Proofiness - Charles Seife [98]
How can engineers assure that the machines are trustworthy— that they’re really counting our votes properly? It’s a damnably hard problem. When you want to ensure that a transaction is recorded properly, you typically get a receipt; when you deposit a large amount of cash at a bank, for example, the ATM gives you a slip of paper that is proof the transaction took place. That way, if the bank fails to record the transaction, you can prove that they made an error. Unfortunately, you can’t give receipts in a vote. If you did, it would open the door to some real nastiness: “Hey, buddy—if you don’t give me a receipt proving that you voted for Gordon Brown in tomorrow’s election, I’ll break your kneecaps.” The moment you give voters any form of proof that they voted for somebody, you create an environment that’s ripe for voter intimidation or for vote selling. So you can’t let us voters get our grubby paws on any sort of receipt.
Another possibility is that the machine prints out a receipt and lets the voter see it, but the voter can’t touch it; instead, the receipt gets stored in the machine. This is a “voter-verified paper trail,” and it has some advantages over purely electronic voting. However, it also has some serious disadvantages. For one thing, it’s more costly, because you have to buy paper and ink to feed the machines. For another, paper jams and misfeeds will make the machines more likely to break down than a purely electronic machine. But these are minor issues. More serious is that the paper trails don’t really solve the fundamental problem. Say that an election official notices that the machine counts are slightly different from what the paper trail shows. Do you believe the machine count or the paper trail? It’s not at all clear. Sure, there might be a bug in the machine that created the discrepancy, but similarly, it might be that a few of the receipts got tangled up in the machinery or were lost by election workers or were simply miscounted (and as we saw, this often happens with paper ballots). Which is the source of the error? More often than not, machines are more reliable than humans, so it’s far from certain that you should overrule the machine count with a human count of the receipts.
Voter-verified paper trails would increase the trustworthiness of electronic systems, but I think that a better solution would be to make the machines open-source. If the software is open to public scrutiny—if any programmer can search through the code looking for bugs—then we’d have a high degree of confidence that the machine is counting votes reliably. The same thing goes for the machine’s hardware. So long as voting machines are proprietary, running buggy software that nobody can inspect, there is very good reason to be wary of trusting your votes to an electronic kiosk.
That being said, there has only been one reported incident of electoral fraud using electronic machines. In a 2007 Kentucky election, a judge, a court clerk, an elections official, and a bunch of other defendants allegedly schemed to fix local elections through a variety of means, including tampering with electronic voting machines. According to the indictment, the conspirators misled voters about how to use the machines. They told voters that their votes had been recorded when they pushed a button labeled “Vote” on the machine. In fact, there was an extra step—one that allowed the voter to review his vote and change it if desired. The vote wasn’t officially cast until the voter pushed another button marked “Cast Ballot.” So when voters walked out of the booth after having pressed “Vote,” the conspirators allegedly waltzed in, changed the votes, and then cast the ballot once the votes had been “corrected.” This scheme was low-tech. It exploited voter confusion rather than any inherent bug in the machine—and it shows that the threat can often come from a quarter you don’t expect.
Electronic voting does indeed pose