[EM] Rules not to consider 'voters', FBC defined wrong

[Craig Carey]

[One of the thoughts at this mailing list was attracting my
  attention: the unpromising FBC rule of Mr Mike Ossipoff]

Partly since I was caught out in an error, and partly because
of the plainly false comment from Mike unfortunate reference to
my sincerity in correctly arguing that FBC was badly defined
and undefined, I investigated this FBC in such detail that I am
now the expert on it. So, Mr Ossipoff, the education of me can
end. In any case, formula do not need to 'understand'. I ask
you to stop telling me that I do not understand. The emphasis
should properly be on defining FBC. It is not defined although
the beginning of something is there.

I can portray FBC to be stupid if used in a society. Of course
it is wide open to such accusations: consider, a government
tester can use FBC to pass or fail any given method (provided
the set of winners alters (etc.)). So the discriminating power
is not high, is it Mike. Don't simply deny this (M.O.), I have
already written a very detailed example. I have other examples
too, that prove points that you might casually say I erred
over. **9**

It is a rather obnoxious rule in that it implicitly requires each
reader of the rule to try and figure out what the rule designer
could only have been thinking of. That idea, "could only have
been thinking of", will be a familiar one to semantics experts
in the societies of English speaking lands, including nations
in the Commonwealth. Where were you born Mike?.

Mike might not realize how unpleasant a task that is: using all
available evidence to reconstruct the reasoning for the bad
definitions, and the distracted or missing responses. The rule
can be written down, and I suggest Mike alone do that.

I don't see that anybody should be anything other than critical
towards the "understand it as is or else I'll have to write again"
microcosm of errors, the FBC criteria.

Here is less than <1/2 of what I have written. The other have
has a lot of real life examples to get the subscribers into the
mood of figuratively aerial fire bombing defective ideas of people
that want to think rules that test feelings ("likes") [that hidden
defect of FBC which is that all methods it tests do not make it
  desirable (or it reasonable?), for voters to vote tactically]



At 04:47 16.03.00 +0000, MIKE OSSIPOFF wrote:
 >
 >>At 10:36 15.03.00 -0800, Bart Ingles wrote:
 >> >
 >> >DEMOREP1 at aol.com wrote:
 >> >>
 >> >> Simple Approval Voting has the rather major defect of
 >> >> having a later vote cancel out an earlier vote.
...
 >>The Approval Voting method, according to Mr Ossipoff, needs
 >>  to be considered undefined when this is not true: there
 >>  are very few candidates and only one winner is to be
 >>  elected. Is that a
 >
 >I didn't say anything resembling that. Craig, I'll thank you
 >not to attribute to me things that only you've said.
 >

I retract my statement, that you had constrained the number of
winners allowed for the Approval Voting method ,to be 1. I
shall call that an error, and I should regard it as a mistake.
In a way analogous to my being unable to stop the entry of
complex numbers into formulae devised only to accept real
numbers. Similarly any statement by you that FBC is defined
well (and that the problem is instead that I do not read and
understand) does not, since false, imply in itself that all
readers of this mailing list are able to fix FBC if they tried.
Is there some way to get FBC fixed, without using truthful
argument?.

  FBC hasn't been getting defined. If the FBC wording (below) is
checked, it is rather obvious that either (a) it doesn't test
the Approval Voting method in general, even in just the one
winner case (which will be the consensus of those that have
read the latest FBC definition), or alternatively, (b) the use
of the word voter is not an accident, but the mistake there is
not accidental. Certainly fixing the wording of FBC is
possible.

  Mr Ossipoff is not fixing the FBC rule. I hope that changes.
At worst, no response could be made to each of the arguments. I
shall mark those strong objections I not, with a **nn** (just
to keep the numbers up after the arguments are decimated with a
precision of reason that is breathtaking to read of).

  FBC is certainly constrained to disregarding methods that are
returning more than 1 winner, according to some of the text in
I saw in 1 message. As far as I am aware, that constraint
wasn't present at all in early March 2000 version. I am not
sure. I believe that a totally major change to the rule
occurred but the definition remained unaltered. **1**


Here is the FBC definition (without amendments).

