Peter,
Thanks for your comments. I'll address them inline.
On 7/22/64 2:59 PM, Peter Zbornik wrote:
> Dear Andrew Myers,
>
> this method looks interesting, as it is proportional, Condorcet and non
> STV-like.
> You write on your web-page, that: "the correctness of the algorithm
> depends on a currently unproved conjecture: that if improvement of a
> committee is possible, it can be done by replacing one member at a time".
> It would be very difficult to gain support for a method, which relies on
> an unproven conjecture.
> I see this as the biggest problem in your proposed method.
We should probably distinguish between the method and the currently
implemented algorithm. The question is whether the algorithm correctly
implements the method -- this is what the conjecture rests on. The
current implementation gives the ability to compare any pair of
committees directly, so it is possible to sanity-check the algorithmic
result.
> I guess that from the presentation every voter votes for M candidates,
> where M is the number of seats, and that the voter uses range-like
> voting for each of the candidates voted for on the ballot.
> I don't understand the two modes - combined weights and best candidate
> and why two modes are needed.
In practice, "best candidate" seems to be the mode most people want. It
supports only ordinal ranking of the choices. The combined-weights mode
is more range-like, but -- crucially, from my perspective -- the
ratings/weights assigned by one voter are NEVER compared to the ratings
f another voter. That, to me, makes range voting a nonstarter.
> You write on your web page, that: "The factor (/k/+1) may be surprising
> in the condition for proportional validity, but it actually agrees with
> proportional representation election methods developed elsewhere; it is
> analogous to the Droop quota
> <http://www.encyclopedia4u.com/d/droop-quota.html> used by many STV
> election methods"
> It could be nice, if you could show a proof on how the method achieves
> proportionality, what advantages it has to standard STV and how it
> tackles strategic-voting/vote management (for instance - give zero
> weight to the strongest competitors).
> I assume it is not used for elections anywhere, so some alpha testing
> could be appropriate.
I agree that more results about this method would be helpful. I haven't
had time to push much on that. But actually, proportional mode has been
used quite a few times for elections in CIVS. At last count, there have
been 292 proportional-mode elections, and none of them have to my
knowledge yielded the wrong result.
As one example, there is a gardening group that runs monthly
proportional polls to pick which plants should be considered "plants of
the month". My impression is that the use of proportional mode is
periodically important for this kind of poll, to prevent, say, the
orchid fanatics from taking over.
-- Andrew
----
Election-Methods mailing list - see http://electorama.com/em for list info
