Douglas Jones is a voting technology expert on the computer science faculty at the University of Iowa who has done extensive study of the ES&S iVotronic direct recording voting machine. In this interview he offers his insights on reports of straight-party voting problems on iVotronics in multiple North Carolina counties.
Q: Are you familiar with this type of problem?
DJ: There have been sporadic reports of “vote flipping” since the iVotronic came into widespread use, but there has never been a good explanation, and I am not convinced that there is just one problem. It may be that several different problems are being confused. In 2004, I was involved in assessing the iVotronic voting machines used in Miami-Dade County, Florida, and one thing I investigated was the possible cause of reports of “vote flipping”. My report is available online (see particularly section 11, pages 20 to 23.)
Since writing that report, I have heard several reports from pollworkers and others who have observed voters who either touched the screen with two fingers or accidentally rested the thumb of one hand on the screen while voting with the other, causing their touches to be misinterpreted in exactly the way I described in that report.
I have also conducted a detailed scientific study of how voters interact with broken touch-screen voting systems. Among the things we did in our experiment was to deliberately miscalibrate the machine and then observe how this interfered with the voteer’s experience. We noted that most voters touch the selection box below the center, so miscalibration that shifted the apparent position of their touch down was more likely than miscalibration that shifted the apparent position of their touch up. The experiment also included deliberate software-induced vote flipping – we switched Obama and McCain on the theory that voters would be more likely to notice that flip than any other. (See here and here.)
Q: Many Republican voters complained that their votes were flipped to Democratic, but few if any complained the other way. The ballot placed the Democratic touch button above the Republican button. What do you make of this?
DJ: This is significant. It is important to understand that “vote flipping” is not necessarily a single problem. It is the popular name for any case where voters report that the selection they tried to make is not what they see reflected on the screen. It’s very difficult to investigate because it’s illegal to stare over voters’ shoulders, and none of the current voting machines record anything useful for post-election diagnosis of what happened. The hypotheses to explain vote flipping include:
* Miscalibration This should be pretty random, up, down, left and right. Furthermore, the user interface of the iVotronic, the last time I tested them (admittedly, that was several years ago) requires the pollworker to select the ballot style using a little tiny checkbox before backing away from the machine to let the voter vote. As a result, miscalibration is far more likely to be noticed by pollworkers than voters, since the pollworkers are trying to touch much smaller targets and must do so before each voter votes. I don’t know what happens if there is only one ballot style. In that case, perhaps, pollworkers wouldn’t notice the problem.
* Voters who steady themselves with one arm while voting with the other. This is the “thumb on the screen” scenario that I described above. People with small hands would be unlikely to have this happen, while big ham-handed people (like me) can trigger this. I found it out by accident while in Miami trying to find other explanations. This can shift things up or down depending on where the thumb lands, but big ham-handed voters will probably grasp the screen higher, leading to an upward shift. It would be interesting to know if those who complained had big enough hands to trigger this problem.
* Voters who vote with their index finger but don’t make a tight fist to keep the other fingers of their “voting hand” off the screen. If you vote with your index finger while making a relaxed fist with the other fingers, a gentle drag of the middle fingernail on the screen can shift the touch position downward. If you hold your fingers out, not making a fist, deliberately touching with the index finger, the pad of the middle finger can brush the screen, shifting the position up (this seems less likely to me, but I’m listing hypotheses here.)
* Fraud. Pollworkers or technicians can, in theory, deliberately miscalibrate the screen. I discussed this in my report. I personally think it would be difficult to do this on an iVotronic because it would have an effect on all the pages of the ballot and also on the pollworker screens used to open and close the polls and to select ballot styles before each voter uses the machine.
The firmware in the machine could also be tinkered with to deliberately flip votes. Most people who talked about “vote flipping” on iVotronics in the early years of the iVotronic assumed that it was this, and in the experiments I conducted above, we simulated vote flipping this way. The other hypotheses listed above seem more likely, but paperless DRE machines don’t record enough information to rule out fraud and they retain little information that would help in a post-election forensic investigation.
Q: An election official has stated that the most common touch screen calibration problems shift votes upward. What do you think?
DJ: As I noted above, it depends on what the underlying mechanism is. True miscalibration tends to be random, up, down, left and right, usually by a small amount but sometimes by a larger amount. Thumping the voting machine (during delivery to the polling place, for example) can change the calibration in any direction. So can temperature changes. In my experience, because voters usually touch below the center of the selection button, it takes a smaller downward shift to misinterpret their touch than an upward shift.
The ham-handed miscalibration hypothesis is different. In that case, it may be that the upward shift is more likely. Nobody has done really good studies of this. The loose-fisted hypothesis would tend to shift things down, but again, these are poorly studied.
One really big problem with scientific study of these kinds of things is that swinging the vote by one percentage point can have a huge impact on the national government. One vote per precinct, nation wide, can change which party controls the US house. So, with elections, we really care about things that might only interfere with one voter in one hundred. Many human factors researchers have told me that if you test a system with 20 people, that’s good enough, because if anything is really wrong, you’ll see it. That may be true for testing an iPhone, but with elections, it’s unlikely to show up the ham-handed voter’s thumb on the screen. For all I know, that might be just only one voter in 400.
One thing we did in our experiment was to ask a number of user satisfaction questions as a followup. An important question in this batch was: “Are you confident that the machine correctly recorded your vote?” The majority of voters were very confident. Curiously, this confidence did not go down much when we rigged the machine to flip the votes for McCain and Obama.
Why? First, note that half the voters did not notice this switch. People felt strongly about their presidential preference, but people, in general, are not very good at proofreading. The summary screen that comes up at the end of the voting session clearly showed McCain for the Obama voters and visa versa, yet they did not notice.
Among people who did notice, the usual reaction was “Huh?” and then they went back to fix the problem. In general, people assume that the machine is right and assume that it was their error, not the machine’s dishonesty. Only a small fraction of our voters (or more properly, experimental subjects) commented on the fact that we’d flipped their votes.
This has important consequences for the real world. Most people do not complain when there is a problem. They assume that it was their mistake and go back and fix it. I assume that things would
be different if the machine did not let them fix the vote flip, but if the machine lets them fix it, the fact that some voters are complaining suggests that there may be far more voters who notice the problem and are silent, and that even more voters may have had their votes flipped but didn’t notice the problem.
It also raises serious questions about voter satisfaction and voter confidence as measures of the effectiveness of touch-screen voting systems. People trust computers, and they keep trusting them even when the computers are programmed to cheat in a way that the voters notice.
Q: The state Board of Elections has emphasized that the confirmation screen protects against these kinds of problems, but you seem to be questioning its effectiveness. Can you say more?
DJ: Our experiment is not the only one to notice that people don’t catch errors in the confirmation screen. Sarah Everett at Rice University did a similar experiment and noticed that only about 1/3 of voters noticed flipped votes on the confirmation screen. (Everett’s experiment was the basis of her PhD dissertation at Rice.) Her experiment used fictional candidates, which is to say, there was no name recognition. We used the real ballot from Johnson County, Iowa and limited our subjects to Johnson County voters, so we had good name recognition. Switching McCain and Obama raised the number of voters who noticed the flip, but it was still only about half. Everett also measured voter satisfaction, and as in our experiment, it remained very high even when the machine was flipping votes.
The lesson here is simple. Review screens are not very effective.