imagine that!

Computer Scientists Cautious of E-Voting
3 November 2004, Rachel Konrad, AP
Sarasota Herald-Tribune

MIAMI - After only scattered problems in electronic voting's biggest day ever in the United States, with roughly 40 million people casting digital ballots, voting equipment company executives crowed.

To them, the relatively smooth election was a vindication of paperless touch-screen systems.

For more than a year, computer scientists and voting rights advocates had vigorously assailed the nation's 175,000 touch-screen machines as insecure and unreliable, prone to software bugs, hackers and hardware failures.

Some naysayers had even predicted worst-case scenarios in which the ATM-like computers deleted or altered votes, machines overheated and crashed under record turnout. But that's not to say electronic voting was trouble-free.

On Tuesday, poll workers in New Orleans had numerous problems operating the equipment. On Election Day and during early voting, several dozen voters in six states reported difficulty selecting candidates, apparently due to miscalibration.

Tuesday's vote was not marred, however, by the problems that plagued primaries earlier this year - power outages, missing memory cartridges, machines that displayed the wrong ballots and suspicious delays in reporting results.

"It was a very positive day for the American voting system generally and for electronic voting machines particularly," said Harris Miller, president of the industry trade group Information Technology Association of America, which represents voting equipment companies. "The machines performed beautifully ... Instead of theories about catastrophes, the simple reality is that the machines produce accurate results and the voters love them."

Computer scientists reserved judgment.

Many acknowledged that the hardware performed well. But software errors may have changed results, they said. The vast majority of touch screens in the United States do not produce paper records. And that means, critics say, that the machines could alter or delete ballots without anyone noticing.

"What has most concerned scientists are problems that are not observable, so the fact that no major problems were observed says nothing about the system," said David Jefferson, a computer scientist at Lawrence Livermore National Laboratory in California. "The fact that we had a relatively smooth election yesterday does not change at all the vulnerability these systems have to fraud or bugs."

Avi Rubin, one of the nation's leading critics of e-voting, said he was relieved and encouraged that the machines didn't fail en masse on Election Day.

But Rubin, who worked in Maryland as a poll judge Tuesday, still supports major changes in election technology - including requirements that the machines produce paper records, and that independent researchers be permitted to examine their software for problems.

"I've been saying all along that my biggest fear is that someone would program a machine to give a wrong answer," said Rubin, a Johns Hopkins computer scientist. "If that were to happen, the machine would still work fine - we just wouldn't know it."

Statisticians at the Massachusetts Institute of Technology and California Institute of Technology are asking county election officials throughout the nation for raw election data and hope to perform "forensic tests" that could take at least a month. The fledgling U.S. Election Assistance Commission is also compiling data and plans to issue a report later this month. Research will include comparisons of the number of voters and the number of ballots cast in random precincts, an attempt to determine whether votes were mysteriously lost.

According to an MIT/CalTech study, 8.2 percent of touch-screen votes in senatorial elections between 1998 and 2000 were lost - more than any other system except lever machines, which lost 9.5 percent of votes.

Other critics said comfortable, sometimes predictable margins of victory in states with electronic voting - Bush in Georgia and Florida, Kerry in California and Maryland - will minimize scrutiny of touch-screen results.

A razor-thin outcome could have prompted a recount, but it would have likely been challenged in court because votes cast on touch screens - everywhere but in Nevada - cannot be manually recounted owing to the lack of a paper trail.

Ohio, where votes were still being counted Wednesday, does not rely heavily on electronic systems.

"We've resolved in our heads to provide scrutiny only when the election is close, and that's a bad way of approaching it," said Lida Rodriguez-Taseff, chairwoman of the Miami-Dade Election Reform Coalition. "We need to apply scrutiny every time so we know we have a healthy process." More than half of Florida voters cast electronic ballots.

David Bear, spokesman for Diebold Inc., which has about 45,000 machines installed nationwide, said more counties will switch to touch screens - particularly for early voting.

Because they can toggle between ballots in dozens of precincts, election officials can consolidate polling places for early voting, reducing lines on Election Day.

The machines also toggle between languages and can be equipped with headphones for blind voters.

Bear dismissed the notion that comfortable margins obscured problems.

"There was no dodging of a bullet," Bear said. "The fact of the matter is, electronic voting is a better way of voting because touch screens are more accurate and they meet people's special needs."