 >>At 17:01 07.03.00, MIKE OSSIPOFF wrote: ...
 >>  >>At 12:23 05.03.00 , MIKE OSSIPOFF wrote:
 >>  >>...
 >>  >> >> >> >> At 16:17 28.02.00 +0000, MIKE OSSIPOFF wrote:
 >>  >> >> >> >> >Favorite-Betrayal Criterion (FBC):
 >>  >> >> >> >> >
 >>  >> >> >> >> >By voting a less-liked candidate over his favorite, a
 >>  >> >> >> >> > voter should never gain an outcome that he likes better
 >>  >> >> >> >> > than any outcome that he could get without voting a
 >>  >> >> >> >> > less-liked candidate over his favorite.
 >>  >> >> >>...




This quote just below is from a recent message and it confirms
that Mr Ossipoff was right in the dispute of the previous
message, i.e. did not regard the Approval Voting method as
being constrained to 1 winner (in the 2nd of these 3 cases:
personal considerations, regarding that limits the
consideration of others, and the actual mathematical reality
[e.g. extension to complex numbers, with inequality redefined,
etc.]).

At 04:06 11.03.00 +0000, MIKE OSSIPOFF wrote:
...
 >>... I note that old FBC definition did not constrain the
 >>    number of winners. Mr Ossipoff said he regarded the
 >>    number of winners to be 1. If that is true then it
 >>    should be in the definition.
 >
 >We're talking about single-winner methods. If Approval, or any
 >method returns a tie, then some tiebreaking procedure would
 >be used to choose a single winner. To the extent that they're
 >applicable, the existing tiebreaking rules already on the
 > books are good enough.
 >
 >I never said anything about whether the method being tested
 >by the criterion is able to elect more than one winner
 >(temporarily, till some tiebreaker is applied). If you feel
 > that that issue is somehow relevant to the criterion's
 > meaning then you'd need to show that.


Ties are boundary values (here).


The word "favorite" is sort of fully undefined.

**2**

A proof of that involves the noting that there is no
definition. (Any comments?). Commentary: I was described to be
not understanding. To the rule owner: is the favourite always
one of the preferences, or approval sub-votes, selected on the
voter's paper?. If you had a mathematical formula, this entire
issue may have been transparent with the use of perhaps less
than 12 characters. There is no way that I could get a question
like this garbed into a symbolic form, so I do agree with Mike
that symbolism is of limited use. Still it meets with approval.

Perhaps the word favourite ("his favorite") is defined in a
dictionary and it is futile to answer. But then what of this
easily made illustrative objection:

  Would it be true or would it be false, to state that Mr
Ossipoff has precisely one favourite English dictionary?. I am
trying to derive an important finding about a method for
testing existing definitions (analogous to preferential voting
methods) and you house might have too many dictionaries. While
I can make an analogy with Mike's motive in creating a similar
problem inside FBC, I can't state why the defect in the test
exists.

  People can quickly see that FBC internally processes three
valued Booleans with the 3rd value occurring when the is not a
singe favourite. Worse, these 3rd Boolean states are operated
on by "if" or implications that are not defined to accept such
values. This is just a little detail suggesting Mr Ossipoff has
never tested his emended FBC method.


**3**

If papers replaced voters then the "favorite" could be taken to
be the candidate of the first preference. Mr David Catchpole
and I have had a discussion in the past and it was a trivial
matter for voter to be altered into paper. The size of the set
of clear problems with using the word "voter" inside FBC is
limited only imagination.


**4** 40,000!.k mayoral elections or else "voter" in FBC does
not mean 'voter':

Suppose FBC was using real voters. I.e. suppose the word voter
means voter. If these voters were electing a city mayor, then
isn't each shift of preferences able to correlate with entirely
new city election. If Mike's FBC is testing a 1 winner method
applying to 40,000 voters, then there would be about 40,000
factorial elections, times whatever constant. Possibly Mike
might say that the papers after the vote could be used. But
after the election the winner will be known. Just to make a
practical comment, some of the voters will want to alter their
favourite or votes. The fraction that decline to state who
their favourite could decrease ("this is protected by privacy
law"). If voters can't put the Mayor through 40,000!.k
elections, then isn't FBC is in danger of appearing to be in
need of being reworded as a purely mathematical idea. Maybe I
have overstepped what is right, by my presuming that Mike means
what he writes.

  People are individual to such an extent that it may be
impossible to come up with a mathematical proof that if FBC
passed voting method A, with say 100,000 Chinese, then it would
continue to pass the same method if another voter was added.
This holds even if the favourite defect is fixed. There may be
no problem if voter does not mean voter, and if voter means
voter, then FBC is more of a certifying that a criteria or
rule.


