Re: General Resolution: Statement regarding Richard Stallman's readmission to the FSF board result

[Felix Lechner]

Hi,

On Mon, Apr 19, 2021 at 2:40 PM Pierre-Elliott Bécue  wrote:
>
> I don't understand how you semantically see 7 and 8 as comparable.

Aside from Bdale's reason for ranking unwanted options below FD—which
were motivated by the voting system—I do: GRs do not decide a matter
with prejudice, even though the weight to bring them again may be
substantial. Therefore, doing nothing is very similar to doing nothing
but talking more.

As we all heal from this divisive issue, I furthermore find it
meaningful that proponents of a shortened discussion, who were at
times accused of pushing the resolution, were actually aligned with
voters: By a narrow margin, people did not want to discuss the matter
at all.

Kind regards
Felix Lechner

Re: General Resolution: Statement regarding Richard Stallman's readmission to the FSF board result

[Pierre-Elliott Bécue]

Le lundi 19 avril 2021 à 12:46:38-0700, Russ Allbery a écrit :
> "Barak A. Pearlmutter"  writes:
> 
> > Sam, you make an excellent point about gaps between options, and that
> > a ranking does not show the strength of preferences. Like, I might
> > prefer ALPHA >>> BETA > GAMMA while you prefer ALPHA > BETA >>> GAMMA.
> > So if it's down to ALPHA vs BETA, my vote should shift things  more
> > than yours, while if it's down to BETA vs GAMMA, your vote should
> > shift things more than mine. And if we do sorta-maybe try to encode
> > this with where FD is in the ranking, it does not actually have this
> > effect.
> 
> > If we wanted to encode this information more fully, we would have to
> > go with some system where people give numeric strengths to each gap in
> > their preferences. And to avoid people just pegging them all to
> > maximum strength, we'd have to put a limit on the total strength in a
> > single ballot.
> 
> I think it's worth observing that this discussion started with "our voting
> system is too complicated and I think some people are making nonsense
> votes because of it" and has now arrived at "we should make our voting
> system considerably more complicated to improve its expressive power."
> 
> This all seems extremely speculative.  Is there some GR whose result you
> think did not accurately represent the correct outcome given the
> preferences of the people who voted?  Precisely what problem are you
> trying to solve here?

I think we are good ad nitpicking and this is some of it. :p

(more seriously, I think our system does quite correclty what it is
designed to do)

-- 
Pierre-Elliott Bécue
GPG: 9AE0 4D98 6400 E3B6 7528  F493 0D44 2664 1949 74E2
It's far easier to fight for principles than to live up to them.


signature.asc
Description: PGP signature

Re: General Resolution: Statement regarding Richard Stallman's readmission to the FSF board result

[Pierre-Elliott Bécue]

Le dimanche 18 avril 2021 à 22:18:22+0100, Barak A. Pearlmutter a écrit :
> The Schwartz set resolution algorithm is absolutely best of breed. But
> there's an old saying in computer science: garbage in, garbage out.
> 
> If we look at the actual ballots, it's really interesting. Options 7
> and 8 were semantically pretty much equivalent. It's hard to see any
> reason for someone to rank them very differently. So if the voters are
> rational, we'd think that nearly all ballots would have options 7 and
> 8 ranking either the same or adjacent. And that if one is ranked the
> same as other options, then they should both be. Yet many of the
> ballots rank one but not the other, or rank them very differently.
> Some voters ranked either option 7 or 8 as "1" and allowed everything
> else to default. It's very difficult to imagine someone who actually
> preferred option 7 being equally satisfied with any of options 1-6 and
> 8.
>
> We tend to assume that the DD electorate is highly sophisticated and
> rational and understand how to correctly express their preferences,
> and how ranking works. But a quick perusal of the actual ballots has
> disabused me of that notion.
> 
> The usual reaction to this sort of thing is to alter the voter
> instructions. But people have intuitions for how voting works, and
> blurbs might not be very effective at changing their behaviour.

Hi,

I don't understand how you semantically see 7 and 8 as comparable.

