Are design issues to blame for e-vote 'flipping'?

Are some touch-screen voting machines really "flipping" votes from one candidate to another, or are the voters who claim their votes are being changed just wrong?

With the US elections just days away, some voters in states including West Virginia, Texas and Tennessee have reported that electronic touch-screen voting machines are "flipping" their votes to another candidate on the screen.

When those allegations are made, e-voting hardware vendors and local election officials usually blame errant fingers, overhanging jewelry or clothing -- or they argue that the touch-screens weren't properly calibrated. But continuing reports about flipping -- sporadic though they may be -- raise questions about the machines themselves, when they were designed and what kind of usability testing was done.

Computerworld asked the four major e-voting machine vendors to talk about how their hardware was originally designed, with an emphasis on whether real-world user testing was done as the devices were being drawn up.

Vendors defend their designs and say they have been proven through sales figures in the marketplace; e-voting critics note the vote-flipping accounts by voters and worry that the machines could cast doubt on the legitimacy of elections.

Chris Riggall, a spokesman for Premier Election Solutions (formerly Diebold) in Texas, said that vote-flipping reports are taken seriously. "It's one of those things that when a voter says this occurs, well, I wasn't there, you weren't there, but we can't say that they didn't have that experience," he said. "The key is that poll workers immediately need to be asked for assistance if that happens" so the ballot can be corrected or cancelled if needed.

A touch-screen machine can be checked for proper calibration to make sure the spot on the screen that's touched registers the proper vote on the electronic ballot -- or a voter can be redirected to another machine in the polling place, Riggall said.

Despite a smattering of reports elsewhere, there have been no complaints of vote flipping in Georgia, where almost a million people have already cast early ballots using Premier touch-screen machines. "There has not been one claimed occurrence of what is really a calibration issue in all of those votes that have already been cast here," he said. "The nature of a touch-screen device [is that], like any other device, there can be an instance where it's not operating properly. What the majority of our jurisdictions find is that where 'vote flipping'..or mis-selection is occurring .., in the vast majority of cases they find that the calibration is not out of specification."

Instead, he said, a voter's "fingernails can definitely have an impact, the length of a fingernail." If a voter has long nails, "there's no question that you can touch the adjoining choice" and make the machine appear to change a selected vote. "There may be instances where that is occurring."

Page Break

Riggall said he isn't familiar with the design history behind the machines, but noted that new models aren't being designed now because optical scanning systems that use paper ballots are growing in popularity. More states are moving to optically scanned paper ballots because they offer a paper trail of every vote cast in case a recount is needed.

Since new touch-screen machines aren't being designed, no usability studies are being done to improve existing designs, he said.

Ken Fields, a spokesman for Election Systems & Software said in an e-mail reply that "certainly, the equipment is designed with voters in mind. I can assure you that in the national level certification process that's taken place before the equipment is ever used, there's great attention paid to the manner in which the technology is going to be used," he said.

ES&S' iVotronic touch-screen machines are designed to prominently display a voter's candidate choices so that they can be reviewed before the ballot is cast, he said. "If the iVotronic has captured what the voter did not intend to convey, but what the voter may have accidentally or inadvertently selected, then the voter is able to see the selection and easily change it if necessary. Every voter has multiple opportunities to see and validate the captured selection prior to casting a ballot."

Peter Lichtenheld, director of operations for vendor Hart InterCivic in Texas, said his company doesn't build or sell touch-screen machines because it felt that existing technologies weren't "up to snuff" when they were created. That could change, he said, because new touch-screen technologies like those used in Apple's iPod music players and iPhone, have shown maturity.

Hart's eSlate voting machines rely on a mechanical wheel that's used to select voting categories, and a button that makes a voter's selection.

When voters are given a chance to try various voting machines in government open houses, he said, the voters often prefer touch-screen devices because they are simple and have a familiar interface. The problem, he said, is that existing technologies are not as reliable as mechanical systems like those used in the eSlate.

"Have you ever used a signature pad" on an electronic check-out system in a retail store, he asked. "It looks nothing like your real signature 99 percent of the time. It looks like crap. That's why we don't use touch-screens."

