E-voting reliability still has a ways to go

By Dan Gillmor

Mercury News Technology Columnist

07 November 2004

It didn't take long after last week's elections for the conspiracy theories to take flight. One, in particular, started to ricochet around the Internet even before John Kerry conceded the election to President Bush and it blamed alleged tampering with electronic voting machines for Kerry's defeat.

I have no reason to believe this scenario. While there were some remarkable discrepancies between the exit-polling surveys and the final tallies in some cases, the discrepancies weren't nearly as consistent as critics charged, nor limited to states where voters used the controversial kinds of electronic voting machines. I think the exit pollsters got it wrong, plain and simple.

I wish we could debunk the theory of e-voting shenanigans more conclusively. Yet the very design of the machines means we cannot at least not in this election.

In a couple of years, the reliability of e-voting technology may be less of an issue. It shouldn't have been an issue at all this year, and that's a scandal in itself.

Media coverage since the election suggests that people using e-voting machines, especially touch-screen video ballots, encountered scattered problems around the nation. One of the more bizarre glitches came in Ohio, where an e-voting system awarded almost 4,000 extra votes to Bush in a precinct near Columbus.

There were disturbing, anecdotal reports of voters ing Kerry but then seeing the machine indicate a vote for Bush. Poor calibration of some machines may well account for these difficulties.

Too bad we'll never know for sure. The majority of this year's electronic voting machines had no auditing system that most experts consider fully trustworthy. They produce a result in their digital circuits, and if you ask for verification, they cough up the same result.

Imagine that you buy a calculator that adds 2 plus 2 and gets 5. You do it again and it gives the same answer. This doesn't make it true, even if the calculator seems to be working. (Voting machines have more safeguards, but you get the general idea.)

Again, I'm not suggesting this happened with voting machines last week. What I am saying is that despite ample warnings from people who know technology and its shortcomings, we have no way of being 100 percent certain that it did not fail.

Even the possibility that an election tally could be undetectably wrong, whether deliberately or by mistake, is utterly corrosive to confidence in our system. Yet the manufacturers of these machines, and the voting officials who grabbed a big pot of taxpayer dollars to install them, have at every step been skeptical of or outright opposed to some simple safeguards.

The voting officials, in some cases, took the word of companies that all was well. But when a raft of convincing data showed that the companies were not providing appropriately secure or reliable technology, too many voting officials reacted with disdain for the whistle-blowers instead of moving swiftly to protect the public trust.

The machines and software running them, apart from the manifest flaws, are black boxes. The code is proprietary, trade secrets of the companies with no serious scrutiny from people who could find the flaws that always exist in software. The machines are certified by companies beholden to the manufacturers.

One interim step toward doing things right is a voter-verifiable paper trail a system by which voters could see a printout (but not handle or keep it), which then would be saved in case of a recount or for random checking of machine accuracy.

Alone among the states using e-voting this year, Nevada required this kind of paper trail. (It's worth noting that the exit polls there also overestimated Kerry's support, a piece of evidence strongly countering the stolen-election theorists.)

California will require a paper trail by the 2006 elections. It should have done so this time. Other states may follow.

What California did do this time was to require, in counties that had touch-screen systems, the option of voting on paper. In my home of Santa Clara County, in the heart of Silicon Valley, voting officials ordered poll workers not to volunteer the fact of this option to voters an outrageous abuse of authority.

A bill before Congress in the last session would have required a paper trail for all e-voting machines. It had no Republican support, a curious fact given the fact that the chief executive of one of the companies, Diebold Election Systems, making the machines was an ardent supporter of Republican causes, and that a Republican senator had an ownership stake in another voting-machine company, Election Systems & Software.

We shouldn't leave voting to the whims of a flawed marketplace. We should require voting machines to be based on open-source software code, so that experts could vet the machines for reliability and accuracy. Then, to be even more certain, we should require a paper trail.

It baffles me that so many elected officials shrink from fixing a broken system. It frustrates me even more that voters are going along with it.

Few people would use an automated teller machine at your bank if it didn't spit out a paper receipt. Yet many of us were blithely willing to trust our votes to similar machines, running flawed software, with no such backup.

America dodged a bullet in the 2004 presidential election. We're lucky it wasn't as close a contest as the last time around. But the gun is still loaded.