**5** FBC undecided over a method if any voter has 0 or more than
        1 favourite.

There is still that detail with FBC, which is that it takes
  only one voter without a favourite, to cause the FBC formula to
  return no finding on a method. I regard that as a defect.



At 04:47 16.03.00 +0000, MIKE OSSIPOFF wrote:
...
 >Sorry, but the meaning of "likes" isn't dependent on the
 >voting system being used or discussed. I made it clear several

**6** Creator of FBC seems to not understand what a "like" is

  Surely Mike, if a voters are to be a real humans, then could
be some correlations that a social statistician could turn up
on investigating. The statement ("Sorry...discussed") is wrong.

  If FBC were a living rule, the correlations between votes and
  asserted likes is able to be a matter for research, and that
  can't ruled out although there would be no issue if the
  definition of "likes better" was never tested because FBC was
  never used. A meaning people could have in disliking a method
  ("voting system") is that they seek to implement FBC as an
  agreeable collective delusion that can create an appearance
  to others that some particular method, say IRV, has been
  dealt with. It may never be a convincing process, given the
  flaws in STV, but sensitive statistical methods could detect
  traces of bias and indirect links with candidates' opinions
  on voting systems. The reality in dispute here is whether or
  not reality has any place in the theory of rules about voting
  theory.

**7** Rules are simple (a string of universal quantifiers, the
       exclusions followed by an implication sign that is made to
       imply what it is that is to be required). FBC proponents
       need to come to the party.

  I note that Mr Catchpole used the word voters and rapidly
  altered the meaning to be that of papers. He was trying to
  define an idea perfectly. Mr Catchpole, do have any ideas on
  why Mr Ossipoff won't fix the FBC definition. I am not happy
  with the direction that Mr Ossipoff is taking this discourse,
  it a little educative and I got the first glimpse at an
  argument that FBC can't be rendered into a form of symbolic
  logic --- not because of a problem with FBC, but (after I got
  the prolixity and what seemed to be traces of personal
  criticism separated from the ideas), because of a problem with
  mathematics.

**8** "likes best" won't always return Booleans, and the method
       is designed to fail to test when that happens

  The whole method falls fully with tri-valued Boolean operator
  errors, if a voter named spock-1 at startrek.com uses inference
  and logic to rank candidates (this is the ranking in thought,
  which need not be the ranking on paper, e.g. if Spock was on
  an undercover mission (check the definition, that is allowed)).

  The words "likes better" are able to return no Boolean value,
  aren't they?. This is the wording: "a voter should never gain
  an outcome that he likes better". It looks like that Mike has
  FBC negate the undefined value, and then logically conjuncts
  it with all others in the sequence. I was told that FBC is not
  undefined, but some of the parts that are defined process
  undefined values (wrongly), and components of the definition
  are wrong ("likes" would not be a word in a legal document,
  etc).

  This FBC started out being one of the properties satisfied by
  that method that is know to have that power spectrum defect
  that leads to unfairness, i.e. the Approval Voting method.

  So weak is FBC that it can't even test the Approval Voting
  method when that is not identical with FPTP. FBC could be
  fixed and that will be a matter I can keep up to date with.

  What is the origin of FBC?. Has this list got somehow a
  damaged copy of a definition that is elsewhere and that is
  not so flawed?. This version has been quoted 15 times. Can
  you investigate into that Mike?, and report findings?.

  Mr Ossipoff: where do we go to do new research into FBC?.

  I request that FBC be withdrawn. There is something incorrect
  with the aims of the defender(s) of FBC. Any one of Mr Saari's
  methods can be trashed with FBC, and so can STV and, sure, Mike,
  the Approval Voting method too. I have reducing interest in
  receving  any more wrong clarifications of the rule embedded
  in a context of disclosed opinions, making it tricky to decide
  if an emendation has in fact been made or not. Every checking
  of the definition shows absolutely no change for all of March
  2000.

In the final analysis, Mr Ossipoff's FBC is not a rule that
  tests methods. Mr Ossipoff will reply: will the idea of voters
  be altered.

**10**
Why is the word "likes" in FBC. Anybody can substitute other
  words, e.g. French language words that are similar but not
  quite the same?. That would create a new FBC. FBC is trapped
  in English speaking lands. What do you suggest be done, Mike,
  to get FBC out to the natives of Vanuatu?.