At the center of the controversy surrounding the U.S. presidential election 2000 are approximately 40,000 so-called "undervotes" cast in the state of Florida on November 7. These ballots, for the most part punch card ballots, showed no vote for president when subjected to a machine count. Manual inspection of these votes, however, allows for the recognition of a voter's impact on some of these ballots. The manifestation of such impact are displaced or deformed chads known as hanging, dangling, indented, pregnant, dimpled, and other types of chads.

It is already clear that these undervotes will become part of American history and, therefore, should be preserved as historical evidence of great interest.

The undervote ballots are delicate documents and require diligent treatment. Each handling of these ballots, e.g., through manual inspection or during transportation, may change the chad's configuration or may dislodge a chad. Even storing the ballots in boxes may flatten indented, pregnant, and dimpled chads over time. Changes in temperature and humidity as well as other environmental effects will further degrade the current configuration of these ballots. There seems to be no way to preserve these ballots in their original, physical state for the future.

Therefore, we propose an effort that would preserve these undervotes in their current condition using state-of-the-art information technology by creating a three-dimensional computer representation of each of the 40,000 ballots. Although this seems to be an enormous number, the 40,000 items of three-dimensional geometry is a rather small dataset if compared to typical engineering representations with several hundred thousand geometry items of usually higher complexity. Since a large number of the undervotes show no voter impact at all in a specific region of the ballot (no vote for president), the effort could concentrate on the remaining, questionable ballots estimated to be just a few thousand (less than 10% of all undervotes).

Three-dimensional laser scanning devices will be used for the precise capture of a ballot's geometry (including the print). Each ballot can be scanned in seconds and without contacting the ballot. The accuracy of modern scanners is sufficient to capture the displacement or deformation of any chad in sub-millimeter detail.

Once the scanned data have been processed, each ballot can be viewed on a computer monitor as a three-dimensional virtual object from any direction. The object can be rotated, enlarged, or viewed against a virtual light. The viewer may even 'fly' through a partially punched-out opening in the ballot and inspect the way the chad has been displaced in detail. Stereoscopic viewing via LCD shutter glasses can be deployed to significantly enhance the perception of any three-dimensional feature of interest. Most importantly, researchers and scientists can develop algorithms that automatically inspect and classify each ballot according to a variety of criteria using three-dimensional measuring and pattern recognition techniques.

The digital preservation of Florida's undervotes requires the testing and selection of scanning equipment, the development of an appropriate and error-free scanning process, as well as overall guidelines. Experts in history, preservation, and constitutional law should be involved in the creation of this unique three-dimensional dataset. The process of scanning the undervotes may require a few days or weeks.

The following summarizes the benefits of the proposed effort:

• The virtual ballots would preserve the current state of the physical undervotes as a historical document for future generations.
  
  
• The dataset will become an invaluable document for future scholars that study the 2000 presidential election in the context of American history.
  
  
• The dataset itself is an objective document and does not imply the legality or illegality of the votes nor does it suggest any criteria for counting these votes.
  
  
• The dataset can be inspected repeatedly without causing damage by an unlimited number of scholars, researchers, and other interested individuals.
  
  
• Any research or inside gained from inspecting these digital ballots can be verified by others. Inspecting the physical ballots can never be reproduced since each handling will change the ballot's condition.
  
  
• The virtual ballots can be of significant help in the development of new standards for vote casting and vote counting in future elections.
  
  
• The digital representation of the undervotes allows for inspection and classification by automatic algorithms using any type of criteria.
  
  
• The dataset can be made available over the Internet for schools and universities, for the media, for many interested organizations, as well as for millions of citizens.
  
  
• The cost of this preservation effort, although still to be determined, will be reasonable and significantly lower than the cost of other preservation projects of similar importance.
  
  

University of Michigan 2600 Draper Road, #225 NA&ME Bldg. Ann Arbor, MI 48109-2145

Phone: (734) 764-4296 Fax: (734) 936-8820 URL: www-VRL.umich.edu