Abstract

Although government officials are well aware of the fact that ballot formatting can play an important role in election outcomes, there has only rarely been any enforcement of good design principles in voting system displays.  Following the November 2000 butterfly ballot fiasco, it was hoped that new election layout standards would be developed, but the new “quick fix” computerized solutions adopted by many communities often have turned out to be less user-friendly than some of the systems they replaced.   Furthermore, many of the checks and balances that allowed voters to verify their ballots are being removed.  Trends indicate that elections may be increasingly controlled by a techno-savvy elite, while some classes of voters are experiencing techno-disenfranchisement.  This paper addresses the serious dangers of Internet and kiosk-style computerized voting, focusing on the user interface issues.

Introduction

Although much ado was made about the now infamous butterfly ballot used in certain south Florida precincts during the November 2000 presidential election, many officials and legislators would like the public to believe that this was a problematic artifact unique to punched card voting systems.  Some voting machine vendors have been quick to suggest computerized solutions, offering promises of ballot entry and tabulation security, user-friendly displays, and improved reliability.  Some new products, though, have already failed in early use, resulting in costly litigation as well as further erosion of voter confidence in the election process.  From a usability standpoint, it is easy to see why these systems have fallen far short of expectations.  Since poor ballot design transcends technology, it is certainly possible to present a misleading user interface no matter what devices are used in its creation.

Nevertheless, the focus in the last two years has been on providing a panacea of increased computerization in elections in order to combat the erroneous assumption that paper-based elections are inherently troublesome.  Many states have introduced new election legislation, the Federal Election Commission has extensively overhauled their suggested voting system standards, and the U.S. House and Senate have enacted voting rights bills, but none of these changes have effectively or even substantially addressed the ballot design issues at the root of many voting system problems.

Accuracy and Auditability

One of the dirty secrets of the election business revealed by the 2000 Presidential election was that every vote is not counted.  Ballots that are voided (because they were prepared improperly, damaged, or otherwise unreadable) are discarded from the totals, effectively disenfranchising the voters who cast them, without their knowledge or consent.  Officials have known for years that the “lost vote” rate across the gamut of system types can range from two to five percent of all ballots cast, enough to affect the outcome of many election races.[VTP01]  Yet vendors and administrators have been allowed to hide behind considerably smaller miscount numbers formulated using pristine data sets, not the typical ballots produced by real voters on election day.  Presently, when an election is deemed to be “close” (and there is often no real definition of what that means), typically a tedious recount process occurs, usually under court order.  Instead, voter error rates should be used to provide a margin of error (similar to that used in statistical polling or scientific studies) and election laws should require the declaration of a “tie” with an additional runoff contest scheduled to determine the winner.

Equally important as accuracy in establishing election vote totals, is the necessity that voters believe that their choices have been recorded and tabulated properly.  Currently, this assurance is usually provided by bi-partisan pollworkers who oversee the precincts on election day.  Increasingly, though, the newer balloting devices (those built as stand-alone kiosks and Internet-based systems) only provide transient, intangible feedback for the voter’s selections.  Ballot images may be generated on paper, after the election, from memory storage units, to serve as an “audit trail.”  This self-checking process was recently described in a memorandum from California Secretary of State Bill Jones, where it was stated (for a Sequoia Edge system under scrutiny) that: “The counting is performed in full public view. ... Also, anyone has the right to observe the counting process.  The counting is so fast that it is not detected by human eye, but at any given time the ballots can be reproduced and printed on paper at the request of the election official...”[JON02]  The sheer ludicrous nature of this statement is underscored by the fact that years ago, Roy Saltman warned against such systems, saying that “the voter is given some reason to believe that the desired choices have been entered correctly into the temporary storage, but no independent proof can be provided to the voter that the choices have, in fact, been entered correctly for the purpose of summarizing these choices with all others to produce vote totals.”[SAL88]

Vendors and purchasers of fully electronic voting systems have been reluctant to provide paper printouts of ballot selections (to be deposited at the polling place and used as a check against the electronically recorded totals), with the excuse that this would unduly increase expenses and time.  At the same time, they insist that the interface will prevent under or over-voting (missed or too many candidate selections).  As it turns out, the need for additional feedback (beyond the electronic ballot face) was demonstrated in an experiment that showed approximately 3% of voters noticing errors in touchscreen choices after subsequently viewing their selections on paper.  This is consistent with the three percent undervote rate discovered in a March 26, 2002 Palm Beach County, Florida election, where only two candidates were on the ballot and “no vote” selections would have been unlikely.[ALE02]

