[an error occurred while processing this directive]

REVISITING THE U.S. VOTING SYSTEM: A RESEARCH INVENTORY

November 27-28, 2006

Convened by the American Association for the Advancement of Science

Scott Robertson

HUMAN-COMPUTER INTERACTION

Usability is critical to all electronic systems that involve interactions with people. Human-Computer Interaction (HCI), the science of understanding how people are going to use a system and the practice of designing in correspondence with that understanding, is especially critical to voting technologies. If voters cannot comprehend what is on a ballot, if they cannot easily remember or recognize their choices, if they have difficulty carrying out the actions required to cast a vote, if they make mistakes, and if they lack trust and confidence in the voting apparatus, then there is little point in insuring the security, privacy, and accountability of their likely erroneous choices. These human issues vary considerably across different voter groups, and so it is critical to conduct usability studies across broad samples to inform the ballot design process.

Usability standards for many issues such as legibility and comprehensibility currently exist to help in the design of ballots. Unfortunately, it is often difficult to achieve good usability even with standards. Electronic voting systems introduce a host of new issues for voters, such as navigability, error checking, and trust in vote recording. Electronic voting systems have also introduced significant usability problems involving other stakeholders such as election officials making technology choices and poll workers who aid voters at election time. It is critical that guidelines for usability and usability testing be developed for electronic systems for all stakeholders at every point in the process including designing the interface, deploying machines, aiding voters, collecting votes, and evaluating performance after elections.

User-centered design is the practice of involving users and users' perspectives in an iterative development process. Guidelines for electronic ballots should include the collection of behavioral data from multiple user and stakeholder groups, and standards should include specific behavioral targets. A system which fails to meet behavioral standards, or which has not been user tested, should be treated the same way as a system that fails to meet security standards or has not undergone security testing, i.e. it should not be fielded. This requirement would have to be met anew for each ballot with significant differences. To facilitate this process, ballot construction and evaluation tool kits should be developed that can be used by local election officials to produce and test various electronic ballot designs.

Electronic voting machines are entering the domain of other electronic consumer products and this introduces at least three important and novel issues:

1) Electronic ballots will exist within a broad context of other electronic information sources and this context should not be ignored during the design process.

2) Electronic ballot design will have to become an iterative process that involves rapid prototyping and that is informed by ongoing user evaluation.

3) An electronic ballot has the potential to introduce new features that are not traditionally part of the ballot. Examples include online help, the possibility of offering different "views" of ballot information (e.g. toggling between a summary and the complete text of a proposition), interaction with completed sample ballots on other devices (e.g. downloading preselected votes from a PDA or cell phone), or dynamic changes in fonts, languages, or other display features. While the initial reaction to such issues is often that they do not belong as part of voting, election designers and planners should recognize that electronic systems will become quickly antiquated and obsolete (both technically and in terms of interaction features) and that redesign cycles could become as frequent as election cycles.

My research involves the first issue in the list above, specifically the context of voting. Research questions include detailed observation of how people gather information from electronic sources, how they share that information, and how to design a political portal that informs, encourages participation and deliberation, is trusted, and integrates well with voting technologies.