[EM] Collective ordering of SW standards

[Mike Ossipoff]

I emphatically urge that we just cover single-winner methods in this
FAQ & recommentation project, which is a SWC project.

PR is too different from  SW methods & issues. They can't be dealt
with in one project. PR is more straightforward & uncontroversial.
Its issues are about details. There are no really unacceptable PR
methods or systems. Single-winner methods is a topic that has never
been adequately discussed, and we all agree that there _are_ 
unacceptable single-winner methods, and that the one currently in
use is one.

***

Yes, I like the idea of voting among the SW standards, for the
purpose of creating a collective ordering of them, and then
using the standards in that order, to evaluate & compare methods
by each standard.

The alternative would be to just have a free-form discussion where
each person could name standards & tell why a method does or doesn't
meet that standard. But this would mean everyone talking on different
subjects, not dealing with other standards. The procedure of dealing
with 1 standard at a time puts each standard before everyone, gives
everyone an opportunity to deny that their method fails a particular
standard. It would always be possible to say "Then why didn't you
speak up when we were discussing that standard & it was said
that your method failed it?"

If we can conduct that vote on standards, and collect rankings
of the standards from the SWC members, that would give us the
best way to deal with standards: 1 at a time, with the whole
committee discussing methods in terms of each standard. That would
be discussion, as opposed to just a set of disjoint reports to send
to ER.

***

I prefer Condorcet's method for the voting method. A points-rating
system, like the Olympic system, even if everyone resisted the
temptation to use defensive or offensive strategy, would still
often give the impression of betrayal, especially in a small voting
group.

***

I realize that we don't all agree on the best voting system, but
fortunately Condorcet's method allows voting as many 1st choices as
one wishes to vote. So there's nothing about it that an Approval
advocate would object to. Voting A, B, & C in 1st place, and
not voting any other rank positions, would cast a vote for those
against all the others, just as in Approval. In general, an advocate
of a method with less freedom would have little or nothing to object
to when using a method with more freedom.

So I suggest a compromise for voting on the standards, to create
a collective ordering of them, to determine the order in which to
apply them to the methods:

Let's use Condorcet's method, with NOTB, and also with the disapproval
count proposed by Lucien. There's no reason why we couldn't include
both of those options with Condorcet's method.

This compromise should be acceptable to everyone, and it will
_greatly_ simplify things if we can just use that method, which
surely has a plurality in our group, rather than having to take
time by voting on a method, or by using the "Combined method",
in which we count the rankings by all the propoed methods, and then
take a 2nd balloting between the winners by those methods. That
need to take a 2nd balloting would add much extra voting when we're
making a collective ordering, so I hope we can agree on a compromise
voting system.

Of course if anyone insists on a vote, then we must have one, to 
choose the voting system. If that happens, and we do have to vote
on a voting systems, I claim that we should do that vote by the
Combined-Method, to avoid favoring any 1 method. As I said, we'd
count the rankings by every method, and then hold a 2nd balloting
between the winners by those methods.

But hopefully it won't be necessary to vote on a voting system,
if Condorcet-with-NOTB-&-disapproval is acceptable to Approval
advocates, & other non-ranked method advocates.

So does anyone object to voting on SW standards by that method,
as a compromise?

***

Of course, when making a collective ordering, from individual
rankings, a natural way would be to do a count by the method,
and, declare that count's winner to be the collective 1st choice.
Then, with that winner no longer considered a contestant, co a 2nd
count among the other alternaives, and the winner is the collective
2nd choice. & so on. I suggest not deleting previous winners from
the rankings on subsequent counts, but merely (of course) not
considering them to still be contestants. In this way, how
beaten an alternative is by a previous winner still counts in 
subsequent counts.

But this means that Condorcet's method doesn't even need more than
1 count: We'd get the same results if we just ordered the alternatives,
in the collective ordering, according to their Condorcet scores in
1 count: If there's an alternative that beats everything, it obviously
has the best Condorcet score, being unbeaten. The remaining alternaives
would be ordered below it according to how well they rate by Condorcet's
count rule.

***

As for how we'd use NOTB & disapproval, for a collective ordering,
1 possibility would be to use them at the beginning, to disqualify
alternatives, at the outset, from the whole collective ordering, and
then to do the single Condorcet count.

