February 17, 2004
Would you like a receipt with that election? Technologists advocate a paper trail for electronic voting machines
By Stephanie Chasteen
Electronic voting has its perils. Imagine this odd scenario on election day. You step inside the voting booth at the local Y and are faced with a red curtain. Behind the curtain is a man who fills out your ballot for you as you tell him whom you want for president, for city council, for mayor. But what if the man writes it down wrong, or switches your vote to a different candidate? Or what if his pen breaks, or he loses your ballot? You would never know, as you don't get to see your ballot and there's no proof of your original vote.
Stanford computer science Professor David Dill says that the man behind the curtain should show you the ballot. He uses this metaphor to illustrate his grievance with completely paperless electronic voting machines, such as touch-screen machines.
"If the machine silently loses or changes the vote, the voter has no clue that that has happened," says Dill. He argues that electronic voting machines should print a paper copy of the ballot, which the voter can inspect and which can be used in the event of a recount. Dill will make the case for this "voter-verifiable paper audit trail" in a Feb. 15 symposium on voting technology at the annual meeting of the American Association for the Advancement of Science (AAAS).
While many, including Dill, think that optically scanned ballots are the cheapest and most reliable method of collecting votes, there has been a rush to invest in electronic voting machines instead. States must decide quickly how to use a rare gift of matching federal funds (allocated in the aftermath of the 2000 election debacle) to upgrade outdated equipment. Touch-screen machines are an attractive option because they're easy to use and handicapped accessible, can be programmed in multiple languages, and allow for quick tabulation of election results. No hanging chads. But most electronic voting machines are paperless, with only an on-screen display. Even an ATM machine gives a receipt, and for obvious reasons. Even in a morally perfect world, technology can fail. Machines can make mistakes.
Numerous problems have been cited in states using electronic machines, such as a vote being recorded for the wrong candidate. And without a paper record of the ballot, it's not always clear exactly what went wrong. In January, for example, touch-screen machines in Florida's Broward County reported 137 blank ballots in a race for a state House seat. That race was won by just 12 votes. It's impossible to go back and verify whether those ballots were intentionally left blank or fell victim to faulty (or malicious) code. There's nothing to recount.
Dill is an expert on eliminating bugs from computer systems. In 20 years of research, he has developed a keen appreciation for just how hard it is to create a virtually flawless computer system. "That's an extremely tall order," he says. "If you have computers recording votes or counting votes, then you have to do manual recounts with sufficient frequency that machine errors are likely to be caught."
Dill also is adamant that experts must study these electronic systems, which are understood by very few. But scientists aren't allowed access to the trademarked software code of the large companies like Diebold Election Systems and Sequoia Voting Systems. A much-publicized report by a team from Johns Hopkins and Rice universities, however, outlined troubling flaws in some leaked source code from Diebold. Several smaller companies promise that their systems are more secure. Several also plan to release the source code for their programs to allow independent reviewers to check for bugs and identify potential avenues for malicious attacks. Hackers, however, would have access to that same code. Andy Neff, the senior scientist for one such company, VoteHere, will be presenting information about VoteHere's system at the same AAAS symposium.
A score of legislators and computer scientists join Dill in his crusade for a voter-verifiable paper trail. He advocates for federal legislation mandating an auditable paper trail on all electronic voting machines. Nevada and California passed similar legislation in the past year, requiring paper trails by 2004 and 2006, respectively. More than 1,600 technologists and 53 elected officials have signed his Resolution on Electronic Voting (www.verifiedvoting.org).
But others have serious concerns about a mandated paper trail. Gadget guru Ted Selker of the Massachusetts Institute of Technology is an expert in human-computer interaction and the co-director of the Caltech-MIT Voting Technology Project. Selker, who will be speaking at the AAAS symposium, likes the idea of having a verifiable audit trail, but not made out of paper. He suggests an electronic voice that reads voters' choices back to them over a headset for verification. That audio track could be recorded, providing an independent record of the vote. That digital recording offers a more secure record than paper, says Selker, to be used in the event of a recount.
Selker also says that pressing problems are being ignored in the wake of the push for a voter-verifiable audit trail. "Nobody's talking about ballot design, we're so embroiled in talking about voter-verified paper trails," he says. The MIT-Caltech Voting Technology Project found that at least 6 million votes were lost in 2000 to confusing ballots, problems with voter registration and polling place violations.
The security of absentee ballots, trusted to the sorting boxes of postal workers, is also an unsolved problem of no small consequence. Charles Stewart, a political science professor at MIT who also will be speaking at the AAAS symposium, says that nowadays most ballot fraud is committed via absentee ballots. Even if paper trails completely solve the problem of computer bugs and hackers, disgruntled or poorly paid postal workers could prove a greater threat to the integrity of election results. Last year more than 29 percent of voters in California and 75 percent of voters in Washington state cast their votes by absentee ballot.
Stephanie Chasteen is a freelancer and doctoral student in physics at the University of California-Santa Cruz.
This release was written by freelancer Stephanie Chasteen. Dill will participate in the symposium "Voting Technology: Current Assessment and Future Prospects" at the annual meeting of the American Association for the Advancement of Science in Seattle on Sunday, Feb. 15, from 9 a.m. to noon PST. A photo of Dill is available on the web at http://newsphotos.stanford.edu (slug: "AAAS").