Abstract
Objective
To describe user-centered voting systems that would support the safe conduct of voting in a pandemic environment.
Background
The COVID-19 pandemic has complicated our democratic processes. Voters and poll workers feel threatened by the potential dangers of voting in business-as-usual polling stations. Indeed, significant problems were encountered in the recent 2020 primary elections in Wisconsin, where the National Guard had to be mobilized because so few poll workers reported to work, and more than 90% of polling places had to remain closed.
Method
We describe a number of possible user-centered solutions that would help protect voters and poll workers in times of pandemic, and also report the results of a survey that asked voters and poll workers about what kinds of systems might make them willing to vote.
Results
Political as well as safety considerations will need to be considered as these safer voting solutions are designed since, surprisingly, the kinds of solutions preferred depend on the political affiliation of the voters.
Conclusion
Human factors professionals have a large role to play in realizing the safe, successful implementation of these user-centered systems. Good human factors analysis can help minimize the risk to voters and poll workers. Moreover, human factors methods can help safeguard democracy by creating safe and well-engineered environments that are conducive to voting in the age of pandemics.
Application
Creating safe and effective voting solutions that protect voters and poll workers during pandemic outbreaks is crucial to the preservation of democracy.
Introduction
Voting has always been an inherently social activity. In early United States elections, men would gather around an election official and yell out their vote for all to hear (Dinkin, 1982). In more recent years, the United States has moved toward a secret ballot, where each voter is alone in the booth and can vote their conscience (Crook & Crook, 2011; Rusk, 1970). However, it’s still a social activity. Our polling stations are in our local schools and churches, and we see our friends and neighbors there. As we leave, we collect an “I voted” sticker to proclaim that we are responsible, democratic citizens who participate in elections.
The recent COVID-19 pandemic is causing us to rethink how we conduct elections in times of infectious disease. One only needs to look at the recent debacle in Wisconsin to understand why a swift and reasoned solution to voting is required if we are going to be able to preserve democracy in such perilous times.
Recall, at the height of the pandemic in the spring of 2020, Wisconsin was holding their primary election. Cases of COVID-19 were steadily rising in the state and despite pleas from voting groups to delay the election, the election went ahead as planned. Poll workers voted with their feet, and thousands of them refused to report to potentially dangerous polling stations. In fact, the National Guard had to be deployed in order to maintain the bare minimum of polling stations in the state (Trevato, 2020). In Milwaukee, 95% of the polling stations were closed due to lack of available poll workers (McCormack, 2020). As can be seen in Figure 1, voters and poll workers did the best they could to protect themselves during the election.
Voters and poll workers try to protect themselves during the 2020 Wisconsin primary election (Copyright 2020 Getty Images, used with permission).
Potential Solutions
Vote by Mail
What is the solution to maintaining democracy when going to the polls could be dangerous and even life-threatening because of a pandemic? A human-systems-integration approach might suggest an obvious solution: mail-in ballots. Voters do not have to leave the relative safety of their homes to procure or cast their ballot. The system is well documented, as five states currently use mail-in ballots almost exclusively and have done so for years. Further, all states have some form of mail-in ballot system operational for absentee voting, albeit with some limitations on who can cast a mail ballot (e.g., persons over the age of 65). Despite the claims from some that mail-in voting favors one party or the other, the data do not support this claim (Barreto et al., 2006). Unfortunately, there are several barriers to implementing mail-in ballots on a national scale.
The first of these considerations is political. In general, Republican leaders (with the exception of Utah and Colorado) are opposed to mail-in ballots, even though almost a quarter of the votes cast in the 2018 elections were mail-in (United States Census Bureau, 2019). The more serious problem, however, is that ramping up a system of mail-in ballots in such a short period of time might not be technically and bureaucratically feasible. The key to mail-in ballots is to ensure that the right ballot is sent to the right person at the right address. However, since many states have never used mail-in ballots at scale, their voter registration information is likely out of date, because voters have moved or passed away. These are all solvable problems, but not necessarily in the few short months between now and the upcoming 2020 and 2021 elections. Vote-by-mail also has a number of human factors issues to consider as well. For example, instructions would need to be exceptionally clear, as many voters have never voted in this manner, and no election official is there to assist the voter. Solutions such as automated and live help lines would need to be considered and designed as part of any widespread implementation of vote-by-mail.
