Objectives
We are building software tools for reliable paper ballot processing as well as developing frameworks for large-scale ballot studies, for examining usability issues involving paper ballots, and for performing reliable and trustworthy recounts of hardcopy election records.


Goals
Our work includes developing and testing software solutions for:

• defining ballot layouts;
• processing voted ballots;
• conducting recounts;
• confirming the accuracy of machine ballot reading;
• supporting large-scale studies of paper ballots and their usability.

Issues
Some of the issues we are attempting to address include:

• Undetected failures in the machine reading of ballots. Optical scan systems do not normally warn users when recognition errors arise, and processing the ballot a second time may lead to a different result.
• Systematic errors due to ballot layout. Our past work has demonstrated that recognition errors are not uniformly distributed across the page. This could be true in the case of ballots, which may disadvantage a candidate based purely on where his/her name appears on the ballot sheet.
• High cost of manual recounts. Recounting all of the ballots in a large geographic area can be expensive, both in time and money.
• Human error and human bias in performing audits and recounts. Human ballot readers may be more trustworthy than machines, but they also bring with them personal biases which may intentionally or unintentionally alter the outcome of an election.
• “Hackers” attempting to manipulate the vote. This fear is the driving force behind the push toward paper ballots, but it should be noted that the electronics of optical scan systems have been proven to be just as vulnerable as DRE systems.
• Traditional ballot-box stuffing. While there is no such thing as a perfectly secure voting system, some approaches are safer than others. Low-tech approaches have undoubtedly resulted in the theft of more elections throughout history than the current cyberthreats that now receive so much media attention.
• Preserving anonymity. Many solutions that come to mind for securing and processing paper ballots place the anonymity of voters at risk. For example, current approaches for providing a Voter Verified Paper Audit Trail (VVPAT) maintain their audit trails sequentially, making it possible to recover a voter’s votes by noting the order in which people use the machine and later obtaining a copy of the paper tape. Likewise, schemes for pre-printing a unique ID on each ballot also fall under suspicion.
  
• Voter error. The range of individuals who vote in a country’s elections (and their abilities) reflects a broad spectrum of humanity. Some voting technologies are more likely to induce errors than others.
• Interpretation of marginal markings. Two different ballot readers – humans and/or machines – may interpret the same marking differently. Such markings are called “marginal.” Legislation is written in terms of voter intent, however; markings that appear to reflect a voter’s desires should not be disqualified for purely technical reasons.
• Testing and certification of electronic voting systems. Current federal and state government evaluations of electronic voting systems are rudimentary at best. These shortcomings were dramatically demonstrated when California and Ohio contracted with independent security consultants to test voting systems used in their states, only to find numerous serious security holes that had passed the original certification process.