Provisions for the inclusion of a physical record, in the form of hand- or machine-marked ballots, or as a Voter Verified Paper Audit Trail (VVPAT), are central to guaranteeing safe and secure elections. However, the processing of such records during the initial counting of votes or in the conduct of audits has raised its own set of problems which span broad technical and social boundaries.
The aim of this project is to study issues that currently make paper records more of a nuisance than an integral component in trustworthy voting systems. Specifically, the principal investigators are working to characterize the statistical distribution of mark sense errors as a function of ballot layout and quality in optical scanning, to examine approaches for unbiased visual auditing based on ballot images, to investigate the possibility that a concept known as homogeneous class display (HCD) can facilitate manual recounts, and to evaluate recognition errors that may arise in processing the VVPAT used with Direct Recording Electronic (DRE) systems. They are also interested in the effects these issues have on procedures for testing the paper handling components of voting systems in accordance with operational constraints, including the modest training received by most poll workers. This work on voting technologies is supported by ? and supports ? a planned survey and focus groups they are conducting to measure voter confidence and acceptance and to identify common misconceptions and concerns, including accessibility to disabled voters. Beyond its broad impact on the development of more reliable and trustworthy voting technologies, this project more generally has implications for the highly accurate computer processing of any information encoded in human readable form.