Monday, September 10, 2007

With Standards This Low....


I know it is hard to get an opinion piece published in the Washington Post, which makes it all the more amazing to me that they so happily publish columns as bad as yesterday's Brookings/AEI piece on how horrible it would be to have a paper trail to track electronic voting. From the start, you know you're not going to get anything which resembles a reasoned analysis:
When early jet aircraft crashed, Congress did not mandate that all planes remain propeller-driven. But this is the kind of reactionary thinking behind two bills that would require that all voting machines used in federal elections produce a voter-verifiable paper record.
Perhaps the high quality research at AEI and Brookings suggests otherwise, but I don't recall that early jet engines were in particularly wide use in passenger aircraft, nor do I recall that the use or nonuse of jet engines posed a threat to the exercise of voters' civil rights. I suspect that if either were the case, Congress may have been much more "reactionary" in regulating early jet aircraft.
Paperless Direct Recording Electronic (DRE) machines, those where votes are entered into computers and stored only in computer memory banks, have encountered numerous failures and no longer inspire public trust.
Well, that's a compelling way to open an argument for doing absolutely nothing about the problem.
Unfortunately, paper records are no panacea for the shortcomings of machines, and mandating paper removes the incentive for researchers to develop better electronic alternatives.
I might cynically observe that the interests which are bankrolling the propaganda effort against paper trails may be inspired to fund such research. Even accepting that paper trails can be problematic, though, is the solution really to ignore the available option because of a pie-in-the-sky hope that somebody, someday, somehow will develop an electronic alternative? One could almost argue....
When early jet aircraft crashed, Congress did not mandate that all planes remain propeller-driven. But this is the kind of reactionary thinking behind two "think tanks" that would bar electronic voting machines used in federal elections from producing a voter-verifiable paper record due to problems with the first generation of printers.
The reasoning for why we should not have a paper trail is rather absurd:
Paper verification looks good on, well, paper, but it is not the cure-all some of its proponents believe it to be. More than two centuries of U.S. elections have shown us that paper is at least as susceptible to chicanery as electronic records. Paper ballots can be modified, counterfeited or destroyed with relative ease. It is not at all clear that they constitute a more reliable medium than electronic records.
But here, obviously, the paper trail is not the primary mechanism for counting votes - it's a safeguard. It is conceivably possible that somebody could tamper with the electronic vote count, inspiring a recount of the paper votes. But what are the odds that the people responsible for the electronic tampering could also gain sufficient access to the paper ballots to subvert the paper vote count as well? And if they have access, what are the odds that they could tamper with the paper ballots such that the faked paper and faked electronic vote counts would match? With a paper trail, tampering with the vote count becomes exponentially more complicated to achieve, and makes it exponentially more likely that the tampering would be detected.
These are not the only problems with paper records. Mandatory paper verification would be a disappointment for blind voters, who could not confirm that their votes were properly cast in the same way that others' were.
I mean no disrespect to blind voters, or to the goal of making everybody's voting experience as equal as possible, but we have a couple of centuries behind us where blind voters have not been able to cast their ballots or confirm their vote in the same manner as other voters. While we can strive for a verification solution that is equally accessible by both blind and sighted voters, our inability to achieve perfection is not a reasonable basis for eliminating safeguards - and it should be recalled that ATM-style voting machines are far from perfect in their accessibility. This particular "problem" seems far from insurmountable - how about having a station with a scanner that can "read" the paper receipt back to anybody who puts on a pair of headphones and scans their ballot? Further, it is absurd to argue that an appropriate "solution" would be to eliminate safeguards for everybody, any more than it would be reasonable to ban voting in any district in which blind voters cast their votes in exactly the same, unassisted manner as any other voter.
Also, the counting of paper ballots, if required by a close election, could prove an unwieldy task and take tens of thousands of hours of work.
When it comes to recounts, that's the nature of the beast. I don't follow why "it's hard" should stand as a justification for banning recounts, or why this "problem" would be somehow unique to ATM-style voting machines.
Further, the printers that produce paper ballots are especially susceptible to mechanical failure; as many as 20 percent fail on Election Day, according to Senate testimony this summer by election expert Michael Shamos.
Wow... so, they're like early airplane jet engines, and Congress should ban them!

