The problem has been reported in U.S. elections since 2004 as states move to e-voting in an effort to make the vote-counting process more accurate. Instead, for many Americans, the change has led to more questions than answers, and suspicions that their votes aren't being counted correctly.
But is vote flipping a real problem, as e-voting critics argue? Or is it caused by user error, machine calibration issues or other factors, as e-voting advocates argue?
Stanford University computer science professor David L. Dill, who founded the nonprofit Verified Voting Foundation and VerifiedVoting.org, has been looking at vote flipping and yesterday called for investigations to determine if there is a real issue.
"People have been way too quick to diagnose the problem," Dill said. Some who have not examined the issue closely quickly call it a touchscreen calibration problem, while others point to different causes. "It could be a calibration problem with the touchscreens, but I'm not sure that anyone really knows yet because no one's looked at it. My answer as a computer scientist is that I want facts ... and all I've heard for two years is speculation."
Dill rejected one theory -- that the problem is a conspiracy to defraud voters of their votes and give the election to the opposition. Once a voter picks a candidate, a review screen shows who they voted for. That ability to review the vote before it is ultimately cast, he said, makes it less likely that fraud is involved.
"It seems to me if you were trying to commit fraud, you wouldn't show [the ballot] to the voter," he said.
One of the possible causes of vote flipping may be voters who place their hands on the side of a machine as they vote, perhaps accidentally touching it with their thumb and erroneously making a selection, he said. In other cases, some e-voting machines use a thumb-operated wheel to advance the electronic screens and when it is turned, it highlights a candidate in the next race on the ballot -- possibly giving a voter the impression that an erroneous choice has been made for them, he said.
The way to figure it out, he said, is to bring together a panel of experts to investigate the issue, confirm it, find ways to fix it and then get any fixes out to voting officials, Dill said. Until then, election officials and watchdog groups across the country will continue to hear reports of vote flipping, he said.
"We know it's going to be a major deal," Dill said. If a voter notices the problem on an e-voting machine's review screen, they can try to go back and fix it, he said. Sometimes it takes multiple attempts to correct, according to reports. But if they don't notice it or try to go backward to fix it, their votes are improperly cast, he said.
"This problem, I think, is a national disgrace," Dill said. "There needs to be a serious independent investigation of this problem ... across the country."
Ted Selker, the co-director of the Voting Technology Project being conducted by the California Institute of Technology and the MIT, has one possible explanation: Experiments show that voters incorrectly choose a candidate on their ballots one in 30 times, even under laboratory settings. "People are just sloppy and make mistakes," Selker said.
Voters may interpret the problems as an issue with the e-voting machines, Selker said. But much of the problem may simply be related to how voters actually use e-voting hardware. Touchscreens are designed for tapping, he said, while voters will often try to drag their finger across a selection, as though they are dragging a computer mouse.
"Vote flipping is a user interface problem," not a technical flaw that is allowing votes for a candidate to be miscast, he said. "I trust that that is not what's happening."
Machine calibration can be a problem because the units are sensitive to user finger inputs based on the user's height and other factors. Taller and shorter voters -- outside of the calibration settings -- can affect the areas where their fingers meet the touchscreens, causing selections that appear to be incorrect, he said.
Better machine designs and simpler voter selection processes, through improved and more intuitive designs, would help a lot, said Selker, who is also a professor of media arts and sciences at MIT and an expert on product design and human interfaces.
To voters, the prospect that their votes might have been switched to another candidate can come from e-voting interfaces that are not intuitive -- such as machines that deselect candidates if a voter touches a part of the screen by mistake. That could make the voter think the machine has removed their vote selections without reason.
Also he said, people can be nervous when they use the machines, which can also lead to user mistakes. And a lack of standardization across the U.S. means a variety of brands and models of e-voting machines are now in use. States and local communities often have their own specific requirements for the hardware, complicating design and support issues for vendors and local governments, he said.
In Brazil, he said, 106,000 of the same e-voting machines are used, simplifying training and use of the machines across the country.
At least one other e-voting expert, Avi Rubin, a professor of computer science at Johns Hopkins University in Baltimore and director of Accurate, an election research center, said the only way to know for sure if votes are being placed for one candidate and given by machines to another is by having paper records of every ballot cast.
"My big worry is that we cannot ever say conclusively whether or not [vote-flipping] happens" due to software glitches, tainted code, machine rigging or other tampering, if there is no paper record, Rubin said. "Whether it happens or not, we won't know it happens."
By having paper records, actual vote totals can be later checked against the vote tallies in e-voting machines to ensure the integrity of an election, he said. "Recounts are then possible," he said. "Without them, recounts are impossible."
Accurate is now studying related election issues -- including how to design voting systems where such vote-flipping possibilities can't occur -- using a US$7.5 million grant from the National Science Foundation, Rubin said. Dill is also involved in the study, as well as professors at several other universities.
Dan Wallach, an associate professor of computer science at Rice University in Houston, Texas, and an associate director of Accurate, said vote-flipping has been a recurrent issue in elections in recent years, but isn't grounded in facts. He believes that machine calibration issues play a large role in the problems most often reported by voters.
"The fundamental problem with any touchscreen is that calibration matters, that angle of touch matters," Wallach said. Different people even use different parts of their fingers to touch the screens, he said.
"If it is calibrated by someone doing it one way and is used by someone else doing it another way, there will be differences," Wallach said. The only way to fix the problem, he said, is to make the candidate selection buttons on the screen much larger and to separate them farther apart, which is how banking automated teller machines are often set up. "If you miss one button, you're not likely to touch another one," he said.
Another way to solve the problem, he said, is to drop touchscreens in favor of other vote input systems. One e-voting machine vendor, Hart InterCivic Inc., uses a wheel on their machines that voters rotate to make their candidate choices, and a separate button to confirm the choice, he said. Another option would be to build systems that incorporate actual buttons for voters to use to make their choices, he said.
Michelle Shafer, a spokeswoman at e-voting machine vendor Sequoia Voting Systems, said the perceived problem of vote flipping is definitely human error or procedural error and not a problem with the systems themselves.
"Sometimes the machines need to be calibrated or recalibrated," Shafer said. "Machines in and of themselves do not flip votes."
Since the hardware displays voters' choices before their electronic ballots are cast, voters can spot any inadvertent errors and should immediately contact a poll worker if they believe an incorrect selection has been made, she said.
What appears to be a flipped vote, she said, is more likely a situation where a voter has too many fingers touching the machine, has long fingernails or is leaning on the equipment. "There's many, many reasons that this type of thing occurs [accidentally]," she said.
E-voting machines are far more secure, accurate and more auditable than the old mechanical lever-operated voting machines and other systems they replaced, Shafer said. She called vote-flipping concerns a "conspiracy theory from activists and bloggers."
David Bear, a spokesman at e-voting equipment vendor Diebold Election Systems Inc., also disagreed that vote flipping is happening with the machines. "It's not a problem," Bear said. "It doesn't exist. This again falls into the 'what if' scenario."
Voters do make mistakes as they use the machines, he said, just as they used to make mistakes on old-style paper ballots. With e-voting hardware, they are at least able to see their mistakes on review screens or when their optical-scan ballots are kicked back due to a user error, he said.
"I don't think these what if scenarios are just related to e-voting machines," because they could have also occurred using paper ballots in the past, Bear said. "I would argue that the technology has lessened these scenarios."
Did you experience vote flipping or other problems related to e-voting in Tuesday's elections? If so tell us about your experiences by sending an e-mail to editor@computerworld.com.