Features

net.wars: Voters in, garbage out

Heard the one about the election mess in Florida?

It was so bad they're having to bring in observers from DC, just like one of those third-world countries that can't be trusted to do elections.

The election in question is not 2000, but the primary held on September 10, 2002. Florida has spent $125 million across many municipalities to update its balloting systems, replacing punch cards with electronics. The result? They had to declare a state of emergency in Dade and Broward counties and keep the polls open extra hours to make up for the amount of time the new machines were unavailable. Florida is now spending an extra $1 million.

Given that the mess inspired a column from Carl Hiaasen, I guess that's the good news. The bad news is that the UK is also considering electronic voting, and you have until October 31 to lodge your comments to the consultation documents. If you're in the US but outside Florida, Congress has just a approved a $3.9 billion election reform package intended to make it both easier to vote and harder to commit voting fraud. Voter ID databases!

Last night, Rebecca Mercuri, who's been researching this area for 13 years, explained at a meeting convened by the Foundation for Information Policy Research why we should be alarmed by and resistant to directly recorded electronic voting. This is not to say she sees no role for computers in voting systems. She just doesn't think the votes should be directly electronically recorded by machines with no checks, balances, or audit trails. After listening to her and privacy advocates, cryptographers like Bruce Schneier and computer security experts such as Peter G. Neumann, I'd agree.

Vendors promise governments that electronic systems will be more accurate, produce faster returns, be more resistant to fraud, and increase turnout. Politicians particularly stress increased turnout. I'm not sure if they think it makes them look inclusive, or if they just believe increased participation means more votes for them. The trouble is, as Jason Kitcat says, that many factors contribute to lower voter turnout, and only one of them (ease of access) can be helped by throwing technology at the problem.

For voters to be engaged, they must believe their vote matters and is counted. The worst part of the 2000 Florida fiasco (and to be fair, there were many other races that had problems but that didn't get the same coverage) was the disenfranchisement of 4,000 voters through what Kitcat refers to as "database abuse." Four thousand voters removed from the rolls; final margin less than 400. But even the more famous chad problem isn't solved by direct electronic recording, since there's no way for voters to check that their vote was as intended. You could argue that the same can be said of paper ballot stuffed into boxes, but the number of observers and the physical objects are reassurance, too. But no one can come along later and find a shoebox full of electronic votes hidden behind the toilet.

The vendors of electronic systems will tell you they've been tested and proved to have a high level of accuracy, but Mercuri's investigations say that the tests are carried out on pristine data in a controlled environment. They do not test for voter error. In Philadelphia, near where she lives, the touch screens proposed for use in the November gubernatorial election fail on obvious, well-known usability principles: the big green button that says VOTE ends your session and is available on first screen with no confirmation; the button labeled NEXT that takes you to the next race is smaller, not as green, and easily overlooked. Some of the candidates in races that appear later on the ballot are considering suing - even before the election. Mercuri's research indicates that some 3 to 5 percent of electronically recorded votes are actually lost, greater than the margin of error in Florida and other states in 2000. And the promised speed of compiling returns? Actually the electronics are slower than humans.

Mercuri's own preference, based on a lot of research plus her extensive experience as an election officer, is the paper-based Marksense system. If we must electronify, her proposal is touchscreens with printers directly attached that issue receipts voters can use as verification. The cryptographer David Chaum, the original digital cash man, has an elaborate scheme for a two-layered receipt printed in dots so the layers together verify the voter's choices but separately show nothing useful. The voter keeps one layer; the election commission the other. Later, the voter could check on a Web site that the vote had actually been counted. It's clever, it's somewhat elegant, and it will probably never happen.

The worst thing about the electronic systems is that vendors are using trade secret law to refuse to reveal details of how they work - and election boardss are being required to sign NDAs in order to buy the equipment. Now, it seems to me there's a simple response: "Fine, I won't buy the equipment. Don't you *want* customers?" But officials are actually agreeing to this, with the result that the democratic process is effectively being privatized. Is that what you want?