Juan Gilbert, Aqueasha Martin, Gregory Rogers, Jerome McClendon, Josh Ekandem
The "finger roll," a well-known shot in the sports world, was once known (and still is) as one of the most difficult to master in the game of basketball. However, in the world of voting and electronic ballots, the same is not true. During the 2008 presidential election in West Virginia, an alleged case of "vote flipping" surfaced, introducing questions related to electronic ballot design. Several videos appeared, demonstrating a voter selecting a candidate on a direct-recording electronic (DRE) ballot by touching the candidate's name positioned in the top left corner of the button. But in doing so, the voter would inadvertently select the button directly above the intended choice. This gesture as it relates to a DRE ballot is also known as a finger roll, because a voter's finger may roll up, down, or sideways depending on the positioning of the touchscreen. The gesture proved not as difficult to master by West Virginia voters as by its basketball-playing counterpart. The larger issue was that some voters did not realize their mistake until it was time to review their ballots.
The election situation presented a few intriguing questions: What are users inclined to touch (a name or a checkbox) and how should the buttons be designed and positioned so that errors are minimized? More specifically, how should the buttons be designed and positioned on electronic voting machines or any other graphical user interface to reduce unintended selections?
In 2007, the U.S. Election Assistance Committee (EAC), founded "to provide help and assistance to the electoral process," issued some guidelines and recommendations for electronic ballot design . The EAC provides eight usability recommendations that should be considered when designing an electronic ballot, including: voter needs, unambiguous process, simple language, color use, rights to review, font type and size, candidate order, and icon. However, these recommendations are voluntary and can be ignored at will. Although the EAC's regulations appear to be based on sound principles of interface design, there has not been sufficient research to warrant recommendations on the placement of text on an electronic ballot. But due to gestures such as the finger roll, such design recommendations may be key to improving voters' interactions with DRE ballots and reducing unintentional voter selections.
One might argue that unintentional voter selections on DREs are not a huge concern, since most electronic ballots include a review screen for voters to catch any mistakes before ballots are cast. However, research suggests that this is not necessarily the case. In a study by Bryan Campbell and Michael Byrne, participants were asked to vote and were instructed to verify their choices during a review process. Before the review process, the voters' choices were altered to determine if they would notice a difference in their selections on the review screen. The findings were that out of 108 participants in the study, only 50 percent noticed changes . A similar study conducted by Sarah Everett at Rice University found that less than 40 percent of participants detected these changes . In cases when elections are close, as little as 10 percent of flipped votes can change the outcome of an election. For example, in the 2008 Minnesota Senate race, nearly 2.4 million votes were cast . The margin of victory was 312 votesless than .5 percentand a winner was declared 239 days after the election.
To get a better understanding of the events in West Virginia, we conducted a study. The goal of the study was to better appreciate how the positioning and layout of controls (e.g., buttons) and related text and objects (e.g., checkboxes) influence where voters touch the electronic ballot. The study aimed to answer the following questions:
- When participants are provided a selection object, such as a checkbox on an electronic ballot, that is similar to the one used in the 2008 West Virginia elections, how many opt to touch the candidate's name as opposed to the checkbox?
- If participants are provided a different electronic ballot design, where text is horizontally and vertically centered, would overall participant interaction with the electronic ballot change?
The study was conducted in two rounds and included a total of 55 participants34 participants in the first round and 21 participants in the second round. The first round of studies included participants ranging in age from 19 to 29, with a mean of 20.12 (SD=2.07). Participants were recruited on a volunteer basis via mass emails to students in computer science courses. A total of 34 graduate and undergraduate students participated. One participant was removed from the sample because of incomplete information. Of the participants included, 26 (79 percent) were male and 7 (21 percent) were female. No participants reported disabilities. Ninety-four percent of participants had completed high school, 3 percent had completed a bachelor's degree, and 3 percent had completed a master's degree.