The other possibility, if that immediate disqualification from the
whole collective ordering isn't acceptable, would be to actually
do the whole series of Condorcet counts, applying NOTB & disapproval
in each count.

I emphasize that either way we only need 1 balloting. And that
even if we do the series of counts, to get the collective ordering,
that's still very little labor, since we have so few voters.

***

Anyway, I ask if anyone objects to that compromise voting
method: Condorcet with NOTB & disapproval.

***

It might seem improper to suggest that a compromise proposal
propsed by me should be the starting assumption for how we
should vote on standards. But Condorcet-with-NOTB-&-disapproval
is actually a combination of proposals from several people, and
is a genuine compromise. Besides, it probably also has a plurality
among the members of the SWC. That along qualifies it as the
starting assumption for how we should vote on standards.

If even 1 person objects to that voting method, then we have to
vote on a voting system by which to vote on standards. If it becomes
necessary to do that, the natural way would be by the Combined Method,
where we count the rankings by all of the methods proposed, and 
then hold a 2nd balloting between the winners by those methods. As
I've said, Approval would be good for that 2nd balloting, but any
method would be ok for the 2nd balloting.

If the different methods proposed used completely different kinds
of balloting--like point ratings as opposed to rankings, or Approval
ballots listing several alternatives unranked, then each ballot
sent in should have: a ranking; a set of point ratings; an Approval
set; etc. That's if we use the completely impartial Combined Method,
if it's necessary to vote on the voting method.

(The Combined method wouldn't be a good choice to vote among the
standards to get a collective ordering, since it would mean we'd
have to vote as many times as there are standards. Using the Combined
Method to vote on a voting system would just require one 2nd balloting.
And repeated counts, by a particular method rather than by the Combined
Method, wouldn't require any repetition of the balloting.)

***

Then, in any small election, there's the problem of frequent ties.
And the need for pre-arranged tie-breakers. I suggest that Plurality
is a good tie-breaker, and I suggest that it be the 1st tie-breaker
to be used with Condorcet's method. The use of Condorcet's method
as the primary method avoids Plurality's problems, and using it
as a tie-breaker brings the benefit of its favoriteness standard.
If Plurality returns a tie, then I suggest the 2nd tie-breaker be
a method that repeatedly eliminates the alternative with fewest
1st place votes, and transfers that alternative's 1st place votes
according to rankings' next choices, like MPV, except that the
process would stop as soon as an alternative acquired a plurality.
So it would be "Plurality-Elimination" as a next tie-breaker, in
case Plurlity returns a tie.

There are lots of possibilities for tie-breakers, but for simplicity
I suggest these 2.

I don't like the long ambiguous wordings for the elimination methods.
Let me state Plurality Elimination more briefly:

Repeately eliminate from the rankings the alternative which 
occupies highest position in fewest rankings. Continue this till
an alternative occupies highest position on more rankings than does
any other alternative.

***

Sorry this is so long. This isn't the 1st version. It was longer.
This is shortened version. I wanted to talk about the possibilities
for ordering the SW standards.

***

p.s. It has just occurred to me that, though NOTB & disapproval
don't give the same results, NOTB is a stronger rule, in the
sense that if being beaten by NOTB disqualifies an alternative,
and if ranking something below NOTB would be done by anyone
who'd give that alternative a disapproval vote, then it wouldn't
be necessary to use disapproval in an election that has NOTB. That's
because, under those conditions, any alternative disapproved of by
a majority would of course be beaten by NOTB.

So, if we agree that anyone disapproving of an alternative would
also rank NOTB over it, then there'd be no need to include the 
disapproval count in our vote on SW standards. If we were going
to include both options, the NOTB option has the disapproval option
covered.

Of course it's _because_ NOTB disqualifies things that disapproval
wouldn't disqualify (since being beaten by NOTB doesn't require
a _full majority_ of the voters ranking NOTB over the alternative,
whereas being disqualified by disapproval requires a full majority
to disapprove of it), that the 2 options, used separately, would
give different results. But using both would give the same results
as just using NOTB.

***

Mike

















--