[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: Null Hypothesis

Nick,

In a perfect world, we would be able to solely examine machine performance and judge it upon its results. Because we do not know and cannot visually see and verify how the machine is actually "recognizing" and recording and then converting that recorded image into a digital version for later tabulation, and then again not being able to visually audit that that tabulation process is not open to errors, all we can judge the machine by is their results.

In Wake County, North Carolina, an ES&S machine that was tested (put thru the wringer of "logic and accuracy tests") and then certified as good to go still dropped 400+ votes and it was only noticed because it happened during early voting and not under the pressures of election day itself. Yet in a test against a paper alternative, the machine would come out fine.

As our friend Al Kolwicz points out, so much more than just results matter. It is the conditions of the election event (not just election day but also the handling of all processes leading up to the election) that can or do affect performance of election systems. It is possible for someone intent on "fixing" or rigging a race to tamper with machines just as much as it is to tamper with the old lever machines or to plan to "lose" a ballot box of paper. The entirety of the process, from preparation to oversight to execution to follow-up, all matter in examining the integrity of the voting process. It is possible that the warehousing function allows knowledgeable employees to alter or insert new instructions into software or into memory cards. It is possible that vendors could be allowed to insert last-minute software version "updates" at polling places, unknown to county administrators and conceivably reasonable to the unaware election judge volunteer who might think this is business as usual.

You are correct that the null hypothesis may be the more proper approach to take (i.e., prove that the machines DO work instead of that they DON'T work) but it remains only one piece of voting integrity. I agree that paper receipts provide the means by which to perform a meaningful recount, but with Colorado law prohibiting non-identical methods of counting/recounting, we would NEVER LOOK AT THE PAPER IN THE FIRST PLACE.

Colorado needs to remove this law from the books. But we also need to reclaim the paper balllot method as the supremely verifiable process in elections, and have those paper ballots be the medium that is itself counted and treated with the full force of law that already exists to protect balllots.

Bob McGrath
Director, Coloradoans for Verified Voting
www.cfvi.org


From: Nicholas Bernstein <nicholas.bernstein@xxxxxxxxxxxx>
To: Robert Mcgrath <mcgrath_mcnally@xxxxxxx>
CC: evan@xxxxxxxx, cmehesy@xxxxxxxxxxx, davide475@xxxxxxxx, david.ellington@xxxxxxxxxxx, summerstorm03@xxxxxxxxxxx, donna@xxxxxxxxxxxx, jpezzillo@xxxxxxxxx, mlambie@xxxxxxxxx, texico86@xxxxxxx, peter.raich@xxxxxxxx, TresCeeA@xxxxxxx, pklammer@xxxxxxx, AlKolwicz@xxxxxxxxx, cvv-discuss@xxxxxxxxxxxxxxxxx
Subject: Null Hypothesis
Date: Fri, 16 Apr 2004 11:08:13 -0600

Bob,

I think you've hit upon a very trenchant issue. I would repose it like this, "In looking at electronic voting systems, who has the burden of proof? Is it up to proponents of electronic voting to prove the superiority of electronic systems, or the detractors to prove the fallibility?"

Proponents say, "Just look at Florida in 2000. Paper doesn't work!"
Detractors say, "Just look at the security holes in Deibold machines. Electronics can't be trusted!"

I would claim that they are both right. Neither method should be trusted based solely on the failures of the other. The only way to reliablilty compare the two is to have them go head-to-head. And the only way to do this is with a paper printout.

Call it a "ballot" or a "receipt". The difference is moot. Until one system has proven its superiority over the other, there should be a complete count of paper and electronic ballots. If a unique ballot ID can link the two, anomolies can be investigated on a case by case basis.

There are those people who are afraid of such a head-to-head comparisons because they worry that the electronic count will trump the paper one. I think the question of which method should be trusted in case of anomolies is minor compared to coming up with a measure of the relative accuracy. If we get nothing else, we learn that there is an x% discrepancy rate. That, in itself, is very valuable.

If we really want to unambiguously resolve discrepancies, then we need a method for letting the voters themselves resolve them. I have proposed variations on this theme before, but here is another try. At the voting place, a SECOND RECEIPT is printed that contains ONLY THE UNIQUE ID. The user can take that home. Discrepant ballot IDs are posted. If your ballot is one of those in question, your second receipt provides you admission to a session where you can cast a clarification ballot. Obviously there are many details to work out, but this system would maintain anonymity, guard against miscounts, and let voters know when their ballot has been invalidated.

Nick



--
If you're not outraged, you're not paying attention.



_________________________________________________________________
FREE pop-up blocking with the new MSN Toolbar ? get it now! http://toolbar.msn.com/go/onm00200415ave/direct/01/