[EM] Comments on Nanson &...) Mike's old rules not tested

Craig Carey research at ijs.co.nz
Sun Sep 10 17:06:26 PDT 2000


I shall leave this up to my colleague, Mr David "where's my pen" Catchpole.
I just rejoined this mailing list and it seemed OK for my first message.
But an unreasonable criticism of the Plurality method occurred. We can't
lose that. I have not studied that method yet. Methods are not bad if they
differ from Condorcet.

I comment on some errors at this page:
http://www.electionmethods.org/criteria.html

Please try to overlook some of the errors in this. There could be many.



At 18:39 00.09.10 +0000 Sunday, MIKE OSSIPOFF wrote:

>Someone wrote:
>
>The election method which is an elimination method based on the Borda count
>   is usually known as Nanson's method. It has the very nice property that if
>    there is a condorcet winner that the method chooses that candidate.
>
>I reply:
>
>If we use "Condorcet winner" to mean a candidate who, when
>compared separately to each of the others, is voted over him
>by more voters than vice-versa, then Plurality meets that criterion.
>
>The Condorcet Criterion can be worded in terms of a sincere
>Condorcet winner (a candidate who, when compared separately to
>each one of the other candidates, is preferred to him by more
>voters than vice-versa):
>
>If there's a sincere Condorcet winner, and everyone votes sincerely,
>then the sincere Condorcet winner must win.

The candidates that are the winners of an election using a formula
applying to votes, are not only the winners that must win, but they are
also the winners that do win.

I will rewrite the statement that M.O. wrote.

"If there's exists X, and Y, then Z.

X = the sincere Condorcet winner, i.e. the winner, or undefined (perhaps)
    if there is no winner.

Y = "everyone votes sincerely". We could replace that with "not Y" and
    not affect the argument.

Z = "the sincere Condorcet winner must win".

My point is that the author seemed to be aiming to emphasize something
that was quite different from truth. A person that wanted to make a
true statement would remove the Y term.



>That wording has the advantage that Plurality doesn't pass, but
>some methods do pass.

(A definition of Plurality Voting:
http://www.ccrc.wustl.edu/~lorracks/dsv/diss/node1.html )

Your test is almost a test 'that a method be Condorcet'. That is not so
bright since whenever Condorcet can pick a winner, it can get the wrong
winner. That is a problem with Condorcet: it will pick the wrong winner.



>Obviously a Condorcet Criterion wording that doesn't have those
>properties isn't very useful.

It isn't an actually useful definition, although it could become
a bit better, as for large enough problems, over 99.9999% of the
space in the simplex of possible elections is full of a Condorcet
0-winners paradox region.

Some of you rules lose power on large problems. Good rules do not
do that. All a politician need to is check the rule and see if its
power tends too fast towards 0 power, as the number of candidates
increase without limit. That takes down a lot of rules, and just
the bad ones. [I need to watch out for my P2]. Why create rules
that are weak?. Mr Schulze calls for them to be discarded, but
various of the rules discard themselves. Mr Ossipoff designed them
to do that (to pass the Approval Vote or something).


>So the Condorcet Criterion stipulates sincere voting by everyone.
>Good luck :-)  That's why the Condorcet Criterion doesn't seem
>to be very useful, even when worded in a usable form. FBC,


I don't believe that statement about FBC is true. FBC never got so
close to being used that your were able to write down its definition
in a way that did not fail, when voters became confused when talking
to your psychologist investigating spirits. Are there such things?.
There has to be some mechanism. If I am wrong do tell me how the
rules get their data?. Maybe the lord of the FBC rule would pay
Demorep to falsify up info for the rule. I could do that. Let's
get the Alternative Vote tested, Mike. I'll oversee the testing and
interpret your rules, and you can check the method that is used in
a real election and together we can find out if the "insincerity"
of the British voters is too much to allow the Approval Vote to
be passed. Being a certifying the data would appear after the
election. Electoral reformers can just pull the method out. If
that is the best that can be done when Mikes rules are used
(unlike Blake Cretney's which can be used before an election), then
that can be done. Mike were you rules ever useful in allowing you
to get the Approval Vote into local government, given that they
can't test the sincerity of a public under an Approval Vote
election until one of those elections is actually held.

