Electronic Voting (was: Re: [lug] Warning about RTENet.com)
J. Wayde Allen
wallen at lug.boulder.co.us
Tue Dec 5 11:22:47 MST 2000
On Mon, 4 Dec 2000, Sean Reifschneider wrote:
> On Mon, Dec 04, 2000 at 04:14:44PM -0700, Nate Duehr wrote:
> >I think a discussion of how to do national voting systems on Linux would
> >be interesting, but not probably very realistic! (Heh...)
>
> There are a number of folks worried that the current election problems will
> cause people to try to move towards a computerized voting system. The
> problem is that it often eliminates the ability to do a manual recount.
OK, there are several problems here. The main one being that in every
measurement/sampling process there are uncertainties. Hand recounts,
etc. don't minimize these uncertainties and may even increase them since
it makes possible the subjective interpretation of the sampling
process. It isn't so much that there is a problem, but rather that we are
trying to measure a difference that is smaller than the sample standard
deviation. Re-counting or re-sampling, whatever, isn't going to solve
this. The only real solution would be to develop a measurement system
that would lower the overall sampling uncertainty, and that may or may not
really be possible.
For example, if we estimate the current population of Florida to be
130,000,000 and say that there are on the order of 20,000 votes in
question (no I don't know what the numbers actually are), that means that
we have an uncertainty on the order of (20,000/130,000,000)*100 =
0.02%. For most measurement processes that aint bad. It just doesn't
make sense to try and distinguish differences on the order of
(600/130,000,000)*100 = 0.0005%. To effectively measure this you'd need
to create a system that would have an uncertainty less than this, and
preferably by at least a factor of two. That is a pretty tall order with
real science and physics. I'm not at all convinced that anyone can
achieve this with a system trying to measure peoples opinions.
So, how about getting back to Linux!
- Wayde
(wallen at lug.boulder.co.us)
More information about the LUG
mailing list