Voter Perceptions

The perception of the voter does play a role in elections.  The sense that one’s vote matters can affect election turnout, which in turn can affect the legitimacy of the representative government.  In addition to the belief that votes are recorded and tabulated accurately, Clifford Nass [NAS02] suggests that the voter should also be able to:

• perform the mechanics of voting properly
• access all permitted information
• be prevented from or warned about possible mistakes
• correct or change mistakes
• use the voting system without any externally imposed bias

Off-site or Internet voting creates additional problems, in part because of the insecurity of the technology, but much more due to sociological factors.  From a convenience standpoint, this voting option seems attractive, but issues related to voter identification and privacy far outweigh any merits that can be provided.  Opportunities for coercion and monitoring of voters increase dramatically.  PIN numbers can be stolen, fraudulated, or even sold.  Some have suggested the use of bioidentification, but others worry that this will disenfranchise certain population groups who fear that the information stored for voting would be used against them in other scenarios.  Large scale deployment of off-site balloting will invariably increase opportunities for the control of elections by political machines, military, religious and labor organizations, and other special interest groups.  Rather than encouraging greater voter participation, it is more likely that the net result of this technology will yield further disillusionment of the public and devaluation of the democratic process in elections.

Usability

At present, no uniform standards exist regarding the usability of voting systems, so states and municipalities are left to devise their own criteria for evaluation.  Usability concerns address ballot presentation [SHE01] and diverse population [ROT98] issues as follows below.

In the area of ballot presentation, guidelines should be used to determine:

• font size
• font type
• text versus background color
• light level at polling station
• display height
• ballot layout
• overvote and undervote warning

• advancement between ballot pages
• position feedback and navigation
• input mechanisms
• recount capability
• malfunction alerts

• age
• accessibility
• training and usage instructions
• literacy (computer and written)

Usability testing must not be overlooked as an essential companion for validation of usability design.[QUE01]  Typically, this has been performed in an ad-hoc fashion, but formal processes should be encouraged.  In order to be effective, though, an appropriate population must test the ballot configuration prior to the election.  Robert Bailey [BAI00] calculated that for the Palm Beach County butterfly ballot, nearly 300 participants would have been required in order to detect 95% of the problems, and over 400 for a 99% assurance.  This is far more than the couple-dozen subjects (or less) that are usually surveyed.  For that county alone, Bailey estimated that the testing cost would have been around $20,000, so price becomes a factor as well.  The need for usability testing is ongoing, since ballot displays are customized for each election in every municipality.

Conclusions

Humanizing voting interfaces is a complex matter that extends beyond the actual ballot presentation, to the underlying perceptions regarding the entire election mechanism.  Checks, balances, and assurance methodologies must be applied throughout the process, in order to achieve fairness for all participants.  The increased attention to balloting issues that occurred following the November 2000 election should encourage communities and election officials to allocate time and funding toward the development and adoption of voting interface standards and usability testing practices.  Ongoing vigilance must be applied, lest new products and designs add additional problems to those they are hoping to solve.

Acknowledgement

Sincere thanks to Whitney Quesenbery for bringing my work to the attention of the usability community and for her efforts in promoting improvements in voting system design.

References

[ALE02] Alexander, Kim, Declaration of Kim Alexander, U.S. District Court for the Central District of California, Western Division, Case No. CV 01-11159 SVW (RZx), May 6, 2002.

[BAI00] Bailey, Robert W., The Usability of Punched Ballots, UI Design Update Newsletter, Human Factors International, November 2000.

[JON02] Jones, Bill, Defendant Bill Jones’ Memorandum of Points and Authorities in Support of His Motion for Summary Judgment, U.S. District Court for the Central District of California, Western Division, Case No. CV 01-11159 SVW (RZx), May 6, 2002.

[NAS02] Nass, Clifford, Perceptions of the Voting Process: Measurement and Relationship to Interface Design, Workshop on Election Standards and Technology, American Association for the Advancement of Science, January 2002.

[QUE01] Quesenbery, Whitney, Voting and Usability: Lessons from the 2000 Presidential Election, STC-PMC News and Views, November 2001.

[ROT98] Roth, Susan King, Disenfranchised by design: voting systems and the election process, Information Design Journal, vol. 9, no. 1, 1998.

[SAL88] Saltman, Roy G., Accuracy, Integrity, and Security in Computerized Vote-Tallying, U.S. Department of Commerce, National Bureau of Standards, NBS (now NIST) Special Publication 500-158, August 1988.

[SHE01] Sherman, Robert M, Usability Standards for Direct Recording Electronic Voting Systems, Undergraduate Thesis, University of Maryland, December 17, 2001.

[VTP01] Voting Technology Project, Caltech/MIT, Voting, What Is, What Could Be, July 2001.