Counting On Computers To Count The Votes
Just weeks before the November election, a session on e-voting at MIT's Emerging Technology conference on September 25th gave me the same level of warm and fuzzy that I got from watching President Bush's recent speech on the state of our economy.
How, I asked myself, can so many great minds screw things up so royally?
The panelists were united on one theme: The Direct Recording Electronic (DRE) voting machines that popped up around the country in the aftermath of the Florida debacle in the 2000 elections are as bad, or worse than those punch card machines that gave us the "hanging chad."
Panelists included California Secretary of State Debra Bowen; Doug Chapin, director of electionline.org; Ronald L. Rivest, professor of electrical engineering and computer science at MIT; and Pamela Smith, president of the Verified Voting Foundation.
It's not simply a question of whether these electronic machines are accurate and reliable, but also whether the voters perceive them as such. After all, many of these machines - at least initially - do not leave a paper trail.
"Nervousness has always been an issue," Chapin said, "even before the 2000 elections."
"The question," Rivest said, is "whether you can provide assurance that the count is providing the right answer."
He added: "Did the voter cast the vote as intended, was it collected as cast, and was it counted as collected?"
The problem, in part, stemmed from well-meaning federal legislation (Help America Vote Act of 2002) which provided funding for states to replace punch card or lever voting machines with systems that are less prone to confusion and miscounted votes.
Unfortunately, many of the machines used to replace these older systems had issues of their own, including vulnerability to hackers and lack of a paper trail to verify votes.
Essentially, voters were (and still are in some states) voting into a black hole. Only the hard drive or memory in the machine records their vote and as computer experts will tell you, software behind any electronic device is subject to all sorts of problems, bugs and limitations. What's more, the software code used in most of the machines is proprietary which makes it difficult to look at even by experts, let alone county election officials who often lack even the most basic technology experience.
Bowen, whose office sponsored an extensive review of machines used in California, questioned the software inside the machines. In 2007, the secretary of state commissioned a "top-to-bottom review" of voting machines that resulted in the decertification of DRE systems from Diebold, Hart InterCivic and Sequoia. At the time of the study, Bowen said that researchers were "able to bypass physical and software security in every machine they tested."
On the MIT panel, Bowen called for the use of open source software that is transparent to anyone with the technical skills to understand it. That may not include the average voter or election office, but with open source code, at least some software engineers have the ability to inspect and even improve code.
Proprietary software is closed to inspection without permission from the vendor who wrote the code. That's arguably OK for office applications used by businesses but not for software that determines who our future public officials will be.
After Bowen's review, California retired paperless voting machines.
"We don't use touch-screen-voting systems that don't leave a paper trail in California," said Bowen in an interview. She added that many states have moved away from touch-screen-voting systems entirely and have gone back to fill-in-the-bubble optical scan systems "because we know they are accountable."
About three-quarters of the states, according to Smith from the Verified Voting Foundation, "either passed a requirement for voter verified paper ballots or have bought equipment that is verifiable without having passed a law."
Unfortunately, Smith said, "that leaves out about a quarter of the states, including some large states such as Maryland, Georgia, large parts of Texas and most parts of Pennsylvania, which is a big swing state." A map on the Foundation's website shows what type of ballots are used in each state.
Both Smith and Bowen prefer paper ballots that can be read by high-speed scanners. If something goes wrong or if there's a challenge, question or recount, the paper is there as a backup.
Because of the way our Electoral College works, states that don't use a verifiable voting system could affect the outcome of our Presidential election. We all have a reason to care about every vote in every state, especially in those "swing" or "battleground" states where a small number of votes could determine who moves into 1600 Pennsylvania Avenue next year.
Whether the Electoral College is a good or bad way to pick a president is beyond my expertise as a technology columnist, but I do know enough about technology to agree with computer scientists who question whether software and silicon is up to the task of reliably recording our votes.
By Larry Magid