Regards,

-- 
Pierre-Elliott Bécue
GPG: 9AE0 4D98 6400 E3B6 7528  F493 0D44 2664 1949 74E2
It's far easier to fight for principles than to live up to them.


signature.asc
Description: PGP signature

Re: General Resolution: Statement regarding Richard Stallman's readmission to the FSF board result

[Russ Allbery]

"Barak A. Pearlmutter"  writes:

> Sam, you make an excellent point about gaps between options, and that
> a ranking does not show the strength of preferences. Like, I might
> prefer ALPHA >>> BETA > GAMMA while you prefer ALPHA > BETA >>> GAMMA.
> So if it's down to ALPHA vs BETA, my vote should shift things  more
> than yours, while if it's down to BETA vs GAMMA, your vote should
> shift things more than mine. And if we do sorta-maybe try to encode
> this with where FD is in the ranking, it does not actually have this
> effect.

> If we wanted to encode this information more fully, we would have to
> go with some system where people give numeric strengths to each gap in
> their preferences. And to avoid people just pegging them all to
> maximum strength, we'd have to put a limit on the total strength in a
> single ballot.

I think it's worth observing that this discussion started with "our voting
system is too complicated and I think some people are making nonsense
votes because of it" and has now arrived at "we should make our voting
system considerably more complicated to improve its expressive power."

This all seems extremely speculative.  Is there some GR whose result you
think did not accurately represent the correct outcome given the
preferences of the people who voted?  Precisely what problem are you
trying to solve here?

> To go back to your restaurant situation, imagine there is one person
> who's deathly allergic to seafood, so really doesn't want to go to the
> dim sum place. Many others do like dim sum (perhaps even a majority),
> but it's just a mild preference, they be happy with many of the
> restaurant options and okay with all of them.

Thankfully, our voting system is not an ideal mathematical model in which
communication is limited to only the votes that one casts.  Someone can
stand up and say "hey, I'm deathly allergic to seafood," and the rest of
us can take that as input into what decisions we want the project to take.

-- 
Russ Allbery (r...@debian.org)  

Re: Re: General Resolution: Statement regarding Richard Stallman's readmission to the FSF board result

[Kurt Roeckx]

On Mon, Apr 19, 2021 at 05:32:40PM +0100, Barak A. Pearlmutter wrote:
> Sam Hartman writes:
> > For me though, even there, notice that we'd be choosing between options
> > that the voters considered acceptable.
> > Because of that, I am not bothered by the cycle.
> 
> If the decision doesn't really matter but a non-FD option must be
> chosen (like a hungry group picking a restaurant) then sure, whatever.
> 
> But for something important, I think we *should* be bothered by a cycle.
> 
> Let me give an example. Let's say we end up with ALPHA, BETA, and
> GAMMA in a cycle: ALPHA>BETA, BETA>GAMMA, GAMMA>ALPHA. So we run our
> resolution algorithm, and it picks ALPHA. Well that's nice. But note
> that if we had not had BETA as an option on the ballot, then GAMMA
> would have been the winner, without anyone changing their votes.

You might want to read:
https://en.wikipedia.org/wiki/Independence_of_irrelevant_alternatives#Criticism_of_IIA

> Not
> only is that odd, but it means that it would be reasonable for a
> die-hard ALPHA supporter, seeing that GAMMA is going to beat ALPHA,
> would propose adding BETA to the ballot. This means that we are
> vulnerable to strategic (rather than honest) behaviour throughout the
> process, including in proposing ballot options.

So that would be:
https://en.wikipedia.org/wiki/Strategic_nomination

If the option is similar to an existing option, it should not
have an effect for the Schulze method we use:
https://en.wikipedia.org/wiki/Independence_of_clones_criterion

No voting system is perfect.


Kurt

Re: General Resolution: Statement regarding Richard Stallman's readmission to the FSF board result

[Barak A. Pearlmutter]

On Mon, 19 Apr 2021 at 16:35, Sam Hartman  wrote:
> I think we need voting reform around how the amendment process works and
> managing discussion time ...
> ...
> Preferences can be of different strengths.
> 
> Which is to say that the gaps between preferences might be relatively
> weak.