Back when I was a food service manager, we had this devices called "cash registers". They kept a "paper trail" of every transaction which was processed by a cashier. During their years of service, despite a complete absense of any "maintenance" beyond the changing of ribbons and rolls of paper, the registers were remarkably reliable. Given that these voting machines are "ATM-style", I will note that the only time I have ever experienced the failure of an ATM receipt was on an occasion when the machine had run out of paper. If the voting machine printers have a 20% failure rate, it seems reasonable to infer that the root of the problem is poor design and manufacture. Perhaps that shouldn't surprise me, given that the companies producing these machines seem extremely hostile to the call for a paper trail, but the unacceptably high failure rate is not something that justifies abandoning a paper trail. It is something that justifies using better manufacturers to make the printers. In the interim, machines could be designed such that printers can be easily swapped out, and upon the detection of a problem the local voting authorities could switch a working printer for one that proved defective.

The author describes two technologies which could take the place of a paper trail:
A system called Prime III, developed by researchers at Auburn University, would employ a separate electronic "witness" in each voting booth. The witness, which would operate independently of the DRE machine, could more efficiently double-check the DRE's tallying of votes while safeguarding privacy and being more accessible to the disabled.
There is some appeal to such a system, if in fact it can be created and implemented while securing the two sets of data against tampering. If the same people have access to both sets of data, and have equal access to change the data or to destroy the audit trail, the safeguard is illusory. Further, this system is vastly more technically involved and costly as compared to a printer. And unless I'm missing something, the "audit" and "recount" mechanism would be to review screen captures of electronic ballots - that is potentially significantly more burdensome than a review of paper ballots (although it should be possible to design a recount mechanism that will recognize the entries on the screen captures).

And then there's the second alternative:
Another system, Punchscan, designed by a team at the University of Maryland, offers an exciting array of features: After casting their ballots, voters can go to a computer and use a receipt to view their individual ballots online.
The Punchscan system is better than that description makes it sound. In this system, each voter receives a paper ballot which they fold and mark with an ink dauber. One half of the ballot can then be discarded, and the other half of the ballot (not a receipt) scanned to verify your vote. The first problem I see with this are that it is a cumbersome verification process, and I somehow doubt that many people would wait through a second line to scan their ballot and line it up with the verification screen - and I similarly doubt that sufficient scanners would be provided such that all voters could reasonably take advantage of verification. Although the system contemplates your being able to verify your ballot from your home PC, this creates additional issues of vote security (IP accounts can be tracked, and may identify individual voters), and given the prior emphasis on the subject it is not clear how accessible the Internet system would be. The second problem is that, although the data may be carefully encoded and safeguarded, in a close election it seems quite feasible that a subset of the data might "somehow" become corrupted and, once again, no recount or verification of the result would be possible.

The author correctly notes that neither of these systems would provide a paper trail, and thus that neither would satisfy any law requiring a paper trail. However, it's also fair to observe that it is perfectly reasonable to base legislative requirements on systems presently available, as opposed to systems which are still under development.

There's a simple technology which creates ballots where voters can easily see the candidates they voted for, which provides for electronic counting (and recounting) of votes, which allows for a paper trail, which is comparably reliable - optical scanning of paper ballots ("Scantron"). Visual magnification devices and an audio interface can be provided for the visually impaired. While implementing a "bleeding edge" voting system can involve a staggering amount of government money, all of which is wasted if the system fails, optical scanning technology is known to work and is comparatively cheap. Just as passenger airlines were content to let others work the bugs out of jet engines before switching over to jet aircraft, it is not at all unreasonable for votes to be cast using a tried and true technology while the bugs are worked out of the possible successor technologies.

1 comment: