Today, four voting system vendors participated an open public demonstration of their voting equipment at the Alameda County Conference Center in Oakland. Each vendor displayed both an optical-scan system (in which the voter marks a paper ballot and feeds the ballot into a machine to be scanned) and a DRE system (in which the voter makes selections on a computer screen). I spent a chunk of this morning trying out the machines, taking photographs, and talking to representatives from the companies, who were all very helpful and informative. The vendors were identified as A, B, C, and D to encourage the visitors to provide impartial feedback, though it was easy to identify the vendors as Diebold, ES & S, Sequoia, and Hart Intercivic respectively.
Here are my notes from the experience, interspersed with amateur photos of the equipment with plenty of glare.
Vendor A (Diebold)
Diebold’s paper scanner, pictured above, offers a slot into which you feed your ballot, and then the ballot drops into the large locked box. A small LCD on the front left corner counts the number of ballots received. A privacy sleeve (essentially a cardboard folder) is provided so you can keep your ballot covered while sliding it into the machine — the ballot is a letter-size sheet of paper, the sleeve leaves an inch or two of the ballot sticking out the top, and the machine grabs that bit of the ballot.
The ballot consists of boxes of varying widths and heights, one for each contest, arranged on the page. In each box, the candidate name is left-aligned with the left edge and has a corresponding oval bubble at the right edge. Each box also provides one or more unlabelled blank lines on which to write in a candidate, with corresponding oval bubbles. The voter fills in the bubbles with a black felt pen.
The instructions printed at the top of the demonstration ballot tell the voter to darken the oval “provided to the left of the candidate”, which is wrong, though i’d hope that this kind of small error would be caught and avoided in a real election. However, the candidate names and bubbles can be far apart, especially in wider boxes such as the box for “President” on this sample ballot, and the names are indeed closer to the incorrect bubbles on the left than the correct bubbles on the right, which is a potential concern.
I filled in a number of bubbles with varying marks and submitted the sample ballot. The scanner accepted my ballot and i saw the counter increment by one, but there was no indication of whether i had undervoted or overvoted on any of the contests.
- What is the likelihood of a confusing error in the instructions on a real election ballot?
- How significant is the risk of voters marking the (incorrect) nearby bubbles on the left rather than the correct bubbles that may be farther away on the right?
- Will voters understand that the unlabelled lines are blanks for write-ins?
- How many voters will fail to notice the words “VOTE BOTH SIDES” and miss the contests on the back side?
- Does the privacy sleeve provide more of a benefit in covering up the ballot than a hindrance in distributing the sleeves and instructing each voter to carefully align their ballot in the sleeve?
* * *
The Diebold electronic system, shown above, is a somewhat complex-looking device, with a touchscreen display, card reader, numeric keypad, headphone jack, and paper tape printer. The only hardware buttons are on the keypad; there are no buttons on the main device. The display is about 14 inches high and 9 inches wide. For each voter, the pollworker uses a small handheld device to activate a card, and gives the card to the voter. The card selects the precinct and also tells the machine whether to activate the audio interface. The voter puts the card in the slot to start the voting session; the card can only be used once and has to be reactivated by the pollworker in order to be used for casting another ballot.
The session begins with a language selection screen, followed by a screen of instructions that also offers buttons to toggle large text mode and toggle high-contrast (black on white) mode. I’d estimate that the text was about 12- or 13-point Verdana at the “normal” setting and about 21- or 22-point Verdana at the “large” setting. Pressing “Next” on the instructions screen brings up the first page of the ballot. The ballot is divided into pages, with the current page number and total number of pages shown at the bottom of each page. There may be two or more contests on a page, depending how many fit.
If the voter chooses a “Write-in” option for a contest, the machine brings up a separate screen for typing in the write-in name using an onscreen keyboard. After the name has been entered, it appears on the selection screen as the name of the option, instead of “Write-in”. Overvoting is prevented — nothing happens if the voter tries to select additional options beyond the number allowed. If the maximum number of selections has already been made, the voter has to press one of the selected options to unselect it before selecting another option. Undervoting is not indicated until the voter gets to the review screen.
After proceeding through all the pages of the ballot, pressing “Next” on each page, the voter is presented with the review screen. Undervoted contests are highlighted in red. At this point, the voter can choose any contest to go back to that contest, or proceed to print the ballot. Pressing the “Print” button causes the ballot to print out on the paper tape. Since the viewable area of the paper tape is limited to a strip about five inches high, only as much as can be viewed at once is printed on the paper. The selections on the ballot that correspond to the printed section on the paper are displayed on the screen, using a scrollbar if necessary. The voter can then make a side-by-side comparison between the screen and the paper. The printing proceeds this way, section by section, until the entire ballot is printed. At each section, the voter can press “Reject Ballot”, which causes “REJECTED” to be printed on the paper tape.
The voting machine can also be operated by audio, using headphones and the numeric keypad. The audio instructions explain that pressing # speeds up the audio, pressing * slows down the audio, and pressing 0 replays the instructions. The pollworker can choose a mode where the visual display and the audio are active simultaneously, or shut off the visual display and activate the audio only. As the voter enters each contest, the contest is described in audio, and then the screen highlights the first option and the first option is spoken. The audio then explains that pressing 5 votes for the option, pressing 6 moves to the next option, pressing 4 moves to the previous option, pressing 8 moves to the next contest, and pressing 2 moves to the previous contest. These directions are repeated after navigating to any option. All the audio segments are interruptible. Whenever the voter navigates out of a contest without having voted for the maximum possible options, the audio announces, “You have chosen to undervote this race,” before proceeding.
For writing in a candidate by audio, the voter chooses letters one at a time with a multi-tap method. Pressing any number from 2 to 9 causes the first letter corresponding to that digit to be spoken, pressing it again speaks the next letter, and so on, cycling through the three or four letters for each digit. Each letter must be individually confirmed by pressing 0. Pressing 1 inserts a space and pressing * deletes the last letter. Pressing 0 again after entering all the letters of the name completes the write-in.
- Why is “high contrast mode” necessary? Why not simply design the ballot so it reads clearly independent of colour, and without unnecessary colour? The paper ballots are fine in black and white, after all.
- Are the pressable areas of the screen sufficiently obvious to voters, despite the visual inconsistency between the buttons on the bottom of the screen, the buttons on the write-in keyboard, and the boxes for the contests on the review screen?
- Are the varying sizes and shapes of the selection and review boxes confusing to voters?
- Does the protruding paper tape printer on the right make it more difficult for left-handed people to use the touchscreen?
- What advantage is there to presenting the ballot on the screen together with the printed paper tape for side-by-side comparison? Since the voter has already reviewed the ballot on the screen, it would seem that verifying the paper is all that’s necessary. Does dividing the voter’s attention between the screen and the paper make it less likely that the voter will notice errors on the paper?
More on other vendors’ machines in the next part.