The Ghost in the Voting Machine

Last week’s close Undergraduate Council election ended with a bang: The Election Commission voted to decertify the results, prompting half of the EC to resign in protest, and the UC President’s email account sent a message accusing Vice Presidential candidate Eric Hysen of fraud—signed by Vice President Kia McLeod—but no one admits to drafting it.

The debate centers on the computer system that tallied the votes, and whether anyone tampered with election results. Electronic voting systems have their advantages in cost, accessibility, and, precision. But last week’s events illustrate that machines are more likely to sink an election than to save it, in part because electronic voting often generates the perception of unfairness, even if everything is played by the book.

Before computers, voters simply had to accept imprecision. In ancient Athens, votes were decided by a show of hands, counted informally. For the framers of the United States Constitution, the Electoral College meant that your vote didn’t really matter that much anyway. In theory, electronic voting systems are a huge improvement because they can count with far greater precision than humans. An computerized system should be able to verify the voter’s identity, count the votes in real time, and yet still leave no record that would allow a person to sell his or her vote.

But even the most sophisticated voting machine can only making voting more precise, not more just. They can’t force certain demographics to turn out to vote in higher numbers or undo the name-recognition advantage incumbents enjoy. Nothing can stop a candidate from changing his last name to “Aardvark” so that he’ll be listed higher on a crowded primary ballot. While there may be no technological solution to combat many of these concerns, sophisticated statistical methods could resolve some of these issues. For instance, computers could randomize the order of the candidates’ names to get around the precedence advantage.

Yet modern voting systems don’t rely on statistics to smooth out voting irregularities, in large part because statistics make people uncomfortable. Statistical methodology throws into doubt the “one man, one vote” assumption and leaves the door wide open to allegations of fraud. For example, statistical sampling was almost banned in the 2000 census, and in debates over the 2010 census this summer, Republicans were vocal in their critique of statistical methods. In the words of Rep. Paul Broun, there is a real fear that “every voter will not be counted and maybe some voters might be counted more often than others.” One could only imagine the outcry if these techniques were applied to a national election.

Often, changes that make an election technically fairer paradoxically give it the appearance of somehow being rigged. Mathematical wizardry can look like real wizardry to the uninitiated. Therefore, scientific assessment of fairness matters much less than the public’s perception of it.

One reason why the UC election proved so contentious was because the vote-counting system itself altered perceptions of the race. The UC election uses the Hare-Clark single transferable vote system, which means votes for losing candidates get redistributed to the top tickets. The system is more sophisticated (and ultimately fairer) than the simple plurality vote we use for the American presidency. But it also means that the vote counting takes place in multiple rounds where tallies rise and fall and early advantages often disappear. The Hayward campaign was winning until the third-place votes were redistributed, and this may have contributed to a sense that victory was taken away from them.

Election officials were also confused about how the software worked, and the email bearing the VP’s signature implied that Hysen’s ability to access the database meant he could also change the results unnoticed. Although there appears to be no evidence of tampering, the suspicion will be tough to dispel.

Yet, just because no fraud seemed to occurr doesn’t make it impossible. No networked system can be fully secure, and modern computers are running so much code written by so many people that a backdoor could exist at any level. Even without malicious attacks, tabulation systems that employ sophisticated verification mechanisms can behave in mysterious ways, which compounds the general problem in computing that no one can fully trust code they did not author.

The failings of electronic voting in the U.S. are well documented: lost absentee ballots, servers freezing in Ohio, and deleted votes in the 2008 election. The scariest part is that with no permanent record of voting input due to secret balloting, we may never know if an election has been compromised.

If electronic voting not only generates undue suspicion, as in the UC case, but is actually more susceptible to fraud as well, then there is ample reason to remove the machines where the election really matters. It may be more expensive and more difficult, but, to paraphrase the Long-Johnson campaign, at least we can get behind it.

Adam R. Gold ’11, a Crimson editorial editor, is a physics concentrator in Adams House. His column appears on alternate Mondays.