PS. Mike doesn't read what I write if it is over 2KB. That is not
permitted by the rules of Rob Lanphier. I don't mind. I may have
to keep repeat a little. This is after all, a pro-Condorcet list,
and the methods are 1 winner methods.

Russ Paielli's page on rules has really no info on how many winners
a method has. The public are expected to know that.

  http://www.electionmethods.org/criteria.html

I shall write about FBC here. I uses the word "favorite". I ask
the reader, if you are voting: how does FBC find out what your
"favorite" was?.

So I now postulate the existence of Mike Ossipoff spirits, because
some people won't talk. That step was logical and easy. I note to
Russ that I would find the page easier to follow if it held this
information: the investigators. In USA they could be men in white
shirts. Or mediums with remote sensing powers that lay on beds.
The origin of the data just needs to be accurate, which in some
countries could be a violation of privacy law.

Suppose the middle management team was about to be sacked and they
voted on each other after a merger of 2 companies. FBC needs to
know which favorite, voters like more. Suppose one company's managers
got bribed and liked themselves but voted for others, just to get
payouts. They like themselves (the candidates) but the money more.
That is another method in trouble. Why design a method to fall apart
until quick consideration.  It might as well be fixed.

WDSC has a similar defect in this part, I guess:
"If a majority of all the voters prefer A to B, then they should
have a way of voting that will ensure that B cannot win"

Suppose A and B are groups and all the A candidates (=voters) vote
for the B candidates but like themselves?.

What is the name of the creator of these rules?. They don't seem
to have been thought out at all. If Mr Ossipoff can't defend the]
bugs in his thinking, I hope Russ Paielli can. Russ this may not
be so quick, because Mr Ossipoff does not realize and fix errors,

I don't want to complain now, since I wasn't a subscriber yesterday.

---

Does the Condorcet "Criterion" really stipulate (= "specify as a
condition", "qualify",) something of voters?. It would reject
mismatching voters?. It would have to access to some information
about voters?, prior to the act and effect of the stipulating.

At worst, just as I am about to put a voting card into the slot,
a spirits would materialise to stipulate that I had not voted
falsely.

So I guess "stipulate" means data is collected.
Suppose a vote starts at 10 in the morning on a Saturday: when does
the accessing of the data on the voters' sincerity begin?. Mike!:
you must have organised a call out: do they hang around for minutes
while trying to find out where the booths are?.

Suppose a noisy crowd of anarchistic supremecists vote in a clump:
how exactly does a Ossipoff spirit investigator resolve the fact
their voting for their own leaders could be insincere if the leaders
are losers, idiots and predators?. Suppose they are cultists and
love their leaders but so stupid they voted for the wrong
candidates?. It is not just Mike's FBC. GSFC says the "majority
should be able to sincerely rank all the candidates, and B still
shouldn't win,". The wording itself presumes they know their own
opinions which may be false. It presumes no vote fixing which may
be false. The very wording rank embeds an undefined voting method
into the rule. Can I put in FPTP and have it ignore all preferences
save the first?. How do you create your rules when you make mistakes
like hiding undefined election methods in them?. Was FSBC ever
tested. Did you ever write it into a symbolic logic form?.

Is it OK if I have a special class of rules that pass or fail methods
like STV and have thta decision affected by the opinions of the
British Labour Party (or whatever data I manage to reverse probe out
of Mike's truth and "sincerity" data collecting emmisaries?. What
of the Plurality vote: can that method unaided issue forth such a
tyrannical protest at the injustice written for it [it had to get
the right number of winners while Condorcet didn't] that the SARC
"then the fact that they showed up" spirits [ibid R.P's page]
themselves never make it out to the voting rooms to find out
enough to know whether the vote was later fixed and less arrived
that actually did?. This is a violation of individuals' right to
privacy. Suppose it was an election in the United Nations and a
representative was using a proxy. I suspect the SARC definition
fails in that circumstance. I presume you have not checked. Can
you get that done?. Here is the problem I suspect:

   "(which we'll call V2) if there's some configuration of the
   other people's votes for which that voter likes the result of V1
   better than the result of V2,"