Safer Polling Station Practices
As with any difficult problem, it is likely that a number of solutions will need to be simultaneously implemented in order to make sure that all the different types of voters and their needs are taken into account. One option is to simply do what we do now, but in a safer manner. Utilize mail-in ballots for elderly and absentee voters, as has been done extensively in many states, while simultaneously employing the best practices that have emerged in the retail sector since the pandemic began. These kinds of best practices are focused on improving the protection afforded to the voter and poll worker alike. These enhanced protections include limiting the number of voters that can come into a polling place at any given time, ensuring that voters are spaced at least 6 feet apart (Centers for Disease Control and Prevention, 2020a) as they queue to vote, making certain that all voters wear face masks and gloves, employing plexiglass guards between voters and poll workers, and following strict sanitation processes for the voting equipment between voters.
A cursory examination of voting behavior in Wisconsin highlights challenges to achieving safe voting practices. Scores of voters were not wearing any type of personal protective equipment, and because of the significant reduction in the number of polling places, spacing was difficult to maintain over the hours that people had to wait to vote (Figure 2).
Unprotected and not appropriately socially distanced voters wait in long lines to cast their ballots in the 2020 Wisconsin primary election (Copyright 2020 Getty Images, used with permission).
Processes are often one of the first things to fail in high-stress, time-sensitive systems. This is why it is important to design systems in ways that have forcing functions that encourage or require compliance with the desired behaviors. For example, in normal queueing situations, people are generally accustomed to standing 15–30 inches apart in the line (Kaya & Erkip, 1999; Nakauchi & Simmons, 2002). However, the Centers for Disease Control and Prevention (CDC) is now asking them to stand at least 6 feet apart, and queueing areas that lack visible 6-feet markings do not provide enough information to voters so they know where to stand, which can lead to undesirable crowding. High affordance designs (e.g., marks on sidewalks, line barriers) and passive protections (e.g., plexiglass barriers and other guards) would need to be employed for maximum effectiveness. Environmental considerations such as these have been shown to have a positive effect on voters’ subjective assessments of the voting environment (Acemyan & Kortum, 2018).
The goal of any voting system deployed during a pandemic is to minimize voters’ exposure to the hazardous virus. There are other user-centered systems that might accomplish this goal better than a traditional polling place where social distancing is practiced and environmental changes (e.g., plexiglass guards) are implemented.
Outdoor Voting
Since some of the potential danger associated with in-person voting is having a large number of people gather in a confined space, outdoor voting is a system that would help mitigate this issue. This would allow for maximum air exchange where voters are casting their ballots, and the outdoor space could provide greater flexibility in maintaining spacing requirements. Human factors professionals would need to ensure that systems implemented to manage voters in this unique voting solution (environmental human factors) and instruct them on required actions (training human factors) complied with the best science. Of course, such a design would be subject to environmental conditions that might make it difficult to implement. For instance, pictures emerging from Wisconsin show voters huddled under umbrellas in long queues waiting to enter the polling station (New York Times, 2020).
Drive-Through Voting
Drive-through (curbside) voting is another potential option. It has been traditionally employed only in the rarest circumstances, such as when a voter is unable to exit their car due to a disability. However, it might be a viable method for large-scale voting with minimal contact.
