rent paperless DREs. Many states have already done so, but many states have not. All voters who go to the polls in Maryland and Georgia are forced to use paperless DREs, as are many voters in other states. Some other states are using paper ballots now, but could decide to convert to paperless e-voting in the future. Without federal legislation, voters in some states will be stuck with DREs for a long time.

Congress should mandate a specific class of paper trails: every voter should mark and cast a voter-verified paper ballot (VVPB). Each ballot can be counted by hand or scanned in the precincts by a scanner that checks it for overvotes or stray marks (the technical term for this type of system is precinct-count optical scan or PCOS). If there is a problem, the voter has a chance to fix the ballot or fill out a new one. Otherwise, the ballot is counted and deposited in secure ballot box. Or ballots can be counted by hand if desired. These systems can be made accessible to voters with a wide range of disabilities through the use of bal-lot-marking devices, which allow paper ballots to be read, marked, and verified via an accessible electronic interface.

Most studies have shown that PCOS systems are at least as accurate as any other voting system. They are less costly than touchscreen machines, and, if they fail, marked ballots can be stored in a ballot box and counted later. Most importantly, the hand-marked ballots can be verified and counted without having to trust computerized systems. Optical scan systems are already the dominant technology in the U.S.—they have been used for many years and the

some have argued
that legislation
requiring paper
ballots would hamper
innovation in voting
technology. But the
main problem in
voting technology
is not a lack of
innovation, but how to
prevent and recover
from bad innovations.

technology is steadily improving.

Why paper and not some other permanent medium such as recordable compact discs? Paper can be read and written by people or machines, and, importantly, by (almost) everyone without machine assistance. Votes on paper cannot be removed or changed without detection. Critical documents on paper have been handled for many centuries and the procedures are easily understood by poll workers and election administrators. For example, it is easily recognized as a problem if a poll worker disappears into a back room for a few hours with a box of ballots.

Of course, simply using paper ballots does not guarantee election in-

tegrity. The ballots must be protected, and the processes for storing, transporting, handling, and counting them must transparent. Crucially, paper ballots enable the routine auditing of elections by choosing ballots from randomly selected precincts or machines and manually counting them to see if they match the machine totals.

Some have argued that legislation requiring paper ballots would hamper innovation in voting technology. But the main problem in voting technology is not a lack of innovation, but how to prevent and recover from bad innovations. State and local governments chose to purchase tens of thousands of DREs in spite of the dire warnings of computer technologists and activists— then the true risks of DREs turned out to be even worse than the warnings. The existing requirements and certification process did little to protect the voting system from this and other bad ideas.

A federal VVPB mandate would channel vendor R&D efforts into improving optical scan technology, instead of developing and marketing lucrative but ultimately dubious systems like DREs. If and when a radically new technology is proposed, the law can be changed— after a thorough debate about the true benefits, costs, and risks of that new technology—a debate that would have averted the disastrous experiment with DREs over the last few years.

 

David L. Dill ( dill@cs.stanford.edu) is a professor of computer science and electrical engineering at Stanford University and has been working actively on policy issues in voting technology since 2003.

Counterpoint: Daniel Castro

ALL VOTERS WANT and deserve secure elections; unfortunately, no voting system currently on the market offers voters verifiable proof that their ballot has been counted. Some activists have been especially concerned about the integrity of votes cast on direct recording electronic (DRE) voting systems, since these devices rely on software that can be difficult to au-

dit. While most individuals agree that voting technology should be improved, many people disagree on the best way to improve it. In particular, a vocal group of activists have popularized the idea of using paper audit trails—basically a paper receipt produced by the DRE— as a countermeasure to fraud and error. Unfortunately, this proposal is an incomplete solution to a much larger problem. Moreover, improving voting systems is not merely a technical challenge but also a public policy challenge.

Computer science is an academic discipline that is based in logic and proof, and we should rely on these valuable methods in our analysis of voting system technology. To understand the scope of the problem one must first understand that the voting process does not end at the ballot box; to have secure elections every step of the voting process must be secure. Specifically, ballots must be cast as intended, collected as cast, and counted as collected. Paper audit trails only provide verification of