Re: Constitutional amendment: Condorcet/Clone Proof SSD vote tallying
On Sun, May 25, 2003 at 02:59:48PM +1000, Anthony Towns wrote:
> From a voting nerd point-of-view, we're not really running a simple
> Condorcet vote here, we're actually running two votes simultaneously. One
> is an approval vote, where we mark every non-default option as either
> approved or not-approved; and we require that a particular proportion of
> the developer body approve each option, and that more developers approve
> it than do not. The other is a Condorcet vote, evaluated on the approved
> (non-default) options. Only if there are _no_ approved options does the
> default option win, which is to say, does the the vote get discarded.
This is a good way to frame the proposed method. I think it also
provides the best ways to attack it:
1. Approval voting has obvious incentives to strategic voting. The
electionmethods people consider it clearly inferior to Condorcet
voting, in part for this reason. Specifically, why don't you
think this is a problem with the proposed method?
The closest I've seen you come to addressing this was to say
that strategically voting something unacceptible is "kind of
petty".[0] But there is no shortage of evidence that people
vote strategically, and I'm sure many of them consider it
perfectly justifiable. What would make it petty in the context
of Debian? I can see possible answers to this, but I think it
remains a very hard question. For example, consider the biases
introduced if there is a contingent of "moral" Debian developers
who vote sincerely, and another of "pragmatic" developers who
vote strategically.
I should note that while the possibility of strategy is itself a
problem (in that it makes it harder to decide one's vote, and
may pose a moral dilemma to some), the more serious implication
is that it undermines the goals of the method. So a win over
the default option may imply that an option is "generally
acceptible", but the converse does not follow. Strategy
basically weakens any rationale for a method, including the
worthwhile ends that you have set out.
2. In combining two methods "in serial", you create a
discontinuity. As the example to which you replied shows, one
vote can be the difference between a "landslide" for A (in round
two), and elimination in round one. This raises questions of
fairness, and makes it difficult to speak meaningfully of a
margin of victory.
You could argue that the "result" of an election is only the
winner, and that concepts of continuity and margin are
irrelevant. But I think this violates most people's intuition
about elections. If the example actually happened, I suspect
many people would be angry and feel cheated.
You could (and I think you have) also argue that these
discontinuities don't happen in realistic scenarios. But in
making this claim, you have implicitly ignored problem 1. I
think that if many voters use strategy, such outcomes are
plausible (perhaps not for the quorum case, but for the defeat
by the default option case).
Andrew
PS. Although I haven't participated in the discussion recently, I
have followed it, and I honestly don't think the above have been
addressed. So please give me a pointer if you think they have.
[0] http://lists.debian.org/debian-vote/2002/debian-vote-200212/msg00101.html
Reply to: