Nov. 3, 2008 -- It's voting season again and that means everybody is concerned about the voting process. Will the right people be accidentally removed from registration rolls? Will the wrong people be able to vote more than once? Will the machines suffer "glitches" that cause them to incorrectly record or tabulate votes? Will malicious hackers (much less unscrupulous election officials) tamper with our voting machines, causing them to silently flip people's votes?
My own expertise is in security issues, and I could have written a lot about how insecure our voting machines are. However, believe it or not, security is not the most pressing problem. No, the biggest problem we face is usability. A significant number of voters simply do not find machines easy to use.
Let's take a trip in the way-back machine to Palm Beach County, Fla., in the November 2000 election. Everybody remembers the infamous "butterfly ballot," where an exceedingly poor visual layout of the ballot somehow induced thousands of voters to accidentally vote for Pat Buchanan rather than Al Gore. Once you got past the partisan rhetoric, the problem boiled down to a debate over whether the voters were at fault, or whether the technology was at fault.
The proper answer is that we should blame the ballot. Our goal in designing a voting system should be to accurately capture the intent of the voters. If actual, real voters can't figure it out, then you need to redesign the ballot. If election officials need to create detailed instructions on how to vote, then the voting process has already failed.
For our country's shiny new electronic voting systems, news reports from early voting in the current election as well as experience in past elections all point to huge usability problems. For example, let's consider a common problem: accusations that voting machines are "flipping" people's votes (also the subject of a funny Simpsons sketch). You press the button for one candidate and it shows that you selected another candidate. We see stories like this every year these machines are used. These problems most likely have nothing to do with security attacks; a competent attacker would ensure that the machine shows the voter the proper thing while silently changing their vote on the inside. Instead, they are most likely the result of poorly designed user interfaces.
Most electronic voting machines use touch screens, much like an iPhone or many bank ATMs. While there are several different technologies for figuring out where your finger might be, they all require a "calibration" process. Unfortunately, calibration depends on viewing angle, and that will vary across different voters. That means some voters will need to "aim high" while others will need to "aim low" in order to hit their target. Some voters will figure this out. Others will make mistakes and may or may not notice.
Related Content:Consider the Hart InterCivic eSlate electronic voting system. Rather than using a touch screen, they instead have an iPod-like scroll wheel. If you vote "straight ticket," it places a check next to all of the party's candidates. You can then scroll down, highlight a race, and click again to make changes. To a seasoned iPod user, all of this makes perfect sense. To somebody who has never used a computer, much less an iPod, they may misunderstand that a straight ticket vote followed by a click on the presidential candidate will actually deselect the presidential race.
To make matters worse, consider the battleground state of North Carolina, which has the inexplicable rule that a "straight ticket" vote applies to every race except the presidential race; no other state has such a rule. How many voters, particularly first-time voters, are going to mess this up? What's going to happen if the presidential margin of victory is smaller than the number of "blank" votes in the presidential race and the outcome in North Carolina ends up being crucial to the 2008 presidential election? It won't be pretty.
When you speak to election officials or vendors, they will discuss the "summary screens" presented at the end of the voting process. These give voters a chance to detect mistakes they may have made and to go back and correct them. At Rice, we built our own voting machine (now available as an open source project) and created an "evil" version that would lie on the summary screen. When we brought test users in from the Houston community, we found that a staggering 63 percent of them failed to notice errors on the summary screen (read the full study for details). We can conclude that the summary screen is an ineffective mechanism for improving accuracy for most voters.
What are we supposed to do? Between now and Nov. 4, there isn't much we can do. Early voting has already begun in some states. If you change election procedures this late in the game, you invite chaos. Yet if you leave them alone, you invite lawsuits when political candidates look at the totals and refuse to accept them as accurately representing the will of the people.
Exactly this sort of lawsuit happened following the November 2006 election in Sarasota, Fla.'s 13th Congressional District, following a screw-up where the ES&S iVotronic electronic voting machines recorded that one in seven voters had no preference in the congressional race; the under-vote rate was two orders of magnitude higher than margin of victory, and any reasonable statistical model that would reassign even a fraction of those blank votes would have resulted in a different candidate serving in Congress. Despite a variety of studies, we still have no conclusive evidence to explain what went wrong. Florida did the right thing and pitched its electronic voting machines, replacing them with optical scan paper ballots. And the two candidates from that 2006 race are facing off against each other again.
Once we get past this November's election, other states will hopefully follow Florida's lead and pitch voting equipment that has been shown to have so many security and usability problems. More realistically, the money isn't there, and there are unacceptable security problems even with current optical scan systems. Voting system vendors claim it takes four years to get a product from conception to delivery, and are pushing to reform the slow certification system. I'll agree that certification is slow, but the vendors are showing no signs of delivering meaningful improvements to their products, much less the evolution that will be necessary to address usability and security concerns.
Dan Wallach is an associate professor in the department of computer science at Rice University in Houston, Texas and is the associate director of the National Science Foundation's ACCURATE (A Center for Correct, Usable, Reliable, Auditable and Transparent Elections). His research involves computer security and the issues of building secure and robust software systems for the Internet. He has testified about voting security issues before government bodies in the U.S., Mexico, and the European Union, has served as an expert witness in a number of voting technology lawsuits, and recently participated in California's "top-to-bottom" audit of its voting systems.
Related Links:
our sites
video
mobile
shop
stay connected
corporate
