Re: Condorcet Voting and Supermajorities (Re: [CONSTITUTIONAL AMENDMENT] Disambiguation of 4.1.5)

[Raul Miller <moth@debian.org>]

On Fri, Dec 01, 2000 at 01:32:00PM +1000, Anthony Towns wrote:
> Here's some possible "fairness" criteria:
> 
> 	* If there are multiple ways of conducting a vote, the outcome
> 	  should not depend on which way is chosen (only on the
> 	  preferences of the developers).

Agreed.

Comment: the current constitution is ambiguous on this issue, for the
case of supermajority.  My original reading of the constitution was that
A.6(7) applied to any ballot which combined an option with supermajority
and an option without supermajority.  The example embedded in A.6(7),
however, indicates an A.6(2) vote and not an A.6(3) or A.6(1) vote.

> 	* Ideally, there shouldn't be any more effective way of getting
> 	  what you want than simply ranking your preferences in order.
> 	  (Although this is, aiui, provably impossible to achieve)

Of course, this is a contradiction -- much discussion could happen here.

> I suspect we've also agreed that the Condorcet winner (if there is
> one) should always win. And we seem to have agreed that the winner
> should be from the Smith set, although that's apparently considered
> debatable amongst the pairwise voting cognescenti.

I'm uncomfortable saying if I've agreed to this.  The Smith/Condorcet
criteria that Buddha posted is not something I've agreed to.  So,
discussion could happen here, to make sure we were using the same
terminology.

> There are two reasonable ways of treating supermajorities, that I can
> think of:
> 
> 	a) A supermajority of developers believe this option is acceptable.
> 	b) A supermajority of developers support this option above all
> 	   others.
> 
> I believe (a) is the better case, and I think it's essentially what the
> N+1 scheme described in the constitution achieves. From some of your
> mails, I'm lead to believe you might prefer (b).

Er.. you've lost me here.

(*scratches head*)

Ok, I think you're basing this off of my earlier response to a specific
example where I thought, at the time, you were talking about a vote
which mixed a supermajority option with a non-supermajority option.

> (b) isn't quite what your system does, though (aiui). If you consider
> the Manoj and Branden's proposed vote, we might end up with people holding
> preferences like:
> 
> 	120 people prefer M, B, S, F  (Manoj, Branden, Status-Quo, F.Disc)
> 	100 people prefer B, M, S, F
> 
> This means that the ratio between M and B would be only 6:5, rather than
> 3:1, so presumably neither should win (that's what (b) would imply to me)?

If we're talking about *my* interpretation of what the constitution
is saying, and what it's trying to say, and *my* understanding of that
particular vote.  And, if we agree that a 3:1 supermajority applies to
changing the social contract:

The way I see it, when determining which votes Dominate which others,
votes which prefer an option with a supermajority are reduced to an
appropriate fraction (in this case 1/3).  But, 40 Dominates 33 1/3,
so M wins.  [Also, 73 1/3 -- the total number of votes for M, exceeds
quorum of 37, so there's not a problem with not enough people voting.]

[since the rest of your message just seems to build on this point,
I've just left it off.]

Thanks,

-- 
Raul