Beware The High-Tech Ballot

Untested technology and poor training can lead to disaster


Even by the standards of Washington newspeak, the Help America Vote Act (HAVA) is in a class by itself. For more than 200 years, the U.S. ran elections quietly, efficiently, and mostly honestly. But since passage of the 2002 law, the simple act of casting a ballot has become the stuff of high drama. There's a lesson here about the use and abuse of technology.

HAVA was a response to the 2000 Presidential deadlock in Florida. But it was based on a misdiagnosis of the problem. The inescapable fact in Florida is that the outcome was a tie. George W. Bush's official margin over Al Gore was 0.009% of the votes cast, and since any method of casting or counting ballots involves error, that difference was just too fine to split. But images of butterfly ballots and hanging chads convinced politicians and the public that the problem was bad technology and that the solution lay in changing how the nation votes. HAVA provided more than $3 billion to pay for high-tech systems, and vendors rushed to fill the demand. Although there was no money for training, "states were forced to go out and buy equipment," says Jim Adler, president of VoteHere, a Bellevue (Wash.) firm that develops ballot auditing systems.

THE RESULT HAS BEEN CHAOS in one election after another, with worse likely in store for November. Much attention has been given to the fact that electronic voting machines, especially those made by Diebold Election Systems, may be vulnerable to attacks by hackers. And Princeton University computer scientist Edward Felten recently showed that the physical lock that protects Diebold machines from tampering can be opened with keys easily purchased on the Internet.

While there have been no reports of attacks on machines that are actually in use, the risks Felten and others have pointed out are serious and highlight the danger of rushing to embrace untested technology. Failures, both human and electronic, are already affecting the ability of people to vote, as I learned in Maryland's Sept. 12 primary. All was confusion at our local elementary school because county officials had neglected to send out the access cards voters use to cast ballots on the Diebold AccuVote-TS machines. I was lucky; I got to cast a provisional paper ballot, supplies of which were quickly exhausted.

In the turmoil, I barely noticed the new electronic poll books, also from Diebold, that replaced an old but utterly reliable paper check-in system. Aviel Rubin, a Johns Hopkins computer scientist who worked as an election judge in Baltimore, reported that the system was incapable of keeping track of who had voted and locked up after about 40 voters had checked in. These results were later confirmed by official state tests. Four weeks before Election Day, Maryland officials are debating how voters will check in and how they will cast their ballots. With no routine for poorly trained poll workers to rely on, there's a great chance of another mess.

There are important take-aways about technology adoption, which go beyond elections. First, gadgetry cannot cure a problem that's not technological to begin with. In the case of voting, this was evident in the ultraclose 2004 Washington state gubernatorial election, where the latest voting technologies didn't prevent a drawn-out legal muddle. Second, while mature technology, applied judiciously, can reduce human error, untested technology foisted on poorly trained workers is an invitation to disaster.

It's important to keep such examples in mind as we try new technologies in other areas. Electronic databases of medical records could be of enormous benefit. And a planned electronic surveillance system on the Mexican border could be more effective and less intrusive than a physical fence. But if we rush into new systems without adequate regard for reliability, security, and privacy, the consequences could be far worse than the voting mess.

For past columns and online-only reviews, go to Technology & You at

By Stephen H. Wildstrom

    Before it's here, it's on the Bloomberg Terminal.