Re: Hand Recounts of votes recorded on DREs

[Ted Selker <selker@xxxxxxxxxxxxx>]

Approach for new machines demonstrating
diversity such as compiling a multiple operating system is an exciting
goal.  See the SAVE paper:

http://www.vote.caltech.edu/Reports/vtp_WP7r.pdf

-Ted

Dick Johnson <dick.johnson@oracle.com>

12/06/2004 10:40 AM

To	"R. Mercuri" <notable@MINDSPRING.COM>
cc	Ted Selker <selker@MEDIA.MIT.EDU>, Arthur Keller <ark@SOE.UCSC.EDU>, stds-1583-stg3@IEEE.ORG, stds-1583-disc@IEEE.ORG, stds-1583-tg1@IEEE.ORG, stds-1583-tg2@IEEE.ORG
Subject	Re: Hand Recounts of votes recorded on DREs

Rebecca has a very good point.  Even more effective
at estimating lost votes would be a voting simulation using the actual
systems used in voting in the past election.  Any system based on
Windows should be capable of being driven by WinRunner; why not simply
do the experiment?  It should not be hard to simulate the actions
of diverse voters, including voter error.

This is my personal and professional opinion, and it does not necessarily
reflect the opinions of my employer.

-- Dick

Richard C. Johnson, Ph.D.
Applications Architect
Oracle Corporation
631-689-3736


R. Mercuri wrote:
As Ted says:
  
-ps.  hand recounts even with 3 people watching
usually have 2 to 3 errors
per thousand,
    

That translates to a "lost vote" rate of 0.3%, which is still
well below the unexplaned "residual" vote rate of 3-5% (at times
as high as 9%) with the unrecountable touchscreen systems.

Also note that Cast Vote Records are not demonstrably equivalent
to the ballots that the voters cast, hence they do not constitute
a "recount," only a "reprint" of GIGO generated by
the computer.

Rebecca Mercuri.