Re: Error rates

From: Richard C. Johnson <dick_at_iwwco_dot_com>
Date: Mon Jun 20 2005 - 09:55:42 CDT

That an error-free election may not be possible does not mean we should "accept" any level of error, whether system failure or malicious in origin. Rather, each detected error should cause an investigation (with public results) into the root cause of such error. Of course, elections need not be invalidated on discovery of error; rather, error needs to impeach only those elections where the error and not the votes determine the candidate elected.
Even the notorious DRE can be tested, should be tested, and the test results should be made public. Can the DRE handle load? Can it be hacked? Interested voters want to know.
It boggles the mind the degree of faith-based reliance on partisan election officials it takes to accept the DRE at face value. Above all, don't dismiss the concerns of scientists, engineers, and the public about electronic voting honesty and reliability without providing for some rational, test-based reason for believing in the adequacy of the machines in use. It seems to me that we have ample reason for distrust, rather than the other way around.
-- Dick

Stephanie Frank Singer <> wrote:
Here are a few relevant excerpts from the 1990 FEC standards. I'm not
suggesting that we revert to these standards, but I do think that
"error-free election" is a Platonic ideal, and that we would be wise to
set (as the FEC did) an acceptable error threshhold.

An error threshhold has no teeth if it is not clear how to measure that
error. So we should also specify the kind of test needed to check for
error. It is also important to say what the consequences of a failed
error test should be.


3.2 Performance Characteristics Performance characteristics for
voting systems represent the combined operational capability of both
system hardware and software. Accuracy, as measured by bit error rate,
and operational failure are treated as two distinct attributes in
operational testing (exclusive of code review). During system
performance, the desired system-level error rate shall be no more than
1 in 10,000,000. Other performance criteria for subsystem accuracy
are presented, as applicable, in sections that follow. Quantitative
system reliability shall be measured by the number of unrecoverable
failures in a timebased operating test consisting of no less than 163
cumulative hours (with no failures). All performance requirements
contained in Section 3 Hardware shall be met under operating and
nonoperating conditions. Recording Accuracy DRE systems shall accurately record
each vote and ballot cast. Accuracy as here defined means the ability
of the subsystem to detect every selection made by the voter, to add
permissible selections correctly to the memory components of the
device, and to verify the correctness of each of these operations. It
also means the ability of the device to preserve the integrity of
voting data and ballot images (for DRE machines) stored in memory
against corruption by stray electromagnetic emissions, and
internally-generated spurious electrical signals. Recording accuracy
may be achieved or enhanced by the incorporation of multiple detection
and memory elements that employ device polling techniques. Corrected
data errors shall in these instances be logged by the system. The
error rate measured by these criteria shall not exceed one part in one
million, as applied independently to the voting data memory and to the
ballot image recording devices. Recording Reliability Recording reliability refers to the
ability to sustain accuracy during the required operating period. DRE
systems shall reliably support the collection and retention of voting
data in the voting device and the transmission of voting data among
voting devices. The retention, transmission, and collection of voting
data shall be error-free for at least 163 hours, as dictated in
Subsection 3.4.3 and Appendix F, Subsection F.4. Memory Stability P&M memory devices, used to retain
control programs and data, shall have demonstrated at least a 99.95
percent probability of error-free data retention for a period of 6
months, under the environmental conditions for operation and
non-operation contained in Subsection 3.4.6.

Stephanie Frank Singer, Ph.D.
Campaign Scientific

OVC discuss mailing lists
Send requests to subscribe or unsubscribe to

OVC discuss mailing lists
Send requests to subscribe or unsubscribe to
= The content of this message, with the exception of any external
= quotations under fair use, are released to the Public Domain
Received on Thu Jun 30 23:17:10 2005

This archive was generated by hypermail 2.1.8 : Thu Jun 30 2005 - 23:17:11 CDT