In a drive-through voting system, voters would arrive at the drive-through polling station and show their credentials through their closed car window. Once checked-in, a poll worker would use an extension device to hand the voter a paper ballot while standing at a distance. The voter would then drive their car into a “voting” parking space so that they could take as long as they needed to complete their ballot. Once the voter completed their ballot, they would drive to the exit and deposit it into a secure ballot box under the watchful eye of a socially distanced poll worker. Figure 3 shows a sample set-up for this system. Similar systems have been deployed by restaurants that have switched exclusively to take-out food during the COVID-19 pandemic even though they were primarily dine-in establishments before.
A drive-through voting system example proposed by the authors; a setup like this would minimize wait times for voters, while maximizing voter throughput.
Ballot Drop-Off Voting
Another potential solution combines some of the elements of the drive-through system with other elements of the vote-by-mail system. In this voting system, voters could download a ballot and print it out at home, or request and receive a ballot in the mail from the county clerk. After completing their ballot in the comfort and safety of their home, they would proceed to a secure ballot drop-off location in their neighborhood. The drop-off location would resemble the drive-through voting system, except that the voter would only check in and drop off their ballot. No provisions would be made for distributing or marking ballots at a drop-off location. This retains the advantage of minimal contact between voters and poll workers, while significantly reducing the complexity of having to hand ballots to voters and ensuring sufficient space for voters to mark their ballots.
Internet Voting
Of course, the holy grail of voting, and one that is frequently discussed as a potential solution, is internet voting. To the average voter, Internet voting seems like the simplest solution. However, there are well-known technology access issues among the elderly and those associated with low socioeconomic status. That said, 80% of U.S. citizens have a smartphone (Comscore, 2017), and research has demonstrated that smartphone voting interfaces are usable and effective (Campbell et al., 2011, 2014).
Voters recognize that they already complete high-security transactions using the internet (e.g., banking, taxes, e-commerce) and so they see little reason why an activity like voting should not be able to take place on the Internet. Unfortunately, in the examples given above, the identity of the person is known and verified by the institution, and all transactions are traced back to that identity. In voting, because of the secret ballot requirement, a voter’s identity cannot be known once they have entered the voting stage. This makes Internet voting extremely complex from a security standpoint. While internet voting has been implemented successfully in several countries, such as Switzerland (Germann & Serdült, 2017) and Estonia (Vassil et al., 2016), and new internet voting security schemes are constantly being proposed (e.g., Kumar et al., 2020; Shakiba et al., 2017), some of the leading security researchers in the United States believe that it may be nearly impossible to field a fully secure and auditable internet voting system (Park et al., 2020).
There are a number of highly sophisticated security mechanisms that might be employed to make Internet voting possible, but given the time frame, it would require a nearly unprecedented level of engagement by the federal government and security experts (on the level of the Manhattan Project, which was responsible for the research and development of the atomic bomb in World War II) in order to make that happen. Further, from a human factor’s standpoint, there would be significant work required to ensure that voters understood how to utilize these new, unfamiliar security processes required to vote. There is strong evidence that voters do not understand high-security protocols well (Acemyan et al., 2014, 2015a, 2015b; Marky et al., 2018), leading to significant disenfranchisement if voters are unable to successfully cast their ballots.
In the most egregious examples from this user research into high-security voting systems, voters believed they had cast their ballot, when in fact they had not, so the failures of the system would have gone undetected in a real election. Security experts and human factor specialists would have to work hand-in-hand from the first line of code to the implementation of the interface to ensure that both security and usability expectations were met (Oltramari et al., 2015). That said, there are examples of highly secure computer-based interfaces that have demonstrated good usability (e.g., Acemyan, Kortum, Byrne et al., 2018).
The internet voting option also has important implications for future voting protocols, since whatever infrastructure is created could be easily reused, unlike some of the ad-hoc virus protection mechanisms described earlier. Aside from the (solvable) technical difficulties, a federally implemented internet voting system might also run into difficulties due to the constraints of the U.S. Constitution (Article I, § 4), which reserves the conduct of elections to the states, not the federal government.
