In a recent article on TechnologyReview.com, Simson Garfinkel gleefully reported that the concerns about electronic voting systems expressed by many of the country’s leading technical experts were overblown (see “Campaigning for Computerized Voting”). Suggesting that those who worry about the conversion from paper ballots and lever machines to newer technologies are like “a group of doctors arguing for the return of leeches because the President of the United States is too important to be treated by modern medicine,” Garfinkel concludes that electronic voting machines may “offer the best hope for escaping the mess inflicted by paper-based balloting system.”Back in the real world, however, the evidence is mounting daily that a lot more work needs to be done before the vote counting process-truly the kernel of democracy-is turned over to devices that lack adequate auditing and operate in secret. One recent study conducted by Johns Hopkins University and Rice University found that the high-tech voting machines made by Diebold Election Systems allowed voters and poll workers to cast extra votes, and also that cryptographic keys, the basic element of system security, were not properly managed. The governor of Maryland has called for an investigation to determine whether the state’s $54 million purchase of these so-called direct recording electronic (or DRE) systems was a wise move.
Another report finds that during San Luis Obispo County’s March 2003 primary in California, absentee vote tallies were sent to an Internet site operated by Diebold several hours before the poll closed. According to election law, officials may not release tallies until voting is completed. An MIT-Caltech study found that regular test forms, which allow for verification, provide higher accuracy than DRE. Considering how much money will be spent in the next year to select the president of the United States, it is remarkable that more money is not being spent to ensure that the new technologies for vote tabulation actually work.
The proposal for voter-verified paper ballots is both practical and intuitive. Electronic recordkeeping systems and paper receipts coexist in many environments. Who would deposit money in an ATM without receiving a receipt? Who would purchase a lottery ticket without an acknowledgement? When was the last time you handed money to a casher or signed a credit card purchase and did not receive a receipt? Even Internet-based commerce requires the prompt transmission of an e-mail confirmation.
In voting, of course, receipts pose a couple of problems. One is that they permit vote-selling. In addition, voting receipts are not an effective technique for auditing since a vote tally reflects the aggregation of many individual transactions. That’s why a group of smart computer scientists has proposed that techniques be developed to allow voters to view the receipt while it remains in the machine; the receipt can then be stored in case there is a recount. Other researchers have developed cryptographic-based receipts that a voter could retain to verify that a vote was included in the final tally but would not indicate which candidate the voter had chosen. Both approaches could do a good job of protecting privacy and preserving the secret ballot.
But some backers of DRE systems seem to run screaming from the room whenever anyone suggests generating voter-verified paper ballots. Apparently, the technology of thermal printers is too complex for the billion-dollar industry. Backers of DRE systems have also tried to confuse the debate by suggesting that voter-verified paper ballots would exclude disabled voters. But this is nonsense. Electronic voting systems, combined with paper ballots, could provide multiple language display and audio for the disabled, and still promote good auditing.
A recent poll on the Web site of the nonpartisan National Commission on Federal Election Reform (run by the University of Virginia and the Century Foundation) asks the question, “Would you feel confident about your vote using a computerized voting system?” More than half of the respondents expressed concern, and 34 percent said they were “not all confident.”
Developers of new voting technology should pay heed. At a minimum, let’s follow the sensible advice of the petition posted at verifiedvoting.org, the site founded by Stanford University computer scientist David Dill. This petition calls for a halt to the adoption of new voting systems that lack adequate methods for auditing, including voter-verified paper trails. Jurisdictions that are seeking to adopt new voting technologies should consider both precinct-based optical scan ballots or touch screen machines that print paper ballots.
Requiring open source software would also be a step in the right direction. Although there would still be many ways to subvert an election, transparent code would remove one of the easiest.
Finally, let’s not forget the central lesson of the Florida election: the technology of voting can determine election outcomes. We should do everything possible to ensure that new voting technologies are accurate, reliable, and transparent. Companies that consider this too difficult should get out of the vote-counting business.