[EM] Open Voting Consortium for e-voting?

[Ken Johnson]

>Message: 1
>To: election-methods-electorama.com at electorama.com
>From: Ernest Prabhakar <drernie at mac.com>
>Date: Fri, 2 Apr 2004 13:26:22 -0800
>
>...
>
>  In the consortium's system, a voter approaches a touchscreen and fills  
>out a form, which is promptly printed out as an official legal ballot.  
>The voter can cast his vote or review choices before the ballot is  
>printed out with the voter's choices checked off. The voter then places  
>the ballot in a private envelope and places it in a ballot box.
>
>  That approach is in sharp contrast to most of the electronic voting  
>machines now in use. Most of today's voting machines use proprietary  
>source code and machines--and leave no paper trail, making it difficult  
>to audit counted votes. ...
>
>  
>
I'm all for paper ballots, but what good are they if you don't actually 
use them to verify the electronic tally? Following is a message relating 
to this topic that I posted recently at 
http://verifiedvoting.org/forum/topic.asp?TOPIC_ID=198 . (Note: Netscape 
is unable to access this site, but IE works.)

Paper ballots present a conundrum: You may never know whether the 
machine count is valid unless the paper ballots are manually counted, 
but you can't justify a full recount unless there is sufficient cause to 
doubt the machine count. The solution to this problem is to mandate a 
"statistical recount" as part of routine election certification 
processes. Each paper ballot would have a piece of information 
associating it with its associated computer database record from which 
the machine count is generated. A small, random sampling of paper 
ballots is manually inspected and compared to the corresponding data 
records to confirm that the ballots are correctly represented in the 
database. Similarly, a small, random sampling of data records is 
selected and compared to the corresponding paper ballots to ensure that 
no "virtual ballot stuffing" has occurred. A very small sampling, e.g. 
1000 out of a total of 10,000,000 ballots, would typically be sufficient 
to validate the machine count with 99.99% confidence. (A conventional 
full, manual recount may actually be LESS certain than this sample-based 
recount because of counting errors.) This process would make it possible 
to routinely validate election results beyond reasonable doubt without 
doing a full recount and without having to validate all of the detailed 
mechanisms and processes (e.g. software) by which the information on the 
paper ballots is transferred to the database.

Ken Johnson