[EM] HR811 and Federal paper trail legislation

[Abd ul-Rahman Lomax]

At 07:57 PM 5/25/2007, James Gilmour wrote:
>I am not familiar with US law on these points, but in the UK there 
>is NO ballot identification
>information on the face of the ballot paper.  That information 
>(ballot paper number  -  and bar code
>if for machine counting) is all on the back of the ballot paper.

Here, there is supposed to be no identifying mark on the ballot at 
all that could be used to identify the voter. I think that election 
rules generally will discard ballots that have extraneous marks on 
them. But an exception is write-in votes, which a voter could easily 
use to make his or her ballot identifiable.

My suggestion here is that an identifying number be added to the 
ballot after it is cast. This should be machine-readable, could be a 
bar code. But it would be on the face and it would be in the images. 
The reason is so that independent counts can be done and compared on 
a ballot-by-ballot basis, which would be far more efficient in 
identifying counting errors than if there is no such identification. 
Of course, if there is a file sequence number, that would do. (The 
118th ballot in batch 1096.) However, marking the ballot -- I'd use a 
sticker, perhaps -- makes it very direct and simple to use. Plus it 
allows independent observers to photograph the ballots and have the 
same identification number, crucial for providing the multiple 
independent validation that would quickly detect attempts to hack the 
system and alter the images.

I find it fascinating that aspects of what I've suggested are already 
practiced in the U.K. Here, we see people dismissing them out of hand 
as impractical and dangerous. To me, they are simply obvious, and I 
don't think of this as rocket science. I'm not an elections expert, 
just someone who looks at things and does not assume that they way 
they have always be done is therefore the best. Sometimes it is! But 
sometimes not, as well.

You don't know until you consider alternatives, and you don't 
*really* know until you try them and work out the new bugs. That's 
what engineers do. They do things that some of these "experts" would 
simply reject because of this problem or that danger. The engineer 
looks at the same problems and attempts to solve them, and then the 
final product is successful if the result is an improvement, overall.

>   Both side were scanned in the
>recent elections, but when scanned images were displayed (on 
>computer screens, on large screens and
>projected onto very large screens) for adjudication, it was only the 
>face of the ballot paper that
>was shown.  No-one could identify the ballot paper from the image.

I don't actually like it. What this does is to depend on the 
reliability of whoever is arranging for those images to be shown. 
That's a point of vulnerability that could easily be eliminated. If 
the ballots were serialized on the face, same process, but now it 
would be verifiable. If someone shows you the wrong image, you could 
later find out!

But the idea of projecting images as described so that many could see 
them at the same time is the kind of thing that should make the 
process more efficient. This is all part of what I immediately 
recognized when the idea of imaging ballots and counting from the 
images instead of from the ballots occurred to me. You can have 
multiple groups counting the same ballots simultaneously, if you have 
images, with no security risk to the original ballots.

>The information that linked the ballot paper to the elector to whom 
>it was issued was not available
>in electronic format anywhere.  That info was recorded on a separate 
>tally list (paper record only)
>that is kept separate from the completed ballot papers.  The ballot 
>paper numbers and the list of
>issued ballot papers with electors' numbers can be brought together 
>for examination ONLY by order of
>a court.

I like this. I've always wanted it to be possible to track down -- 
with serious safeguards -- an actual vote. It would be very difficult 
politically to implement this here. I've gotten a lot of flack just 
from suggesting ballot imaging as somehow -- I can't imagine -- 
compromising voter identity.

Serializing ballots so that the images can be quickly and directly 
connected with the paper ballot is one thing that could make it all 
work very efficiently, and I see no security risk from it.

It would be cheapest to serialize them before giving them to the 
voter, but this, then, *does* create a risk to the voter's privacy. 
Better if the ballots are serialized immediately when the ballot box 
is opened, first task. Cheap and fast with stickers, but other means 
are possible. For example, the ballots could be run through an impact 
printer or a printer with red ink to add a number. Also fast and very 
cheap, cheaper even than labels, with less human handling.