Letter To Kevin Shelly, Originator Of The Adhoc Task Force For DRE In California.

March 7, 2003
The Honorable Kevin Shelley,
Secretary of State
1500 11th St., 4th Fl
Sacramento, CA 95814

Dear Mr. Shelley,

It has come to my attention that you have called for an Ad Hoc touch screen task force to establish the value of adding a step of voter verified paper audits for Direct Record Electronic (DRE) voting machines.

As you may be aware I have been involved in studying voting for the last two years as part of the Caltech/MIT voting project. I am a contributing author in Voting What Is and What Could Be, which was published in July of 2001. This document reported the problems of the presidential elections in 2000 steamed from:

1. Registration problems (1 to 3 million lost votes),
2. User interface problems on the ballot itself (1 to 2 million lost votes),
3. Polling place operations (1 million lost votes), and
4. Absentee ballots (an unknown number problems).

I have been doing research and development in human computer interaction for 25 years. I have worked at Xerox PARC, Atari Research, and taught human computer interaction research at Stanford for more than 10 years. In my position at IBM Almaden research center in San Jose, California I was a technical executive (IBM Fellow) involved in emerging technology strategy and for many user interface products at IBM. These included special needs products, graphical interface products and physical interface products.

My students have worked with me to prototype voting solutions to demonstrate possibilities for

1. Using best of breed human computer interaction knowledge to layout ballots
2. Creating a voting system that uses multiple software agents at each stage in the system to check on each other and create electronic audit trails
3. Creating a simple inexpensive system for checking registration databases with online resources.

I have been working with academics, election officials, and other organizations concerned with voting for the past 2 years. I have been in a position to ask these people to help expose the human sides of computing problems of voting. I have been trying to generate interest within the professional community of human computer interaction researchers for working on voting. I chaired a panel sponsored by National Science Foundation; WEST02 conference on voting technology in Washington DC last year. I am leading a panel on the topic at the Association for Computing Machinery's Computer Human Interaction 2003 conference in April of this year.

In the last month, I have been pursued by David Dill to sign a resolution in support of his position to require a voter verified paper trail for DRE voting equipment. Of the dozens of computer scientists Dill's resolution has enlisted, virtually none have experience with any of aspects of voting administration that they are commenting on.

For some reason, the voting and research communities have not been putting much effort into solving the user experience problems of registration, ballot design, running a vote, and absentee ballots. Instead many people are chasing their predictions of the possible security problems of the future as the number one issue of voting. The history of the idea of physical audit trails was developed a few years ago by two task forces lead by security experts who focused on their academic interests not the problems that have presented themselves.

Redundant separately recorded electronic audit trails are in all certified voting machines. Separate audit trails are in fact of value, but handling paper is notoriously unreliable. The idea that computers are not the best tabulating machines of our time is confusing. Banking quality has improved tremendously in the past few decades by computerizing the transactions and audits. Counting paper records by people is error prone. Such recounts have typically taken many tries corroborate computer results and has had to be given up or redone multiple times to be equivalent to computer records.

Separate redundant computer records are in place and can be reliable. I have been led to believe that the DRE-based Brazilian election in 2000 was able to report an unprecedented 99.8% of 106,000,000 votes electronically.

Requiring voting machines to be secure will eliminate fraud. I also agree that mass fraud is of special importance. Observations indicate that anything that complicates voting reduces reliability. The approaches advocated by Dr. Dill and others do complicate the process. The problems of recent elections are almost all caused by complicated local problems.

My major concern is that the people that have been assembled for the California Secretary of State Ad Hoc Touch Screen Task Force do not represent sufficient experience with human-computer interface research or voting administration to make a defendable decision. The danger is to make decisions that will have enormous costs and not improve the voting quality.

I am at your disposal to help evaluate ways to improve voting in your historic opportunity to solve major problems with voting in our democracy.

Sincerely

Ted Selker

Professor, MIT Media Lab
Caltech/MIT Voting Project
20 Ames St., Cambridge MA, USA 02139
selker@media.mit.edu
617-253-0291,
FAX 617-258-0910

Member Benefits | About AAPD | Join | Disability Resources | News | Contact Us | Calendar | Home