[EM] First U.S. Scientific Election Audit Reveals Voting System Flaws but Questions Remain Unanswered

[Abd ul-Rahman Lomax]

At 02:20 PM 5/8/2007, Kathy Dopp wrote:
>First U.S. Scientific Election Audit Reveals Voting System Flaws but
>Questions Remain Unanswered

Indeed. Though there is one question that has been answered. The 
Diebold machines and procedures are a disaster.

Ironically, unless we accept some kind of conspiracy theory, it is 
mysterious why voting machines are used at all. Classic cost/benefit 
analysis would generally advise against investing in relatively 
expensive equipment that is going to be used perhaps once a year or 
less, when there are much simpler methods which are accurate and 
auditable. Yet there was this rush to voting machines.... why?

Further, if one *is* going to machine-count ballots, the equipment to 
do so is essentially free. Any fax could be used to scan a paper 
ballot and the software to read marked ballots has been around for a long time.

To top it all off, it is feasible to allow images of ballots to 
become public record. A proposal has been made that not only would 
normal counting take place through officially obtained scans of 
ballots, with the actual ballots being sequestered and secure for 
later audit if necessary, but at the same time election observers 
could make scans of ballots with private equipment, providing the 
general public access that was heretofor reserved for a few observers 
and officials. If these latter scans are made public as well, the 
public will have multiple independent sources of validation of 
reported election results. As part of this proposal, all ballots 
would be assigned a scannable and automatically recognizable serial 
number (probably after being cast) so that independent counts can be 
correlated, the number would be part of the scan image. It then 
becomes possible for individuals to manually count a few ballots and 
for validation databases to be compiled piecemeal with little labor 
on the part of any given individual.

I am sure, however, that news media would want to conduct their own 
machine-assisted counts for rapid prediction of outcomes. The 
existence of such massively-redundant counting should reduce counting 
errors to the minimum that are produced by ambiguous ballots. The 
result would be vastly increased public confidence in the accuracy of 
election results, a situation which has been rapidly deteriorating 
here as more and more information becomes available about apparent 
election manipulation or, at least, incompetency.

Given that the public-access part of this is essentially free (that 
is, costs are covered privately), and that the governmental part 
could be less expensive than present practices -- once on is counting 
images of ballots instead of actual ballots, the stringent security 
that is standard practice, and which is very expensive, becomes 
unnecessary. It is cheap to keep piles of ballots locked up, what is 
expensive is to have many eyes watching whenever counting is going 
on. It's outrageous that when people want to verify counts, they have 
to pay county officials to stand and watch them to make sure they 
don't alter ballots. The precautions against alteration make sense 
when there are no images.

If there is fear that ballot images have been altered -- in spite of 
what we assume would be caution in making sure that observers are 
allowed to scan the same ballots -- then sampling of actual ballots 
could discover major manipulation, and manual verification of ballots 
against images could also be done rapidly. That is, it would not be 
necessary to totalize the ballots but simply to look at them and 
compare them with images. There is blink-processing equipment which 
would make altered images stand out like a sore thumb, and the 
auditors could move quite rapidly through many ballots.

The work of Stephen Unger should also be noted.

http://www1.cs.columbia.edu/~unger/articles/sarasota5-2-07.html