: Richard's criteria

[Craig Carey]

These rules are likely to be rejected for being undesirable and unfair.
Most likely small examples get into disputes with some rules (OSSIPOFF
criteria) and then whichever rule has to be rejected. That particular
step should have been done near 2000 AD.



At 2004-02-13 09:38 -0800 Friday, Richard Moore wrote:
>Last night I wrote that I would add a provision to my definitions to
...
>SFC:
>For any set of ballots that either are ranked ballots or allow 
>rankings to be uniquely calculated, let X be the Condorcet winner of 
>this set of ballots. 

No need to read further since the Condorcet winner is not known to be
desirable. I recall Mr Schulze asserting thaet he was unaware that
the Condorcet winner can be shown inconsistent with monotonicity when
there are only 4 ballot papers. 

>Let Y be any other candidate. Transform all of 
>the ballots into ballots that can be accepted in the method under 
>consideration, subject to the following two constraints: (1) If X is 

To use the word "transform" just seems to be most unacceptable. It
means any change. There surely is no need to have a wording that
allows everyone a way to ruin the definition with any idea out of all
possible ideas.

As I mentioned earlier at this mailing list, the Block Vote ballot
papers can be described as having a 1st prefernece, and that
preference is a set of the names of candidates. There is no 2nd
preference.

E.g.
   Block Vote:    ({A,B,C,E})
   STV:           (A,B,C,E)

Mr Moore is being unclear over these cases. 

 (1) Preference X is the last preference showing on the paper;

 (2) The ballot paper does not name candidate X with a preference.

It looked like Richard was going to construct ideas out of the
nonsense wordings of Mr Schulze. Instead he is making it an opportunity
to mallot readers with ideas including "'over' means 'over' of course".

Mr OSSIPOFF was trying to be inadequately precise on what the word
"over" meant in 2000 AD. Currently Mr OSSIPOFF and Mr Schulze are
withholding the exact same idea: the 5 numbers, but Mr Ossipoff uses
the word "over" and Mr Schulze uses the words "strictly prefer".

I assume that the meritless ambition of making people be selected
wrongly, is what they want to withhold.



>voted higher than Y on a ballot, then for any two candidates A and B, 
>A is voted higher than B on the new ballot if and only if A is voted 
>higher than B on the original ballot, and (2) If X is not voted higher 
>than Y on a ballot, then for any two candidates A and B, A is voted 
>higher than B on the new ballot only if A is voted higher than B on 
>the original ballot. However the ballots are transformed subject to 
>these constraints, Y cannot win an election with the resulting set of 

The whole rule MUST BE REJECTED now since it was said that Y cannot
win. Note that the number of winners can equal the number of candidates.
Also the rule would be not be infinitesimal and fair if it is stopping
a candidate from holding its win-lose state stable.

Richard can't be any good at designing rules. Since Mr OSSIPOFF had
wording so defective that he might not even have outlines of rules,
it could take a large number of messages from Mr Moore before some
actual rules appear.

As far as I know, Mr Schulze has held an untrue belief that he has
at times, understood some rules of Mr Paielli and OSSIPOFF. He didn't
mention the OSSIPOFF rules when producing a PDF making "it passes"
arguments about an undefined preferential voting method.

The words "voted higher than" have no meaning. There is only 2 things
that provide numbers about preferences: their "rank" and their "index".
Those 2 numbers are the same. So the term should get interpreted to
mean "ranked higher than". So obviously the preference of two listed,
that is nearer the end, will be the higher preference. Of course,
their is no failure to appreciate Mr Moore's unarticulated intent
to have the public found to be in the wrong if not thinking
untruthfully. MIKE OSSIPOFF has some "I can't change" belief that
substitutes for an acceptable statement of the reasoning. However
Mr Moore seems to have a purpose of correcting MIKE OSSIPOFF's
mistakes.

Actually there seems to be a big new mistake MIKE didn't make, of
allowing any possible transform, to be made to the papers. To
simplify things, don't allow any sympathizer of Mr Moore & co. to
define the transforms. Etc. It sounds like Richard was writing to
his friends rather than getting OSSIPOFF's electionmethods.org
"criteria" actually defined. Those are not the rules actually
used in the (never accessible proofs): for those he seem to access
memory and ignore the wordings of the online versions. That could
be why the final wordings that Richard Moore produces are possibly
going to be kept offline.