Sam, you make an excellent point about gaps between options, and that
a ranking does not show the strength of preferences. Like, I might
prefer ALPHA >>> BETA > GAMMA while you prefer ALPHA > BETA >>> GAMMA.
So if it's down to ALPHA vs BETA, my vote should shift things  more
than yours, while if it's down to BETA vs GAMMA, your vote should
shift things more than mine. And if we do sorta-maybe try to encode
this with where FD is in the ranking, it does not actually have this
effect.

If we wanted to encode this information more fully, we would have to
go with some system where people give numeric strengths to each gap in
their preferences. And to avoid people just pegging them all to
maximum strength, we'd have to put a limit on the total strength in a
single ballot.

That's a very interesting idea. I wonder if we could elaborate upon it
to build a more expressive, and more robust, voting system.

To go back to your restaurant situation, imagine there is one person
who's deathly allergic to seafood, so really doesn't want to go to the
dim sum place. Many others do like dim sum (perhaps even a majority),
but it's just a mild preference, they be happy with many of the
restaurant options and okay with all of them. It would be nice if the
allergic person were able to express that in a ballot. Right now,
they'd put everything-else>FD>DIMSUM, but that doesn't really have the
expressive power we'd like, which is that this one voter could put
*all* their expressive power against DIMSUM instead of being forced to
distribute it between all their preferences even though their
preferences between the other restaurants are, by comparison, very
small---and not doing so just wastes the power. What we need is for
people to be able to express mild preferences
SUSHI>DIMSUM>ITALIAN>TAI>..., but the one person who really cares to
be able to go {SUSHI,ITALIAN,TAI}>>>DIMSUM so they can really move the
meter on DIMSUM, at the expense of their ability to express other
preferences.

In an informal group setting this happens naturally. That's why we
discuss which restaurant to go to, rather than voting. We want to
gauge the strength of people's preferences and take that into account.

--Barak.

Re: Re: General Resolution: Statement regarding Richard Stallman's readmission to the FSF board result

[Bdale Garbee]

FWIW, I didn't consider 7 and 8 at all similar.  

After watching the strain the pre-vote discussion introduced, I decided making 
no statement as a project was the best outcome.  But if the project were to 
make a statement, I wanted to express preference between the acceptable to me 
statements, then put the unacceptable to me options below FD.  

Bdale

On April 18, 2021 3:18:22 PM MDT, "Barak A. Pearlmutter" 
 wrote:
>The Schwartz set resolution algorithm is absolutely best of breed. But
>there's an old saying in computer science: garbage in, garbage out.
>
>If we look at the actual ballots, it's really interesting. Options 7
>and 8 were semantically pretty much equivalent. It's hard to see any
>reason for someone to rank them very differently. So if the voters are
>rational, we'd think that nearly all ballots would have options 7 and
>8 ranking either the same or adjacent. And that if one is ranked the
>same as other options, then they should both be. Yet many of the
>ballots rank one but not the other, or rank them very differently.
>Some voters ranked either option 7 or 8 as "1" and allowed everything
>else to default. It's very difficult to imagine someone who actually
>preferred option 7 being equally satisfied with any of options 1-6 and
>8.
>
>We tend to assume that the DD electorate is highly sophisticated and
>rational and understand how to correctly express their preferences,
>and how ranking works. But a quick perusal of the actual ballots has
>disabused me of that notion.
>
>The usual reaction to this sort of thing is to alter the voter
>instructions. But people have intuitions for how voting works, and
>blurbs might not be very effective at changing their behaviour.
>
>--Barak.

-- 
Sent from my Android device with K-9 Mail. Please excuse my brevity.

Re: Re: General Resolution: Statement regarding Richard Stallman's readmission to the FSF board result

[Barak A. Pearlmutter]

Sam Hartman writes:
> For me though, even there, notice that we'd be choosing between options
> that the voters considered acceptable.
> Because of that, I am not bothered by the cycle.

If the decision doesn't really matter but a non-FD option must be
chosen (like a hungry group picking a restaurant) then sure, whatever.

But for something important, I think we *should* be bothered by a cycle.