That seems to be all about the internal opinion of the proxy and
not the nation being represented. Nations can be reluctant to use
the word "likes". Has Mike Ossipoff got synonyms for the word
likes?. I asked once before and referred me to a dictionary. In
the last day I downloaded a remarkable researcher's dictionary
named WordNet, from
http://www.cogsci.princeton.edu/~wn/obtain/
But it would help me understand FBC more accurately if you told
me the range of sysnonyms that the spirits would interpret
coinciding with the likes of a nation. I don't want to leap
beyond what readers would think reasonable, as Markus Schulze
did, and give you advice on the rejection of the rules (and the
trailing vanloads of psycholgists that appear with notebooks at
elections [or handhelds, if the 'preferential voting method'
certification agency has the funding for that)



>That wording has the advantage that Plurality doesn't pass, but
>some methods do pass.
>
>Obviously a Condorcet Criterion wording that doesn't have those
>properties isn't very useful.
>
>So the Condorcet Criterion stipulates sincere voting by everyone.
>Good luck :-)  That's why the Condorcet Criterion doesn't seem
>to be very useful, even when worded in a usable form. FBC,
>SARC, WDSC & SDSC offer absolute guarantees to the voter, without
>any stipulations. SFC & GSFC only stipulate that no false preferences

This is a second use of the word "stipulated". He means it:
The method "stipulates", or requires. It gets data from voters
through hyperspace like in Babylon 5, and if the method doesn't
like it, the method can't come to a conclusion. It is sounds like
such an untrustable procedure that I am sure I speak for members
when I say that no-one would want Mike to be handling all this
data from the shadows of hyperspace.

These a times of new theories. The dawn of a new age, when it is
unsafe to leave too much power to those that have not proven that
they can both write a rule and test it at least 1-2 times.

But Mike can fix that by precisely telling the list which of his
rules (listed on Russ's page(s)), pass, or fail, the Nanson
method.

I request that under the rules of the mailing list and I ask that
subscribers keep me and Schulze informed on the true facts of the
holdup in the testing of the Nanson method. The rules are
certifying so they need to be redone at each election, and if
that is slow, Mike should have an explanation (otherwise they
may be impractical). Or else a think-tank operation could be
started. Unless of course, all subscribers but one regard the
Russ rules as largely/entirely of no merit. However they could
be repaired. I have had a very bad experience trying to repair
Mike Ossipoff's FBC. SARC is a missing idea. SARC says:

"If a group of voters share the same preferences,...".

That could be a multiwinner election. A logic program would
never get to receive the SARC rule for testing. It is not defined
when there is more than one winner. There can be >1 winners in SARC.

------

What about a well defined method of Demorep that might not get
exercised by anybody except him, in an election that took a fraction
of a second, 2,000 years in the future, in the troll's hall checking
during a project trying to find big errors in human democracy.

Why such a long wait Mike?. It is customary to make rules usable.
A rule is not usable if there are 200,000 voters and their
sincerity needs to be sampled. How do you estimate the undetectable
biases.

Why even find out if voters voted sincerely?.

Suppose I voted for candidate 6 in an election under a new method,
devised by the best theorist in the list.  Mike read the 6 as a 9
and thought I sincerely wanted candidate 5 and was just trying to
sabotage the dual test. The shadows told him that (Russ can you
update the page). Are readers (Lanphier, Schulze, DemoRep, BC,
Mr I, and the many remaining), satisfied that Mike has not told
about procedures to keep him unable to exclude the testing of
methods. If you want to test governmental methods Mike, you need
to be able to check and double check. Instead of a 6 being
sincerely replaced with a 9, suppose the Prime Minister made
a lot of good speeches. Would you make speeches yourself Mike?,
or do the rules just become unusable?. I suggest that
Rob Lanophier oversee the investigation into the procedures that
would make it safe to translate Mr Ossipoff's thoughts into the
world where mathematical formulae can be harmed (starting with
the Plurality vote). How long would it take Rob to report back
to me via the list?. This strikes me as being a better idea than
Mr Schulze's "ditch some of the ideas" suggestion.