Unless this goes downhill, Richard Moore will be providing a
demonstration of the usability of his quantifier logic definitions.

They must correctly handle internal ties correctly. The best way
may be to be explicit on making the inequalities of the method be
strict and lax. The effort that Mr Moore puts into clarifying the
rules does not count: a mistake at a single election point can
result in the loss of a rule. They are sure to be discarded anyway
for being unfair and not able to be worded in a way that makes them
more desirable, so maybe to save them they have to be shown to be
consequences of fairness.


>ballots in the method under consideration.
>
>GSFC:
>For any set of ballots that either are ranked ballots or allow 
>rankings to be uniquely calculated, let X be a candidate in the Smith 
>set of this set of ballots. Let Y be a candidate outside of that Smith 
>set. Transform all of the ballots into ballots that can be accepted in 
>the method under consideration, subject to the following two 
...

Those are not rules but deliberately unclear on what the weighting
numbers for the 5 categories are.

(...X...Y...)
(...Y...X...)
(...X...)
(...Y...)
(...)

Mr OSSIPOFF seemed to be hoping that Mr Moore would actually define
the rules. I guess MIKE won't be able to "understand" the final
rules, not that I suppose that Mr Moore can get them defined.

Mr Schulze is still hiding what his 5 real numbers are using the words
"strictly prefer" and he also hacked up his algorithm so that no one
can get the numbers from that.

MIKE OSSIPOFF has been doing exactly the same but instead using the
word "over". Now Mr Moore copying MIKE's regime of being excessively
unclear on what "over" means. To the competent method designer it would
mean that the method is too unfair to put time into. Mr Schulze has
a now GPL-ed method parameterized by 5 real numbers.

Maybe the 5 numbers are known for the original Condorcet method. But
when Mr Moore replaces that with the most unclear word "over", then
he is not saying that people must use the same wrong numbers.

I am expecting that Mr Moore has no intention of producing real
definitions.

----------

The rules should run correctly for all numbers of winners or else they
would be rejected.

Here is some of the text:

| SFC:
| For any set of ballots that either are ranked ballots or allow 
| rankings to be uniquely calculated, let X be the Condorcet winner of 
| this set of ballots. Let Y be any other candidate. Transform all of 

Here is an election:

  Number of candidates = 1

  Number of winners = 0

  The ballot papers (only 2 kinds of paper)
  -------
   z * ()
   a * (A)
  -------
  "a" and "z" are Real numbers.

Why is the rule internally computing a 1 winner election while it is
supposed to be testing a 0 winner election. The wording does not
rule out zero winner elections.

That sounds so strange that it would be guessed that the Condorcet
method of the SFC "criteria" has to find the same number of winners
as the method being tested has to. But that is presumably worse since
(if) Condorcet malfunctions in 0 winner 1 candidate elections.

To summarise, Mr Moore should expect to be ignored after
 putting this idea into the things:

  * apply to most data any transform that any of my critics can think of.


That is just done to avoid making the 1st preference be a set.

That would get the checkboxy Approval past a "Truncation Resistance"
test since every candidate of a marked checkbox is named in the
set of the 1st preference.

MIKE has to do better than what he did in 2000 AD, which seemed to be
to 'visualize' that Approval got STV papers and let it affect much of
his writing, while he would switch to denying that Approval got STV
papers if asked. That is similar to the 'sincerity' idea except that
the imagination that the papers are STV papers, got replaced with
claims that voters are living people. Real people have real names and
I am confident that MIKE never can answer questions on surnames.

MIKE found it easier to imply that he had an army of Oss-spirits
who could could say if a ballot paper (in a 0-winner election say) is
sincere or not, instead of having the 1st preference name more than
one candidate. At this point, I guess that Mr OSSIPOFF simply won't
define a rule correctly. Just how hard can it be to extrapolate the
definition of Truncation Resistance over to the new case of their
being only 1 preference ?. However that is a rule that MIKE (and
Richard too ?) just don't like to write on.

Is that "over" a "strictly prefer" type of "over" or just an
OSSIPOFF "over" ?. One is associated no right to know.

Keep going Mr Moore. Naturally all your efforts would be rejected at
the end for requiring unfairness. But I need some real attempts to
get MIKE's definitions salvaged and floated up.



Craig Carey

http://groups.yahoo.com/groups/politicians-and-polytopes