Hi Jameson,
--- En date de : Mar 8.3.11, Jameson Quinn <[email protected]> a écrit :
I think that this kind of investigation of strategy in realistic monte-carlo
simulations is important. Two comments:
1. Do you plan to share your source code? I'd encourage you to do so,
preferably under some kind of open-source license (including just public
domain).
It's not out of the question, but I won't do it while I'm still working on it.
The main reason
I would do it is to open the results to scrutiny, or to clarify how something
works.
2. I've been thinking of how to extend Yee diagrams to show strategic
vulnerabilities. So far, I'm thinking of starting with a interactive
one-dimesional Yee diagram with three candidates, and using method DNA to show
separate strategic and counterstrategic possibilities in separate lines. In
those terms, runoff-style methods (including my recently-developed MCA-Asset
and GMCA) are somewhat confounding, because a first-round strategy doesn't
carry over into the second round, so they effectively expand the range that the
DNA must cover to include both rounds (although I think that certain ballots,
such as A>B>C and then C>B>A, can be discounted).
For your simulation, I wonder if it would be possible to include such methods,
by assuming that voters would always be honest in the second (two-candidate)
runoff round? Of course, pushover strategies and counterstrategies would become
important for such systems.
I should note I think, that this new simulation doesn't use DNA. I have to code
the
methods. I created actually three predecessors that do use DNA, but I kept
running
into difficulties.
Skip unless interested:
One problem is that I wanted factions to be able to compromise and just vote
for the
second-favorite, automatically electing them under all considered methods. If
you use
a "faction-first" approach, you have the problem that multiple factions may
want to
compromise, and the DNA doesn't (easily) tell you how to resolve that, so it's
unclear
what result (as utility) to report to the voters. If you use a "scenario-first"
approach, so
that factions see a result and may unilaterally change it, you can ban "multi-
compromises" but you'll never see a burial attempt backfire because no one will
knowingly "move to that space."
Anyway, to answer your question, it is technically possible to write an election
method that can read voters' minds. There's probably nothing wrong with it
either,
in the case of a second round between two finalists. So, yes, I'll consider such
methods. I'll start with top-two runoff, and a couple of types of VFA ballot
runoffs.
Thanks.
Kevin
----
Election-Methods mailing list - see http://electorama.com/em for list info
