When Voting Machines Misbehave
Nov 7, 2016 2:25 PM EST
By Faye Flam
Interference by hackers is just one of the nightmare scenarios that worry computer scientists about the upcoming election. The other is a race so close that calling the result is beyond the capacity of today’s voting technology.
Experts who’ve delved into the accuracy of these apparatuses -- from punch cards and mechanical levers to electronic voting machines -- say that no system is perfect. In most cases the error rates are unknown, or are only measured in artificial test settings and not as they would be used in the real world.
Computer scientist Douglas Jones of the University of Iowa, who co-authored the book “Broken Ballots: Will Your Vote Count?,” came to realize that voters usually blame themselves when something goes wrong in the voting booth -- a tendency that could mask intentional hacking or equipment error. When Jones set up experiments with electronic voting machines rigged to switch votes away from the subjects’ choices, the people casting ballots assumed they had done something wrong. “People tend to trust the machines,” he said -- even when the machines don’t work.
Electronic voting machines use proprietary software, making it hard for outside researchers to get a measure of their error rates, according to computer scientist Rebecca Mercuri, founder of the company Notable Software and an expert on electronic voting systems. ...
Voting has surely advanced since the 19th century, ... Today’s scanners and voting machines ensure privacy and make it possible to count millions of votes. But there was a surge of concern about their reliability after the 2000 presidential race, ...
It all came down to Florida, which was using a common technology of the time, the Votomatic -- a system of punch cards read by machine. Until 2000, people thought the punch-card system was high-tech, said Jones. It was also cheap. But a partial recount in the too-close-to-call race revealed flaws. Some votes hadn’t been counted because punches were incomplete -- hence, hanging and pregnant chads.
Shortly after the Supreme Court decided the race, Jones got access to a Votomatic and showed that if it was not properly maintained, the chads tended to accumulate, preventing the mechanism from properly punching the cards. “It leads to a very nice pregnant chad,” Jones said.
The chad problem was what scientists would refer to as random error -- noise in the system that limits its capacity to get an exact count. Critics also pointed to what they saw as systematic errors -- problems in the way overseas and absentee ballots were counted that might have skewed things in one direction.
Over the next decade, many states replaced punch-cards with electronic voting machines. The problem there, Jones said, is that “they replaced a system with flaws with one whose weaknesses are unknown.”
Manufacturers sometimes subject sample machines to testing, Mercuri said. But that’s like sampling just a few cars of each make for safety and emissions inspections rather than requiring all cars go through it. ...
And while close calls are rare, they happen. Mercuri testified as an expert witness in a 2008 case in which two candidates for office in Orange County, California -- both named Nguyen -- finished only seven votes apart. ... A full hand recount would have been the best way to decide the race, but it was never done.
On the bright side, Mercuri and Jones agree it would be difficult for anyone overseas to hack into this election, since voting machines aren’t connected to the internet. However, scientists have shown how easy it is to hack into a voting machine from the inside.
Jones warns that if overseas hackers got access to online voter registration lists, they could disrupt the election by changing addresses so people would be turned away -- told they are not registered or are in the wrong precinct. The best hackers, he said, wouldn’t leave any evidence behind. He warns voters to be prepared, and not to automatically assume they made a mistake if something goes wrong.
This column does not necessarily reflect the opinion of the editorial board or Bloomberg LP and its owners.