[SGVLUG] Vote stealing

[Chris Smith]

On 9/19/06, Emerson, Tom <Tom.Emerson at wbconsultant.com> wrote:
> Of course, should a machine be compromised, noone would know UNLESS a
> recount was requested/required, and then it could still come in within
> the "margin for error", and if it does, which way do you think people
> will believe? The machine or fallable humans counting by hand?

The fallable humans. Though they are fallable and slow, having triple
verification tends to bring the error rate down to a consistent and
generally fair level. The problem is that it is very time consuming.
This is exactly why the paper ballots would be the votes of record,
and the computer system would simply for getting a quick and
reasonably accurate vote. If the results are surprising, someone will
ask for a recount anyway.

> Personally, I'd say that any "recount" situation would invalidate the
> entire set of results for a given location -- the mechanics of counting
> by hand almost completely guarantees the totals won't match, and in that
> case, you cannot believe EITHER source as "accurate"  (kind of like
> carrying [or relying on] an odd number of clocks on a boat...)

You'd be surprised. In another life I worked at a cashier in an
eletronics store. I probably processed at least a few hundred cash
transactions a day, and after a while the cash I had in the till
seemed to precisely match the cash the system thought I was going to
have. It turns out that when it comes to reading and talling numbers,
even one human is surprisingly accurate, and three humans tend to
produce repeatable results.

> You see, I'd consider it "odd" if a recount matched the machine count
> exactly for anything over a few hundred votes. ;)  [of course, depending
> on the readability and irrefutability of the "printed" version of the
> votes cast, not to mention that one ballot might stick to another and
> not get counted, or get dropped on the floor and re-counted, etc.]

It's worth noting that human verification of hand written ballots tend
to produce consistent results no matter how many times you recount.

-- 
Chris