The Role of Human Factors in Making Voting During a Pandemic Safer
In all of the different voting scenarios described above, human factors professionals can contribute to the final user-centered solutions. We can be instrumental in performing appropriate task analyses to determine how the new COVID requirements change the operation of the polling station. This might include running simulations, or performing GOMS (Goals, Operators, Methods, and Selection) analysis to help select between a number of different solutions. As each solution is developed, flow, link, and timeline analyses will need to be conducted so that we understand how voting patterns and wait times will be affected. This is critical to ensure that voting times remain reasonable, and allocation of voting resources can be properly made. Important variables such as wait times, voting times, polling station capacity, and poll worker workload should be as good or better than the standard voting systems they are supplementing or replacing.
We will have a large role in creating instructions so that voters will understand the voting options available to them and their various implications. Carefully developed instructions will also help voters to know what to do when they step inside a polling station with novel procedures and layouts that are now unfamiliar to them. Contingencies will need to be developed for disabled and elderly users who may not be able to conform to these new polling station designs. Poll workers (many of whom have worked for years in this capacity) will need to receive well-engineered training on how to function in this new environment and perform all the new procedures that are now required to keep themselves and the voters safe. Training needs to not only be effective, but also efficient, as voting officials may have to re-train hundreds or thousands of poll workers in relatively short periods of time. Human factors involvement will be essential in ensuring that all of these changes are user-centered and align with cognitive capabilities and limitations, especially within the constraints of tight timelines, when communities’ health and safety is on the line.
Once the preliminary polling station parameters are set (for whatever form of voting method is being designed), usability specialists will need to assess the systems to ensure that everyone can perform their appropriate roles effectively and efficiently. Mockups of potential voting configurations could be constructed and tested for usability. Effectiveness and efficiency measures for the voters should be equivalent to current voting methods to the greatest degree possible. If user-centered processes have been followed during the design phase, then usability and safety requirements should be able to be achieved.
Finally, risk management techniques and safety methods will also need to be employed to ensure that voting goals do not compromise safety and that safety does not compromise the ability of voters to cast their ballots. For example, if the voting technology being used requires skin-to-screen contact to work, then voters might have to remove their finger coverings to actually select their candidates, a violation of safety protocols in pursuit of the voting goal. Safety accommodations must also be compatible for voters who have hearing, vision, or mobility disabilities, so their right to vote is not compromised.
If states decide to move toward internet voting, human factors engineers/human computer-interaction scientists will need to design and develop new interfaces to support a wide range of users—from computer novices to experienced users. These systems will need to be rigorously tested to make certain that security and usability have been seamlessly integrated, and that users of all abilities can use, understand, and have confidence in these sophisticated new systems. These new interfaces should be as familiar to users as possible, so that voters can use their past voting experiences to guide them as they vote with these newly designed systems.
Throughout all the human factors activities, human factors professionals need to remember that their users are dedicated citizens who want to vote as simply and as safely as possible. These goals should guide our design and evaluation efforts as we advocate for the wide variety of users who will vote in the upcoming elections.
Voter and Poll Worker Preferences
In addition to identifying numerous potential voting methods that could be used during a pandemic, it is essential to understand which of these systems voters and poll workers find most palatable for the conduct of elections in time of widespread infectious disease outbreak. Any attempt to perform user-centered design must include at least some understanding of the potential users. Accordingly, we conducted an Institutional Review Board–approved poll of 1002 likely voters (45% democrat, 37% republican, 18% independent) and 1791 poll workers (49% democrat, 39% republican, 14% independent) in Harris County, Texas, whose county seat is the City of Houston, the fourth largest city in United States. Interviews were conducted over the telephone (voters) and using internet-based surveys (poll workers).
