Can This Machine Be Trusted?

By Michael Duffy
Time Magazine

Sunday 29 October 2006

The US's new voting systems are only as good as the people who program and use them. Which is why next week could be interesting.

The Diebold Accuvote TSX electronic voting machine. (Photo: Brent Humphreys / Redux for Time)

A woman walked into a polling place in Peoria, Ill. last week and proceeded to use one of the new electronic voting machines set up for early voting. She logged on, went through each contest and seemed to be making her choices. After reviewing each race, the machine checked to see if she was satisfied with her selections and wanted to move on. Each time, she pressed YES, and the machine progressed to the next race. When she was done, a waving American flag appeared on the screen, indicating that her votes had been cast and recorded.

But there was a problem. The woman had not made any choices at all. She had only browsed. Now when she told the election judges she was ready to do it again-but this time actually vote-they told her it was too late. Pressing the last button, they said, is like dropping your ballot in an old-fashioned ballot box. There's no getting it back.

So What?

So this: In one week, more than 80 million Americans will go to the polls, and a record number of them - 90% - will either cast their vote on a computer or have it tabulated that way. When that many people collide with that many high-tech devices, there are going to be problems. Some will be machine malfunctions. Some could come from sabotage by poll workers or voters themselves. But in a venture this large, trouble is most likely to come from just plain human error, a fact often overlooked in an environment as charged and conspiratorial as America is in today. Four years after Congress passed a law requiring every state to vote by a method more reliable than the punch-card system that paralyzed Florida and the nation in 2000, the 2006 election is shaping up into a contest not just between Democrats and Republicans but also between people who believe in technology and those who fear machines cannot be trusted to count votes in a closely divided democracy.

Perhaps the biggest fallacy in this debate is the notion that elections were perfect before Congress decided to hold them on computers. They weren't. "Stuffing the ballot box" is not an expression from the world of fiction. The problem with overvoting punch cards existed for decades before the dateline PALM BEACH COUNTY became a household term. Peoria County clerk JoAnn Thomas says she routinely tossed out several hundred twice-punched ballots every year. That represents roughly 1% of all registered voters in her jurisdiction.

The 2000 election reminded Americans that every vote makes a difference, and scrutiny of polling practices intensified. So just as America has moved to a process of electronic voting and tabulation intended to make voting more accessible, reliable and secure, trust in the system has actually gone down. Says David Orr, clerk of Illinois' Cook County: "We used to have a problem with giving people the wrong ballots. And if we were lucky, we'd catch it before they voted. Now, if the same thing happens with a touch screen, it's a conspiracy."

So far, at least, Murphy's Law has been a bigger problem than fraud. Many jurisdictions, especially those with long or bilingual ballots, have struggled to program their computers perfectly, and there have been scattered reports of glitches. In three Virginia cities, for example, electronic voting machines have inadvertently shortened the name of the Democratic candidate in one of the tightest Senate races in the nation. In Charlottesville, Falls Church and Alexandria, James H. Webb's name will appear on the ballot summary screen page simply as "James H. 'Jim'" - with no last name. Sounds like a crisis - except that the same thing happened in the June primary and Webb still won.

A bigger worry concerns something that is least likely to happen-that someone will somehow meddle with the devices and manipulate vote tallies. It's not impossible. Princeton computer scientist Edward Felten and a couple of graduate students this past summer tested the defenses of a voting machine made by Diebold, a major manufacturer of such devices. Felten's team found three ways to insert into the machine rogue programs that allowed them to redistribute votes that had already been cast. In one instance, the testers had to take the machine apart with a screwdriver - an act likely to draw the attention of poll workers. But in two others, they were able to quickly infect the device with a standard memory-access card in which they had installed a preprogrammed chip. Other computer scientists have also breached electronic voting machines. Congressman Vernon Ehlers, a Michigan Republican who has been holding hearings this fall, says manufacturers "have produced machines that are very vulnerable, not very reliable and I suspect fairly easy to hack."

Concerns about fraud are heightened by the fact that with some electronic voting machines, there is no such thing as a real recount. When asked again for the tally, the computer could spit back the same response as the first time. For that reason, at least 27 states have built in a backup that requires electronic voting machines to provide an attached voter-verified paper trail - a running ticker that allows voters to see on paper that their votes are recorded as cast. That way, if there's a question about the electronic tally, the paper records can be counted by hand.

It was just such a paper trail that enabled Marilyn Jo Drake, the auditor in Iowa's Pottawattamie County, to suss out an anomaly in a county-recorder race she was monitoring in June. She noticed that a 20-year incumbent was being beaten 10 to 1 by an unknown newcomer. Sensing a glitch, Drake cross-checked the electronic results against the totals on the paper vote and discovered the veteran was actually well ahead. The problem, it turned out, was the way the candidates' names had been ordered and coded into the access cards that activated the machines, which were made by Omaha's ES & S. Drake says she should have caught the problem in the pre-election test runs. "It was human error both on their end and my end," she notes. Not every county will have an auditor as sharp-eyed as Drake - or an outcome as transparently false as the one she uncovered. "We were just plain lucky," she says.

Rather than waiting for results to be contested, some states are requiring election officials to conduct random samples of electronic results next week and compare them with the paper printouts. Minnesota's secretary of state, Mary Kiffmeyer, plans to audit the tally from two precincts in each of her state's 87 counties to make sure the electronic tabulation matches the paper trail. Audits, says Kiffmeyer, "just build confidence." In Los Angeles County, officials aren't waiting for the election to start running their tests. They will soon conduct random audits of 5% of the devices used in early voting, which began in earnest last week.

County election officials who spoke to TIME reported that most of the fears they field about the new machines come from Democrats, who have not won a national election in three cycles. It may be that a solid Democratic win in 2006 will allay some of their worries. It follows, of course, that if the Republicans lose, they will take up the charge. In fact, that's already happening in some places this year.

In a country of 300 million, it is far preferable for partisans, poll workers, defensive voting-machine manufacturers and voters to adjust to the new technologies, eliminate their weak spots and work to keep human errors to a minimum. In that way, voting by machine may someday be no more mysterious than making a visit to the ATM.

--------

With reporting by Hilary Hylton/Austin, Laura Locke/San Francisco, Jackson Baker/Memphis, Christopher Maag/ Cleveland, Melissa August, Mark Thompson/Washington.

-------