today I wanna point out this project: "Election Transparency Project".
"The 2008 June 3rd election was the first time that ballots were
scanned and published on the Internet for a Humboldt County election.
Scanning began on June 23rd and took about 5 days. A set of 3 DVDs
containing images of 30,025 voted ballots (around 8.2 GB of data)
became available on July 3rd.
"The November Presidential Election was the second time that ballots
were scanned and published by the ETP. During this process, the ETP
discovered that 197 ballots were deleted by the Diebold/Premier
software used by Humboldt County to tally the vote. This software
glitch resulted in the certification of inaccurate election results.
More information about this issue can be found in the "News" section
of this website."
"BALLOT IMAGES produced by the Humboldt County Election Transparency
project can be obtained from the Elections Department on DVD."
"These images are also available on a website hosted by the Election
Administration Research Center at UC Berkeley:"
The Ballots are publics, you can download them here.
To (re)simulate the counting phase, using this particular software, The Ballot Browser, is easy to count scanned ballots.
I totally agree with this sampled audit system. The goal of this project is to verify the machine's work by (re)counting the votes in some small cities, where results no heavy the ballot scanning procedure. The end of the experiment is pretty easy, if the results are not equals, there is a counting error in some ways.Of course we wont say that if the totals are the same there are no errors on the procedures and on the machines but if the totals are different, we are sure that something wrong happened. Until the (re)counting process is Transparent, it's easy to check if the fault is in it. No faults in the (re)counting procedure ? The fault is in the voting device.
Using this procedure "The Election Transparency Project" discovered: New Errors in County election Results that caused huge political and social consequences.