These data illustrate that some of the solutions proposed in this paper are strongly preferred over others (Figure 4). Among Democrats, mail-in ballots are the clear favorite. Republicans, on the other hand, are most comfortable with the current methods of in-person voting, so long as social distancing requirements are met. Persons over 65 (63%) strongly preferred voting by mail than persons under 65 (51%). A small majority (53%) of white voters strongly prefer voting by mail, while more than two-thirds (68%) of African-American voters strongly prefer voting by mail. Novel methods, such as drive-through voting and internet voting, have moderate Democratic support, but far less support among Republican voters. Drop-off ballots are not the preferred solution for any voters. For the most part, Independent voters tend to more closely mirror Democrats.
Percentage of voters, divided by political affiliation, as a function of preferred system type.
Notably, when voters are asked if they would participate in an election where a given method is used, there is nearly consistent support for all the methods (~70% support), suggesting that voters will vote with the methods that are provided to them at the time of the election even though they may have (partisan) preferences about those methods. This was amply demonstrated in Wisconsin, where voters turned out despite suboptimal and potentially dangerous voting conditions.
Poll workers also have clear expectations about what will create a safe environment for them on election day, with access to personal protective equipment at the top of the list (Figure 5). Surprisingly, even among poll workers, there appear to be differences in preferred solutions along party lines, with Republicans strongly preferring social distancing requirements, and Democrats preferring drive thru voting by nearly 3-to-1 over Republicans. Outdoor voting and one-person-at-a time voting are at the bottom of the preference list, likely reflecting the pragmatic side of poll workers who recognize the difficulties for these specific systems.
Percentage of poll workers across political parties as a function of preferred voting method.
Data from Wisconsin suggest that there are now documented cases of COVID infection among voters and poll workers tied directly to the conduct of the election (Stracqualursi & Phillip, 2020). As this issue becomes more well understood, the willingness of poll workers to place themselves at risk may diminish, and lead to greater poll worker shortages.
Conclusions
Although final decisions about which of these systems to implement should not be based solely on voter/poll worker preference, compliance and willingness to utilize systems is often predicated on user acceptance. Accordingly, voter and poll worker preferences should be strongly factored into any decisions. Given the very short time that states have to implement solutions, it is likely that the solutions and systems employed may, by necessity, be “all of the above,” an approach that the CDC is now advocating (Centers for Disease Control and Prevention, 2020b). Further, multiple solutions will need to be considered because of voter demographics. For example, not everyone owns a car, and that would exclude them from the drive-through voting solution.
Human factors professionals have a large role to play in realizing the safe, successful implementation of these user-centered systems and good human factors will be essential in helping to minimize the risk to voters and poll workers. How can you make a difference in your community? Call your local county clerk and volunteer your skills to support the upcoming elections. Get involved with civic organizations that are already involved in helping to run elections. Volunteer to be a poll work or election judge. Together, human factors professionals can help safeguard democracy by creating safe and well-engineered environments that are conducive to voting in the age of pandemics.
Key Points
The COVID-19 pandemic has altered the way we need to vote due to the hazards to which voters and poll workers may be exposed.
User-centered systems must be designed and/or deployed to meet the challenge.
Some system design considerations may be political rather than user-focused, so system solutions must align with that reality.
Multidisciplinary teams will need to work together to create well-engineered systems that conform to human system integration principles.
Footnotes
Acknowledgments
This research was funded in part by a grant from Rice University’s COVID-19 Research Fund.
Author Biographies
Philip Kortum is an associate professor in the Department of Psychological Sciences at Rice University. He received his PhD from the University of Texas at Austin in 1994.
Robert Stein is a professor in the Department of Political Science at Rice University. He received his PhD from the University of Wisconsin-Milwaukee in 1977.
Claudia Ziegler Acemyan is an adjunct assistant professor in the Department of Psychological Sciences at Rice University and the founding Principal of Post Hoc LLC, a human factors and safety consulting firm. She received her PhD from Rice University in 2014.
Dan S. Wallach is a professor in the Department of Computer Science at Rice University. He received his PhD from Princeton in 1999.
Elizabeth Vann is the Director of Programs and Partnerships for the Center for Civic Leadership at Rice University. She earned her PhD from the University of Virginia in 2003.