Page Break

And though it didn't show up in the design stages of the eSlate, a usability issue did once turn up in California, he said. "In one county, sometimes people were relaxing and leaning on the 'cast ballot' button and casting their ballot unintentionally."

To fix that, election officials increased voter education efforts, reminding them not to lean on the machines while casting their votes, he said.

A spokesman for Sequoia Voting Systems could not be reached for comment.

David Beirne, the executive director of the Election Technology Council, a trade group that represents e-voting vendors, said that "most if not all of them did some sort of usability testing before production."

But one of the requirements for the touch-screen machines -- high-sensitivity to touch so they can be used by physically-impaired or disabled voters who want to vote on their own -- could cause some of the problems experienced by able-bodied voters, Beirne said. "Sensitivity is tied to the [voting] actions that are structured for voters with disabilities, so if you have loose clothing or jewelry, it can make [an errant] selection," Beirne said. "They are designed to be sensitive."

When such incidents are brought to the attention of election officials, they usually try to replicate it so they can determine what caused it, he said. "Critics call it 'vote-flipping.' That term is used by e-voting critics to mean that the machine is trying to work independently of a voter's wishes" in a manipulative way. Instead, such problems usually go back to the local procedures that are used to verify proper machine touch-screen calibrations throughout election day.

"That's why there are procedures to follow in response to it if reports come in" about problems, Beirne said.

The Election Technology Council Wednesday posted on its Web site a list of tips and reminders for voters about using the machines.

Representatives of several e-voting watchdog groups, however, vigorously dispute the vendors on these issues. Lillie Coney, coordinator of the National Committee for Voting Integrity, a project of the Washington-based Electronic Privacy Information Center (EPIC), said the main problem is that existing standards for e-voting machine designs have been voluntary, yielding designs that don't work well for all voters.

Page Break

"You have usability testing for all kinds of [consumer] devices," Coney said. "The people who sell [consumer goods] want to make sure that people can use them and use them easily...so they don't buy a competing product."

With e-voting machines, that's not the case, she said. "There's a disconnect between the people who make the voting systems and the voters themselves" during the initial purchasing decisions by state and local government agencies, she said. "There were no standards in place. There was no usability testing required [although vendors] may have done some on their own."

Elections officials didn't seek such testing because they had limited budgets, wanted equipment that they could afford and had little time to dig into issues like usability, Coney said. "The issue for them was what fits the budget," she said. "A lot of the usability issues were just 'throw the voters in there and let them be the beta testers."

What has happened, she said, is that the vendors have simply blamed voters and poll workers for problems that arise. "That's been the generic scapegoat...from the beginning of these machines being used."

The sensitivity issue is a prime problem, she said, because of the disparity between the needs of physically-impaired voters and others. "The way usability works, you don't make the voter conform to the machines. You build the machines to work the way the user uses it..., to fit whatever a typical voter might do in that environment."

"The existing systems are made hyper-sensitive for the few voters with extreme physical limitations, so then people without physical limitations get all these flaky responses from the machines," Coney said.

What is surprising about touch-screen problems, Coney said, is that the technology has been seemingly proven in other industries. "They should know how to do this," she said. "This is a technology interface that you find at kiosks, you find at banks, that you find all over the place, and they're not messing up. So why is this happening?"

What's required, she said, is mandatory usability testing for e-voting machines. "It's the only way you're going to know" what the problems might be. To ensure that that happens, she said, federal funding should be allocated. "Otherwise, it's not going to happen," she said. "And until we do serious research-and-development, you know, go-to-the-moon-level R&D...it won't improve," she said. "Nobody's ever tried to serve the whole diverse population of voters" with such machines. "This is a neat challenge in and of itself."

Ellen Theisen, the founder of VotersUnite.org, an e-voting watchdog group, said that vendors only want to talk about calibration when there are clear design flaws. One problem has been that when machines are tested by independent groups, they only test one or two machines, which is inadequate, she said.

"I'm not convinced that there aren't serious design flaws in the software, too," Theisen said. "That is not checked, either."

Vote flipping has also occurred in eSlate machines, she said, which use software but not touch-screens. Whatever the design, "when we see [vote-flipping] happen over and over and over, there's some problem," she said. "If it's a poor user interface, there's a problem."