[kwlug-disc] What is all this about systemd?

[Hubert Chathi]

On Thu, 20 Nov 2014 10:56:59 -0500, Ralph Janke <txwikinger at ubuntu.com> said:

> The problem is that the options while they may be interpreted as
> independent, are not really that. So what is not included in the vote
> is the strength of the conviction for or against an option.

That's true, and measuring strength of conviction is difficult to do,
especially in an objective manner.  The only voting method that I'm
aware of that attempts to do this (there may be others, but I don't know
of them), is one in which every voter is given a number of votes, and
can assign the votes to the options in whatever way they want.  So if
they feel very strongly about A over the other options, and their
preference for A overrides their preference for the order of the other
options, then they might assign all their votes to A, and none to the
others.  Or if they have a mild preference of A over B, and B over C,
they might assign most of their votes to A, some to B, and none to C.

Unfortunately, I don't know what the name of that method is.

> Ranking is not always adequate. While i.e. I might prefer Option 1,
> but would also accept Option 2, but don't want Option 3 or 4, I seem
> not be able to express that.

In Debian, you would rank Option 1 first, Option 2 second, Further
Discussion third, and Options 3 and 4 last.  Any option ranked beneath
"Further Discussion" is considered a vote for "I don't want that option
no matter what."  This takes effect through the fact that the winning
option must meet its majority requirement over the "Further Discussion"
option.  In a normal vote, this seems obvious, but it makes a difference
when an option requires a supermajority (for example, a 2/3 majority) to
win.  In that case, an option could be the most popular, but fail the
majority requirement, and still lose.

> That's why I tried to find a way to add the votes for Option 1 and 2
> in some way together. It was clear to me that simple adding would not
> work in this system, but my argument was more on the lines that the
> difference between Option 4 and Option 1 was very small, and the
> contextual closeness of Option 1 and Option 2 should be able to
> overcome that.

You could try to count the number of ballots where either option 1 or
option 2 was ranked over option 4 (the tally sheet is available[1], so
you can see what the actual votes were, and do the counting yourself),
but I would expect that option 4 would still come out ahead, assuming
that most people voted sanely.  This is actually a property of the
Schulze method that Debian uses, that having similar options on the
ballot should not make a difference to the actual outcome.  Of course,
mathematical properties do not always hold true when applied to human
behaviour...

[1] https://vote.debian.org/~secretary/gr_initcoupling/tally.txt

> In his voting system, this is not possible and in IMHO therefore, the
> question and options were not given in an unbiased way (not matter if
> it was intentional or unwittingly) and hence presents the limit of
> such votes.

> That was all I wanted to say, sorry if I was not clear enough, I was
> not criticizing the voting system itself, but more the usage thereof,
> and trying to indicate that democracy is mostly a very flawed process,
> even if it is deemed the best available one.

Yup, it's not absolutely perfect, but it's the way that Debian
Developers have chosen to make decisions, and for the most part[1] have
agree to abide by the results, and Debian in general considers it to be
a fair enough method.

[1] I say "for the most part", not because I suspect that anyone will be
working against the results, but because I don't want to make absolute
statements without evidence.