The second round of studies included participants ranging in age from 49 to 91, with a mean of 69 (SD=4.2). We recruited participants on a volunteer basis through the Osher Lifelong Learning Institute (OLLI) at Auburn University. A total of 22 OLLI members participated. Thirteen females (59 percent) and nine males (41 percent) participated, of whom two reported disabilities. Of the 22 total participants, 28 percent reported high school as the highest degree obtained, 27 percent had completed a bachelor's degree, 27 percent had completed a master's degree, and 18 percent had completed a Ph.D.
We simulated the design and layout of the electronic ballot used in the 2008 West Virginia elections, since this is where the concern first surfaced. This served as the control ballot. The candidate's name, along with a checkbox, was placed at the top left-hand corner of the buttons (Figure 1). When a voter selected a candidate, a check appeared in the preceding checkbox. For the experiment prototype, the text was positioned in the center of the buttons without a checkbox (Figure 2). Each prototype included a mock election ballot that contained three contests and one proposition.
The prototype system included a feature that recorded where participants touched the screen. A screenshot provided the view of the electronic ballot at the time the participant touched the screen. In addition, it included a blue dot representing where the participant touched the screen (Figure 3). Screenshots were taken for each interaction with the electronic ballot, except for cases when the participants chose to write in a candidate. Screenshots for each interaction were stored in either a control or experiment folder for each participant.
Results from the first round of studies indicated a total of 98 touches from the control interface and 98 touches from the experiment interface. The second round of user studies indicated a total of 58 and 55 touches for the control and experiment interfaces, respectively. A total of 309 touches were recorded between the two rounds of studies. Because write-in votes did not capture the participant's interactions with the controls or text, write-ins were not included in the tally and were excluded from the screen recordings.
As shown in Figure 4, in the first study, when presented with the control interface, 63 percent of the time participants selected the checkbox when choosing their candidate. Thirty-one percent of the time participants selected the button, and only 6 percent of the time participants selected the candidate's name. However, when the checkbox was removed in the experiment interface, 61 percent of the time participants selected the candidate's name, followed by participants selecting the button itself 39 percent of the time.
In the second round of studies, 86 percent of the time participants selected the checkbox when it was provided as an option. When the checkbox was removed, 84 percent of the time participants chose the candidate's name.
In total, 55 people participated in the studies. Out of the 156 recorded touches in the control studies, 72 percent touched the checkbox, 6 percent touched the candidate's name, and 22 percent touched the button. Out of the 153 recorded touches in the experiment studies, 31 percent touched the button, and 69 percent touched the candidate's name.
Interestingly, regardless of the presence or absence of a checkbox, a percentage of the time (5 to 6 percent) study participants elected to touch the candidate's name as opposed to the supplied checkbox. In a real election, if 5 percent of the voters touched the candidate's name and some of these selections resulted in unintentional voter selection, this could be more than enough votes to change the outcome of a close election, given the significant number of people who do not verify their ballot choices [2,3].
The findings suggest that ballot design may influence how voters interact with controls on an electronic ballot. In both studies, more voters selected the checkbox when it was present, followed by selecting the button itself, with an even smaller number touching the candidate's name. However, when the checkbox was removed, more voters selected the candidate's name than the button.
The results of this study suggest that the use or absence of checkboxes may influence user interactions with an electronic ballot. However, some users still elect to select a candidate's name. We therefore suggest that candidates' names be centrally located on buttons, both vertically and horizontally to minimize unintentional voter selections. If checkboxes or other images are used, they should be vertically centered as well.
Due to limited screen real estate, ballot designers may minimize gap space between buttons to accommodate all candidates in a contest. This design choice provides the opportunity for finger roll, as a voter's finger may roll up, down, or sideways depending on the positioning of the touchscreen. The recommendations presented here are based on the use of sufficiently large buttons within the design. Sufficiently large buttons are buttons that are large enough to provide space for a voter's finger to roll on a selection without rolling into the area of the adjacent or surrounding buttons. Illustrations of recommended designs are presented in Figure 5.
