Re: GNOME Foundation Elections - Official list of candidates

[danilo gnome org (Danilo Šegan)]

Hi David,

I promised (myself) not to engage in this discussion anymore, but I
have to point out a flaw in this example for everybody's
consideration.

I acknowledge that this is mostly uninteresting topic, but lets not
make false statements in order to advocate any of the voting
systems.

Yesterday at 19:56, David Neary wrote:

> An example might help illustrate. Let's say there's a quota of 4,
> with 5 seats, and 6 candidates, and there are 5 Novell
> candidates (OK, it's contrived, but it will illustrate the
> problem).
>
> Now with 5 equally weighted votes, if I vote for all the Novell
> candidates my vote will effectively be wasted, because whatever
> happens 4 Novell people are going to get elected along with the
> 5th non-Novell person.
>
> In this situation, even if I like all the Novell candidates and
> dislike the other person immensely, it is in my interest to not
> vote for the Novell person I like least. This is because my 5
> votes all have equal weight, so my 5th pick gets the same weight
> from me as my favourite.

I understand how employing a certain tactics would cause problems in
theory, but have you actually voted that way last year?  I mean, this
tactics is pretty much flawed, since now you're removing one vote for
a candidate which might be ranked better with that vote than your
other Novell candidates, thus 2 of your favourites losing a spot
instead of only one (eg. if that vote to other candidate pushed
him above 2 Novell candidates for one vote, and there were at least
7 candidates).  Yeah, your tactics would work in your example of 6
total, but it would fail if there were 7 candidates, with 5 being
from Novell.

This tactics only makes sense if you're absolutely sure which
Novell candidates are going to get less votes, so you don't vote for
them (that would be a wasted vote).  And are you positive of
the ranking between Novell candidates this year?  I'm not, and I
doubt many are (otherwise, we wouldn't allow more than 4 of them to
run for the Board).  Close call from last year supports this.

So current system would be "particularly bad" for non-Novell
employees instead, if everybody voted as they felt (unless somebody
uses indicated "tactics", which I demonstrated to be seriously
flawed in my example :).

> If, on the other hand, there is a preferential voting system,
> then I can safely vote for all Novell candidates in prder of my
> preference, thus expressing my vote fully without discriminating
> against my favourite candidate. I vote 1,2,3,4,5 for the 5 Novell
> candidates, and since the #1s will be counted first for all the
> candidates, my vote goes to my favourite candidate and all is
> well. If my favourite candidate gets eliminated from
> consideration, then my second favourite gets my vote, and so on.

I understand the advantages of preferential voting system, but I
don't see them working toward Novell candidates in *practice*. On the
contrary, they might even work better for non-Novell candidates.
One can talk about tactics even here.  Voting systems should be
considered in their practical applications, not in theoretical
corner-cases.

Basically, one can find examples for both equally-weighted and
ranked-ballot being better for Novell candidates.  That was my point
from the beginning.  

Example doesn't show that it is always so.  OTOH, (counter-)example
can be used to prove negation, i.e. my giving example of Novell
employees being favoured in current system proves that assertion
"current system is 'particularly bad' for Novell employees" doesn't
hold.  Proper "proof" that any system is better on average in Gnome
Foundation case is much more complicated and involved, and I leave
that to better mathematicians than I am :).


Cheers,
Danilo