I recall that attempting to adjust and repair Mikes errors in
definitions led to a comment that I would need to give my version
a new name. Why not just rename all of the methods at Russ's site,
by attaching the words "version 0.001". It would be best if Mike
fixed them himself. But unfortunately this message exceeds Mike
Ossipoff's 2KB limit on what he can permit himself to read.


-------------------------

GFSC is an error, I'd say:

"GSFC extends SFC to cases where there's no Condorcet winner. "

So whatever is wrong with GFSC might also be a problem in SFC (check
that later).

GFSC starts with:  "If there's no falsification of preferences".

That is ambiguously worded.
Suppose voters voted quite sincerely, but ballot boxes were
replaced?. That is a falsification of votes. The words "falsification
of preferences" has, or almost has, two different meanings. One can
name the replacing of papers, and another, the voting in a way
that corresponds to lying. SFC and GFSC focus upon the act of voters
lying. Voting booths have toilets and information from those rooms
essential in the opinions of other theorists, to judge the fate
of STV. Why design your methods so that lying, e.g. in reply to
questioning indicating whether responses to questions about
sincerity, are accurate and the shape and moments of the statistical
distribution. I think SFC and GFSC might as well be rejected.
Otherwise maintaining a mind full of lies as a defence to the Mike
Ossipoff spirits who later misuse the information for something
other than GFSC, is going to be a surreal experience. I do suggest
that GFSC and SFC are hoaxes. However it might be just

-----

SFC says this:

"SFC: Strategy-Free Criterion

     If no one falsifies a preference, and if a majority rank the
     Universal Winner over candidate B, then candidate B shouldn't win."


Suppose the falsification occurs with a probability of 10**(-12).
That is reasonable. I guess SFC makes no use of probability. Since
some US citizens will "falsify", SFC could not ever test the
Approval Vote for an election of a US president.

FBC is similarly worded so that a single US voter out of millions, can
crash and fail the test of a voting method, and make it be rejected.
Mr Ossipoff seems have it worded to do that.

"FBC: Favorite Betrayal Criterion

      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."

-------

What does this mean in the GSFC "Generalized Strategy-Free Criterion" ?:

     "The members of that majority should be able to sincerely rank
     all the candidates, and B still shouldn't win, if the above
     conditions are met."

Suppose they were retarded infants and although they were subscribed
to the election methods list, they just couldn't clearly get the
meaning of the word "rank" right?. GSFC needs to be preceded with
explanations about why ranking has to be sincere. The Australians
are as insincere as their politicians (who always like to vote) when
have the party list rankings in their STV elections. SFC says:
"that majority should be able to sincerely rank all the candidates".

Why require that over 50% rank all the candidates when perhaps 80%
want to just use the parties ordering?. It is a bad definition.


>are voted, which is a less demanding stipulation than sincere
>voting, which, with a rank method, means sincerely ranking all of
>the candidates, and forbids truncation.
>
>By the way, I'd like to add that the Smith Criterion,
>Condorcet Loser, Majority Loser, & Mutual Majority, all
>need to be fixed in the way that Condorcet's Criterion needs to

... fixed or lost.

>be fixed, in order to be meetable, but not met by Plurality.

Fixed? How would that be done?. If you make a rule that requires a
method to be the same as Condorcet then you have a rule that makes
a method bad.


>Mike Ossipoff

This is the theory: all the badness in a method that cannot be
expunged, has to collect itself into the region when Condorcet
can't find a winner. We know for a fact that types of
insincerity can not be removed from methods. I presume including
types Mr Ossipoff says he does not like.

Everyone understand: bad properties that are impossible to remove
are required to be present peculiarly in the region where Condorcet
can't find a winner. Why is it desirable to not spread out the
problems while at the same time reducing its intensity?.

Mike's belief the the Condorcet winner is good, is undesirable and
unproven, and in somewhere in the future, the word false could be
used to describe it. Of course I am arguing against Tideman and
Schulze or SSD as well. They haven't been getting promoted well
here, but it may not matter.

I hope the website of Russ Pialli is improved. The defitinions are
bad. I can't call them false because they are not well defined. That
might not be fixable (e.g. if they lose power is the number of candidates
rises).

I leave this to others.




Craig Carey




More information about the Election-Methods mailing list