In design option 1, the text is horizontally and vertically centered within each button (Figure 5a). Centering text in this way reduces unintentional voter error by allowing space for selection while minimizing the chance that a voter's finger inadvertently rolls outside the button area. This allows the voter's finger to roll up, down, left, or right upon selecting the button while providing ample space between the text and the top or bottom of the surrounding buttons. Once a selection is made, a checkmark appears beside the user's selection, providing a visual confirmation of the selection (Figure 5b). If checkboxes must be used, we recommend design option 2. Both the checkbox and text are vertically centered and left-justified (Figure 5c).
Electronic ballot design is a subset of user interface design, and like any other type of interface design, design principles and usability testing are essential for providing voters with effective and efficient systems. The EAC provides a guide for designing electronic ballots for DRE systems; however, the guide does not include detailed recommendations for the layout of controls, related text, and objects with which a voter must interact. Our research supports the hypothesis that placement of text affects how users interact with an electronic voting system. Therefore, as in West Virginia, poor placement of text could potentially leave voters vulnerable to unintentional selections of candidates. Prior research has shown that nearly half of voters do not notice such anomalies, which may affect the outcome of an election [2,3]. Subsequently, the results of this research suggest that in addition to the design recommendations provided by the EAC, recommendations are needed that focus on the placement of controls and related text and objects, as they may influence how voters interact with the electronic ballot interface presented to them.
The use of touchscreen technology is growing in many domains (e.g., voting, finance, healthcare), including those where interface design is critical. Research that focuses on touch-based interface design and its influence on how users interact with an interface is also growing; however, only a small subset of this work focuses specifically on the design and proximity of buttons and other controls. Nevertheless, advocates argue for the importance of examining such characteristics because of the impact they may have on various user groups . The principles derived from this research can therefore be applied to other touchscreen systems.
This material is based upon work supported by the U.S. Election Assistance Commission (EAC). Opinions or points of view expressed in this document are those of the authors and do not necessarily reflect the official position of or a position that is endorsed by, EAC or the Federal government.
This material is based in part upon work supported by the National Science Foundation under Grant Numbers IIS-0738175. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation.
2. Campbell, B.A. and Byrne, M.D. Now do voters notice review screen anomalies? A look at voting system usability. Proc. of the 2009 Electronic Voting Technology Workshop/Workshop on Trustworthy Elections (EVT/WOTE '09).
4. Doyle, P. Judges rule Franken winner; Coleman to appeal. April 5, 2009; http://www.startribune.com/politics/national/senate/42932907.html
5. Sesto, M., Irwin, C., Chen, K., Amrish, C., and Douglas, W. Effect of touch screen button size and spacing on touch characteristics of users with and without disabilities. The Journal of Human Factors and Ergonomics Society 54, 3 (2012), 425436.
Juan E. Gilbert is the IDEaS Professor and chair of the Human-Centered Computing Division in the School of Computing at Clemson University. He is the principal investigator on the U.S. Election Assistance Commission Accessible Voting Technology Initiative and the lead for the Research for Alliance for Accessible Voting (RAAV).
Aqueasha M. Martin is a Ph.D. candidate in the School of Computing at Clemson University. Her research interests include human-computer interaction, health informatics, and information retrieval. She is interested in the ways that technology can be designed and used to promote and inform healthy living.
Gregory Rogers is a lead ETL developer for Chevron. In this position, he designs, builds, tests, and maintains databases. He graduated from Auburn University with a master's degree in computer science and a bachelor's degree in software engineering. His interests include spoken language systems, usability, multimodal interfaces, human-computer interaction, and databases.
Jerome McClendon is a Ph.D. candidate in the School of Computing at Clemson University, where he is a member of the Cyber Innovations Lab. He received a B.S. and M.S. in computer science from Auburn University. His interests include spoken-language systems, machine learning, and natural language processing.
Joshua Ekandem is pursuing a Ph.D. in human-centered computing in the School of Computing at Clemson University. Joshua's research interests include the design and development of natural user interfaces, especially in automotive and mobile contexts. He is interested in achieving natural user interaction through multimodal interfaces and culturally situated design frameworks.
©2012 ACM 1072-5220/12/11 $15.00
Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, to republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee.
The Digital Library is published by the Association for Computing Machinery. Copyright © 2012 ACM, Inc.