Let me give an example. Let's say we end up with ALPHA, BETA, and
GAMMA in a cycle: ALPHA>BETA, BETA>GAMMA, GAMMA>ALPHA. So we run our
resolution algorithm, and it picks ALPHA. Well that's nice. But note
that if we had not had BETA as an option on the ballot, then GAMMA
would have been the winner, without anyone changing their votes. Not
only is that odd, but it means that it would be reasonable for a
die-hard ALPHA supporter, seeing that GAMMA is going to beat ALPHA,
would propose adding BETA to the ballot. This means that we are
vulnerable to strategic (rather than honest) behaviour throughout the
process, including in proposing ballot options.

If ALPHA, BETA, and GAMMA are restaurants, then it doesn't matter. But
if they're different directions for the future of the project, or
otherwise of great importance to people, then it becomes a big deal.
And if it calls the legitimacy of the voting process into question,
then it becomes an even bigger deal. Like "hey, who proposed BETA, and
were they actually an ALPHA supporter? I call shenanigans!"

In the RMS GR, people proposed or seconded ballot options that they
themselves did not support, as a way of ensuring fairness and coverage
of opinions. I think that was very healthy and considerate, and I'd
like to thank the people involved for trying to make the process
inclusive and be sure everyone had an option that they could stand
behind. But if we had a cycle, it could be misconstrued, and I think
that would be a shame.

--Barak.

Re: Re: General Resolution: Statement regarding Richard Stallman's readmission to the FSF board result

[Sam Hartman]

> "Barak" == Barak A Pearlmutter  writes:

Barak> The Schwartz set resolution algorithm is absolutely best of
Barak> breed. But there's an old saying in computer science: garbage
Barak> in, garbage out.

Barak> If we look at the actual ballots, it's really
Barak> interesting. Options 7 and 8 were semantically pretty much
Barak> equivalent. It's hard to see any reason for someone to rank
Barak> them very differently. So if the voters are rational, we'd
Barak> think that nearly all ballots would have options 7 and 8
Barak> ranking either the same or adjacent. And that if one is
Barak> ranked the same as other options, then they should both
Barak> be. Yet many of the ballots rank one but not the other, or
Barak> rank them very differently.  Some voters ranked either option
Barak> 7 or 8 as "1" and allowed everything else to default. It's
Barak> very difficult to imagine someone who actually preferred
Barak> option 7 being equally satisfied with any of options 1-6 and
Barak> 8.


In my mind the ballot options are not similar.  First, things above FD
are things I don't mind being in a cycle.  If it's ranked above FD, I'd
rather be done with a decisdiscussion and have that option win even if
it is not my preferred option.  Options below FD are options I'd prefer
not make their way into a cycle.

Second, FD implies that the question is still open.  I might be able to
convince people to choose something more aligned with my option in the
following discussion.
In contrast, option 7 is final; we've made a decision.

So, in filling out my ballot I rank:

1) Options that I like--where I'd be okay with any of those options
getting chosen.

2) fd

3) Options that are in the general direction I like, but are weak enough
that I'd rather have an opportunity to ask people to do something
stronger than choose those options.

4) no statement

5) options that are in a direction I disagree with.

Re: General Resolution: Statement regarding Richard Stallman's readmission to the FSF board result

[Sam Hartman]

I'm writing to present an alternate interpretation--the one under which
I think our voting system is doing a good job of choosing among complex
ballots in the last couple elections.
I think we need voting reform around how the amendment process works and
managing discussion time, but I am very happy with how the actual voting
mechanism  has worked.
That's true even though my preferred option didn't win in either the rms
election or the systemd election.

> "Barak" == Barak A Pearlmutter  writes:

Barak> If the winning option in an election is part of a preference
Barak> cycle, then it (by definition) has the property that there
Barak> exists some other option that a majority of the voters
Barak> preferred. In some elections that is unavoidable: we need to
Barak> pick one DPL, and if they're in a cycle so be it; if there's
Barak> a tie we can just toss a coin. But in others, like the RMS
Barak> GR, it seems like it would be a rather bad property and we'd
Barak> be better off treating it as FD and trying again later.

