Carp of the Day

home | archives
"Indeed, the rage of theorists to make constitutions a vehicle for the conveyance of their own crude, and visionary aphorisms of government, requires to be guarded against with the most unceasing vigilance."
     -- Joseph Story
     Commentaries on the Constitution of the United States
     Book III, § 1857.
 

Tuesday, August 05, 2003

Rick Hasen on the litigation whirlwind in California. Hasen sets out his opinion that, of the various suits challenging the recall, or parts of the recall, only one has apparent merit: The suit predicated on the idea that the use of punched-card ballots in some places, but not in others, violates Equal Protection guarantees. And why does it violate those guarantees? Because, according to the litigants, punched-card ballots are more likely to be uncounted, or counted incorrectly, than other ballots.

This argument makes no sense to me. I've worked with computer systems using punched-card input, and it was never been my perception that the card readers were inaccurate. Slow, yes; tedious, yes; bulky, yes; inefficient, yes; expensive, yes, yes, yes; inaccurate, no way. The reader might balk at a particular card -- say, because it had been folded, or had been processed so many times that it was fraying -- but you just copied the card, plugged the new card into the deck, hit the START button, and off you went. Once a card went through the reader, a faithful image of that card sat in core, ready for processing. Some of the systems I worked with put critical binary data on those cards (and it was a real bitch to copy those if the reader refused them), and thus would not have been able to tolerate having a card read wrongly, and yet I do not recall a single instance of a program bombing because the card reader actually accepted a card and transcribed it wrongly into memory.

And we're not talking, here, about such things as program decks, where the order of the cards is important, so that dropping a tray and having the cards scatter all over the floor is a first-rank catastrophe. These are votes; the order in which they are processed does not matter; you just pick them up again, stack them, make sure they're all right-way-up and front-forward, square the deck, and drop them in the hopper.

Now I can see how the occasional necessity of transcribing a card (because the reader refused the original) might require some controls: Recording the original and replacement ballot card numbers, for example, so that they might later be compared in an audit, and having multiple persons watch the copying, so that the risk of collusion is reduced. And I can also see how laxity in these matters would reduce confidence in the integrity of the result. But these problems aren't specific to punched cards; they are inherent in any balloting system, especially those which seek to preserve ballot secrecy.

All that I have addressed here is the hardware that reads the card into the machine. I haven't touched upon the software that takes the memory image of that card and tallies the votes represented therein. Just because the card reader is honest, doesn't mean the software is, too. But that objection applies -- even more so -- to wholly electronic means of balloting, because they produce no evidence subject to wholly independent audit. At least, with punched cards, one can, in the last necessity, eyeball the cards themselves, and so determine whether the machines are functioning properly. If the election is wholly electronic, though, what is there to look at? The audit trail was written by the software itself; if someone took it into his mind to skew the tallies, he has the means at hand to skew the supposed evidence as well.

(And yes, yes, I'm familiar with all the protections against such things offered by modern database systems. None of them are proof against a sufficiently determined user with system admin privileges. It would take a lot of work to cover up the trail completely, but hey, there's a lot riding on the results.)

What it comes down to, in the end, is, nothing can substitute for the integrity of elections officials. The best we can do is to provide means of spot-checking their veracity. These suits, so far as they have any merit, are predicated upon the existence of some magic gizmo that will still do its job properly in the face of every attempt that can possibly be made to pervert a tally, when the simple fact is, no such gizmo exists, and the ones being offered up are significantly worse, on the integrity score, than those challenged punched cards.

-- posted by Clayton 8/05/2003 09:13:00 PM


Comments: 0

Post a Comment

<< Home

 

Links

 

For the more forensically inclined