[E-voting] Question on optical scan

[Catherine Ansbro]

Brian,

These are exactly the concerns I am worried about, too, in relation to 
use of gummed stickers. 

They'd be hard for many people to handle, and there could be chaos with 
OCR machines getting gummed up, or people not getting them stuck on 
fully, or putting them on crooked.

It amazes me how many problems there were/are in the USA with the 
OCR-reading machines.  In some of the documents that have been posted, 
the manufacturer specifically points out that each ballot has to be 
visually inspected because there are a wide range of problems the 
machine won't catch.  This means you need training and absolute 
vigilance on the part of the poll workers.  (And in the USA it appears 
the poll workers didn't bother checking the ballots at all.)   Just one 
example that arose in the Ohio revote, was an oval that had a tiny dead 
bug on it, (and a different oval filled in by the voter), but to the 
reader it looked like an "overvote" (2 votes) and was disqualified.  
There were also problems having to do with setting for the thickness of 
the ballot, problems if the ballot had been folded, problems if the OCR 
sensitivity levels were not correct, etc. 

Some of these could be managed, but the sticky label part worries me, 
and also having a 30-sec time limit to change your choice is also too 
restrictive.  And if the label can be changed multiple times, then 
you're asking for problems with labels coming off--accidentally, or 
perhaps even repositioned  fraudulently. 

The ironic thing is, I'd like the proposed DREs if they were just used 
for marking a printed ballot and nothing else (or perhaps, for an 
indicative count only similar to a tally that would have /*no*/ official 
weight).  The trouble is, if there were an indicative count people might 
get very lax about attending and scrutinizing the formal count of the 
paper ballots.  And no doubt their would be opposition to having 
expensive machines just to print the ballot.

Surely it would be possible to have a really cheap, small "device" just 
for printing a paper ballot, but not counting them?  There wouldn't be 
the same security storage issues, it should cost a lot less because the 
programming would be very simple and there'd be no need for anything 
proprietary.  Shouldn't be too different from a cash register type of 
device.

Using such a device to print a legible ballot would make (manual) 
counting faster, it could reduce the few inadvertent spoiled votes, and 
would reduce the number of contested ballots that have to be reviewed by 
the candidates.

ANY kind of counting device will involve potential for user error due to 
many possible reasons, and it would be impossible to know how many 
errors had been made unless you did a manual recount--so why not just 
keep the counting by hand?

Catherine


Brian O'Byrne wrote:

>Quoting Margaret McGaley <Margaret.McGaley at redbrick.dcu.ie>:
>
>  
>
>>On Wed, Jan 19, 2005 at 02:33:32PM +0000, evoting at stusoft.org wrote:
>>    
>>
>>>Stickers sound a little messy to me.
>>>
>>>      
>>>
>>Why?
>>    
>>
>
>I'd be worried about problems like:
>- Stickers falling off immediately after the OCR step.
>- Stickers getting folded in such a way as to stick two ballots together.
>(Consider where the center of the sticker is fully and properly stuck but an
>edge has been folded back. Another ballot lands on top and gets stuck.)
>- Ballots getting folded so that one sticker is stuck on two places. (A
>variation on the same theme. The sticker has a folded edge or corner; not
>enough to alert the scanner but enough to be a problem when the ballot is
>folded.)
>- Arguments arising if people find odd stickers in the ballot box.
>- Mechanical/reliability problems with the OCR equipment. It is one thing to put
>a single flat piece of paper through a machine, it is another to put through
>stickers. I could imagine rollers getting gummed up or stickers transferring to
>rollers by accident.
>
>Some of these are a little paranoid, and could be overcome with a good design. I
>keep reminding myself that punch cards seemed like a reliable and well-trusted
>technology before the 2000 US presendential elections...
>
>Brian.
>  
>