Again this election cycle, stories have emerged about “vote flipping”, most notably in Texas, where a video of erratic touchscreen behavior was posted on several sites, and in several North Carolina counties. (linklinklinklink) As voting technology expert Douglas Jones wrote several years ago, it seems unlikely that vote flipping is evidence of intentional hacking. However, these incidents do highlight the lack of transparency of software-generated election results and undermine confidence in elections generally. Vote flipping can be caused by a voter touching the screen in two places, for example resting one hand on the machine while making selections with the other (see pp. 20-22 here), but the most likely cause of “vote-flipping” is miscalibration. As Rice University computer scientist Dan Wallach explains in a post at ACCURATE:

The screen shows pictures of buttons with labels for the various candidates, which the voter selects by touching the screen with their finger. Some voters using these machines have reported problems where they pressed the button for one candidate and a different candidate was selected. These issues are most likely the result of poor touchscreen calibration rather than any security problems with the voting machines’ software.

The clear, touch-sensitive layer is separate from the part of the screen that displays the buttons. The thickness of the touch-sensitive layer directly implies that when different voters are looking at the screen from different angles, they will naturally want to touch the screen at different locations. This can be partly addressed by “calibrating” the touchscreen in advance. The calibration process, familiar to anyone who owns a PDA, involves the machine displaying a series of cross-hairs and asking the user to press on the center of each cross-hair. The machine then computes a correction to ensure that selections are mapped to the correct part of the screen below. Of course, if the calibration was done incorrectly, or even if the voter is notably taller or shorter than the person who did the calibration, then presses on the screen might still be misinterpreted. Furthermore, different voters may use different parts of their finger (ranging from the fingernail to the whole finger), which may differ from how the system was calibrated. (See also “Touch Screen Usability: Election Edition!” and “Vote Flipping and Touchscreens“) Vote flipping was investigated in several articles during the 2008 election cycle. Computerworld interviewed both voting machine vendor and election integrity activists for “Are design issues to blame for vote ‘flipping’ in touch-screen machines?” and Wired magazine posted an article about the potential for maliscious calibration as detailed in the Ohio EVEREST report.

In 2008, Verified Voting and the Brennan Center for Justice wrote to Secretaries of State in 16 states that use ES&S iVotronics, recommending that pollworkers re-calibrate iVotronic touch screens each morning during early voting. The North Carolina Board of Elections recently advised all county election directors in the State using the ES&S iVotronic voting machines be recalibrated at the beginning of each day of early voting, a practice that should be adopted by all jurisdictions still using touchscreen voting equipment. Such recalibration will have no effect on votes already cast in the machine and in most cases should not take more than a minute or two. Recalibration can, and should, be done during election day as well if voters report problems that could be the result of miscalibration.

It is worth noting that the touchscreen technology incorporated into fielded voting systems in the US is nearly a decade old and not that found, for example, on current generation smart phones. The machines themselves are also several years old, so we can expect problems resulting from aging touchscreens to increase with each election cycle.