Barak

Preferences can be of different strengths.
Imagine we were using the Debian voting system to decide where to go to
dinner before a conference in six weeks.
It might well be that we had five or six options that were generally
acceptable to most people (and perhaps a couple options that were
unacceptable that got dropped).
We don't have to go to dinner, and we don't even have to use the voting
system to make our decision.
So, unlike the DPL election, a decision is not necessary.
And yet, I suspect many people might well prefer to be done with things
and to have a decision even if there is an option that a majority of
voters prefer  to the selected option.
Which is to say that the gaps between preferences might be relatively
weak.

I think we've tried to encode that in our voting system with the
majority requirement.
We never select options that the voters consider unacceptable.
And among the options that the voters do consider acceptable (if any),
we'd prefer to make a decision than not to make a decision.

Consider for example if we had a cycle between options 2, 3 and 4.
That would be a clear desire to make some sort of statement, and the
debate would be over how strong of a statement to make.
I don't think we would be well served in such a situation to make no
statement at all.

It gets more complex when you add option 7 (the no statement option)
into the cycle.

For me though, even there, notice that we'd be choosing between options
that the voters considered acceptable.
Because of that, I am not bothered by the cycle.



signature.asc
Description: PGP signature

Re: General Resolution: Statement regarding Richard Stallman's readmission to the FSF board result

[Jonas Smedegaard]

Quoting Don Armstrong (2021-04-19 00:39:12)
> On Sun, 18 Apr 2021, Barak A. Pearlmutter wrote:
> > If we look at the actual ballots, it's really interesting. Options 7 
> > and 8 were semantically pretty much equivalent. It's hard to see any 
> > reason for someone to rank them very differently.
> 
> 7 was a decision to not issue a statement ["There's no statement on 
> this issue that I want Debian to issue"]. 8 was a decision for further 
> discussion ["There may be statement on this issue that I'd want Debian 
> to issue, but it's not here."]
> 
> When there isn't an explicit "no decision" option on the ballot, 
> further discussion encompass both, but that is not the case here.
> 
> > It's very difficult to imagine someone who actually preferred option 
> > 7 being equally satisfied with any of options 1-6 and 8.
> 
> Here's an example thought process that works: "I want Debian to stop 
> discussing this issue and anything more that Debian does on this issue 
> is equally bad."
> 
> Or another one: "I know that I prefer this option, but I'm not 
> comfortable with the rest of the options to decide what the project 
> should do, so I'll defer to the project's judgement."
> 
> Not to say that there aren't voters who are confused, but you should 
> contact them to figure out why they voted the way they did before 
> assuming that they didn't know what they were doing.

In case anyone wants to pick a live brain on this, I volunteer:

At first I voted 7-8: I explicitly preferred Debian to *not* issue a 
statement.

Then, after reading the discussion on this list about the concern over 
people leaving options below FD "blank", I changed my vote to wade 
through all those options I did *not* want Debian to make and try rank 
the severity of their badness - while being worried that my vote is 
public so I expose my priority of evil thoughts to the World.


 - Jonas


P.S.

This is *not* an invitation to rehash a debate over which ballot options 
are or are not sane. My offer is that if you have trouble understanding 
why someone deliberately choose to vote by the two _patterns_ described 
above then I am willing to reflect on that.  Perhaps you even manage to 
point out to me that I am a fool and did what I did for the wrong 
reasons.

-- 
 * Jonas Smedegaard - idealist & Internet-arkitekt
 * Tlf.: +45 40843136  Website: http://dr.jones.dk/

 [x] quote me freely  [ ] ask before reusing  [ ] keep private

signature.asc
Description: signature

Re: General Resolution: Statement regarding Richard Stallman's readmission to the FSF board result

[Jonathan Carter]

On 2021/04/18 23:36, Bernd Zeimetz wrote:
> Complaining about the
> voting system because you don't like the outcome or because you could
> announce the outcome in an awkward way is not helpful.

Who complained about the voting system because they didn't like the
outcome of this particular vote? I've literally not seen one instance of
that.

-Jonathan