robert bristow-johnson wrote:
On Jul 14, 2010, at 5:38 PM, Kristofer Munsterhjelm wrote:
As part of tinkering with my simulator, I have found that for certain
methods, it's having problems finding disproofs of criterion
compliance.
[snip]
why need every permutation of ranking be put on the ballot (where the
voter simply checks off one)? why not just list the candidates (in a
random order) and then, beside each name, would be a row of ovals to
fill in (with 1, 2, 3...)? i have to admit that i am a partisan for
optical scan voting machines, but the same question would apply for
other ballot technology.
This isn't about actual ballots, but rather about computer simulations.
When one is doing a computer simulation, one needs a way of actually
generating ballots so that one can test whatever it is one's testing
(frequency of cycles, whether a certain method fails a certain
criterion, etc). One such model is the "impartial culture", which, when
truncation is not permitted, means that every permutation is equally
likely. I was asking how to extend that model to the case where
truncation is actually permitted, while still preserving the property
that every permutation is equally likely.
Other models exist and they may be more realistic, but IC has done well
for me in non-truncated situations so far - hence my question.
As for actual real-world ballots, I seem to remember some country (which
was it again?) that used OCR and had the voters number the candidates,
where ballots with unclear readings were forwarded to a manual check.
Such a ballot would have next to no limitation on the number of ranks,
assuming you could fit two or three digits into the space given. If the
limit given by optical scan is no concern, then you can of course use it.
----
Election-Methods mailing list - see http://electorama.com/em for list info
