On Sat, Jun 30, 2007 at 10:00:47AM +0300, Antti-Juhani Kaijanaho wrote:
> On Fri, Jun 29, 2007 at 02:43:24PM -0500, Manoj Srivastava wrote:
> > It should be relatively straight forward for Devotee to find the
> > winner, take the winner out of contention the next round, find the next
> > winner (ignoring any pairwise contests dealing with any candidate no
> > longer in the contest), and continue until the number of candidates
> > desired has been reached.
> This is no doubt true.
> As I mentioned in another mail, this procedure does have the problem of not
> delivering proprtional results.
> A scenario.
A simpler scenario. A bunch of candidates divide themselves into
essentially two parties, people focussed on free software, and people
focussed on our users. As it turns out, one group has 60% support within
the project, the other group has 40% support within the project. There
are six candidates, and three places to fill. Votes go along the lines of:
60% [ 1 ] A-1
[ 2 ] A-2
[ 3 ] A-3
[ 4 ] B-1
[ 5 ] B-2
[ 6 ] B-3
40% [ 4 ] A-1
[ 5 ] A-2
[ 6 ] A-3
[ 1 ] B-1
[ 2 ] B-2
[ 3 ] B-3
Condorcet gives the winner as A-1. Excluding A-1 gives you a Condorcet winner
of A-2. Excluding A-1 and A-2 gives you a Condorcet winner of A-3.
A more desirable outcome, IMO, would have given B-1 a seat in the above
circumstances. Which is what proportionality is all about...
Whether A is "free software" and B is "our users" or vice-versa is left as
an exercise for the reader. ;) Other plausible scenarios might involve
soc-ctte candidates promoting "freedom of speech" versus "improving
signal:noise".
Cheers,
aj
Attachment:
signature.asc
Description: Digital signature