[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Debian's custom use of Condorcet and later-no-harm

2014-02-28 17:50 GMT+01:00 Ian Jackson <ijackson@chiark.greenend.org.uk>:
Thue Janus Kristensen writes ("Re: Debian's custom use of Condorcet and later-no-harm"):
> I am not completely sure, but I think both ways accomplish the same thing,
> if you always only use the >= criterium.

Actually, they are not identical. Your proposed change would give the
casting vote more power than mine, by allowing the casting vote to
override FD when voting for B in the example below.

> My way seems more flexible though, since you can use it with >= or >, or
> 2/3 majority over FD requirement, and still get sane results.

Consider a TC ballot with 2 options, A and B, plus FD.
   4 voters: A > FD > B
   4 voters: B > A > FD

In the current system, the winner is A.  This is arguably a result of
manipulation by the A-preferers.  Voters should not be incentivised in
this way to prolong "discussion".

I feel that whether a group consisting of half the committee is allowed
to prolong discussion is a separate issue. The point of my proposed
change is that you can no longer win in this way by manipulating FD
(though there are of course still well-known ways to manipulate Condorcet).

In your proposal, the outcome is, I think, still A.  The Schwartz set
is A, B, but B fails the >FD criterion.  (I assume you mean to apply
A.6.3 before the casting vote, rather than afterwards.

No, I specifically said that the FD criterion should be applied after the
casting vote.
If afterwards,
then the casting voter gets to choose beteen A which becomes A and B
which becomes FD, which is bizarre.)

I had actually considered this exact scenario (in an private email to Russ
Allbery), and I don't consider that result bizarre:

* If the casting vote chooses A, then there is 4+casting vote for A, better
  than B, and A passes the FD criterion. A winning is fair.
* If the casting vote chooses B, then there isn't any majority for B over
  FD, so FD winning is fair. A doesn't have a majority over B or FD either,
  so obviously A shouldn't win either.

These result are consistent with what I would expect an abstract
magically fair voting system to produce, given the votes. Yes, you can 
force FD if you control half the votes, but that can be an honest vote,
and not just used for manipulation, so has to be allowed.

Regards, Thue

Reply to: