Dan Tokaji's Blog
Professor Dan Tokaji
Election reform, the Voting Rights Act, the Help America Vote Act, and related topics -- with special attention to the voting rights of people of color, non-English proficient citizens, and people with disabilities

Dan Tokaji's Blog Links Publications & Working Papers
Equal Vote
Tuesday, May 2
 
Voting with the VVPAT
I've been writing about the "paper trail" debate for a long time, but today was the first time that I had the opportunity to vote on an electronic voting system with a contemporaneous paper record (aka, "voter verified paper audit trail" or "VVPAT") in a real election. This is the result of a law that the Ohio legislature passed in 2004, amid concerns regarding electronic voting security. Here are my impressions on using that system.

Franklin County, Ohio -- the county in which Columbus is located -- is using the "ES&S iVotronic with Real-Time Audit Log" for the first time in today's primary election. It's a touchscreen direct record electronic, or "DRE," system that prints out the voter's choices on a strip of paper as they're made. That's the VVPAT ... more on this shortly. Before this election, Franklin County used an early model "full face" DRE, in which voters made their choices by pressing on a touchpad rather than a touchscreen. Though this may sound like a minor change, the user interface was actually much different. More information on Franklin County's new system can be found here, including a nifty video on how the voting process works using this machine.

One of the first things I noticed when walking into the room this morning is that the way in which the machines were place in the room had the potential to compromise privacy. On each side of the room, there was a row of machines set up -- three on one side and two on the other. (See my rough schematic drawing to the left, "Actual Set-Up -- the smiley face shows where I voted.) The machines were placed so that, if two machines in the same row were in use, the voter at one machine would be facing the back of the voter in front of him, as both voted. This could allow the voter in back, or a voter passing through the aisle between the two rows, to see the choices o of the voter in front. This is a problem, however, that could be very easily fixed, by placing the machines so that the wall is behind each voter as he or she votes. (See Figure B, "Alternative Set-Up.") There would be no way that others could view a voter's choices under such a configuration.

Actually voting on the iVotronic was quite interesting. After I gave my party affiliation to a pollworker at the check-in table, I was given a slip of paper indicating what ballot I should be given. I then gave that slip to another poll worker, who in turn inserted a card into a slot in the touchscreen unit that activated my ballot. The first screen displayed included three-step instructions on how to vote using the new machine. There were also some instructions printed on the left-hand privacy shield of the machine, which seemed more detailed -- although I could be mistaken, I don't think those instructions mentioned how to "verify" one's choices using the VVPAT.

There's a yellow button marked "next" at the bottom of each screen, which the voter can touch to advance to the next screen. There's also a "back" button that can be used to return to a prior screen. In this election, for which I think there were around 15 contests, all of them could be shown on two screens.

After reviewing the instructions on the touchscreen, I advanced to the first screen, which included the "top-of-the-ticket" contests such as Governor, Attorney General, and U.S. Senator. The voter makes a choice for each contest by touching a box immediately to the left of the candidates name. That causes a check-mark to appear in the box, indicating that a choice has been made. This was relatively straightforward and intuitive. If a voter changes his or her mind, the vote for one candidate can be changed by touching the box by that person's name a second time to cancel that choice. If a voter tries pressing by another candidate's name without de-selecting the first candidate chosen for a given race, the machine won't allow it -- it notifies you on the screen that you can't select more than one candidate for a race, or "overvote," without first de-selecting the prior candidate chosen.

Each time the voter touches inside one of the boxes on the screen, that choice is printed on a reel-to-reel strip of paper immediately to the left of the screen and inside the voting booth. That strip of paper is behind a transparent screen, so the voter can see but not touch it. This is necessary in order to preserve the integrity of the audit trail. If the voter cancels a prior choice that is also recorded, as would be the selection of a new candidate. The paper strips also indicates if a voter attempts to "overvote," as I did. The choices are displayed in pretty small typeface -- I'm not very tall, about 5'8" in heels -- but it was difficult for me to read the print out without bending over.

After each selection (or de-selection) is made, there's about an inch separating that selection from the next one printed on the paper audit record. Only a few voter selections are visible through the transparent screen at any one time. What that means is that the voter must actually check the strip of paper each time he or she makes a selection on the touchscreen, in order to ensure that the print-out matches the voter's intended choices. For example, for a ballot with twenty total contests on it, the voter would have to look back and forth between the touchscreen and the print-out that many times. Needless to say, this would get tiresome very quickly and I doubt many voters will do it.

One of the questions I asked the poll workers is whether there's a way to "scroll back" on the print-out and review all of one's choices at the completion of the voting process. The poll workers confirmed that there's no way of doing this.

After making all of one's choices, there's a "review" button that the voter can press. That caused the ES&S iVotronic to pull up summary screens through which I could "check my work," making sure that the votes recorded matched my intended choices. The system also allows one the opportunity to correct mistakes, by touching the desired contest and re-voting that race. I had deliberately "mis-voted" one item on the ballot, in order to test the review and correction process. Changing votes through the review process was, I thought, straightforward and easy. The machine displays the contest, and allows the voter to correct his or her choices in the manner described above, and then press the "review" button to go back to the review screen.

When a voter makes a change, the printed audit log will display the cancellation of each prior vote and each new choice as made. What struck me as a I was voting is how enormously difficult it would be to actually go through and recount these curled up strips of paper. That's particularly true, given that a correction for a particular race won't necessarily appear immediately below the prior choice for that race. For example, if I voted for one candidate for Governor, and then changed my choice once getting to the review screen at the end, my initial choice would appear at the top of the paper print-out, and my correction at the bottom. I feel very sorry for the poor election officials who would have to actually go through and decipher these print-outs in the event of a manual recount. It would be extremely time-consuing and laborious, making Florida's 2000 "hanging chad" controversy look like a picnic. Don't get me wrong. There's no question that touchscreen machines are a major improvement over the punch card, in terms of accurately recording voter choices and preventing mistaken overvotes and undervotes. But the audit trail would be a major pain to decipher.

One thing that I found a little confusing was the need to actually press buttons twice at the end of the process, before finalizing one's vote. After the last summary screen, there's a button on the screen that is pressed to "cast your ballot now." But even after that, there's another screen that comes up which is necessary to "confirm" the casting of one's ballot. This requires either pressing a big green confirmation button on the touchscreen, or pressing a flashing red light immediately above the touchscreen. I can easily imagining voters thinking they've completed voting, without going throught the final "confirmation" step. This is a problem the Columbus Dispatch discussed in this story yesterday.

When voting today, I waited several seconds before pressing the final confirmation button, in order to ask a poll worker a question. Interestingly, the machine started making a noise when I did so, alerting poll workers that someone had failed to confirm. They very politely and professionally explained that I had to confirm in order to finish. But when I then tried to press the big green "confirm" button on the screen, it wouldn't allow me to do so. Fortunately, I was able to press the flashing red light at the top of the screen, which confirmed my choices and resulted in my vote being recorded.

After confirming my ballot, the print-out scrolled upward and what appeared to be a bar code was generated. This scrolling would prevent a subsequent voter from seeing my choices. However, if someone had access to the paper print-out as well as a list of which voters voted on which machines, it would be theoretically possible to figure out which votes are associated with whcih voters -- thus compromising voter privacy. I say this is "theoretically possible" since I think it's unlikely to happen in a large urban precinct like the one in which I voted. In fact, I'm not sure that poll workers keep track of which voters used which machines. But it might be possible if there were a smaller precinct that had only one machine in use, or if poll workers did track which voters used which machine.

In short, I found the iVotronic touchscreen to be straightforward and user-friendly. Checking the contemporaneous paper print-out was another matter. This would be difficult for all but the most meticulous voters, and ones with good eyesight or the willingness to repeatedly bend over. Also troubling is the difficulty that would inevitably be entailed in recounting the strips of paper that would be generated by this device. Put more simply, my experience suggests that the so-called "voter verified paper audit trail" leaves much to be desired in terms of both verifiability and auditability, at least in this incarnation.

Powered by Blogger Site Meter


Moritz College of Law The Ohio State University