A New Way To Vote: Accessible, Affordable, Available

ABSTRACT

The Florida election fiasco of 2000 has spawned thousands of articles on what is wrong with our voting systems, hundreds of ideas on how to fix the problems, and one radically different voting system that is arguably more secure and less expensive than any other system in existence.

The system described here is primarily software, and runs on ordinary PCs as old, small and slow as a 20 MHz 386 with 4 Mb of memory.  Such computers are free.  Security is obtained not by secrecy, but by openness; everything is public knowledge except how any individual votes.  Fully functional software is downloadable from the web.

Accessibility is trivial for anyone who can read a computer screen and use a standard computer keyboard (perhaps with mouth stick or head stick).  For those with even less mobility, only two keys are absolutely essential (if not doing any write-in votes); simple mechanical or electronic keyboard interfaces could be designed to operate those keys.  One totally blind person and two legally blind people have voted successfully with this system.  They refer to a ballot in another form (Braille, voice tape, or large print) and request tones from the computer telling them current page number, line number, and whether or not that candidate is selected.

INTRODUCTION

Problems with existing voting technology were recognized long ago. It was known that machines failed, that recounts (where possible) almost always gave different numbers, that invalid ballots were common, and that people with various disabilities could not vote without assistance. Mechanical lever machines, once very popular, pose difficulties even for people who are merely short! Nevertheless, not until the too-close-to-call presidential election of 2000 did those voting problems really gain national public attention.

Since November of 2000, every newspaper in the nation must have carried at least ten articles about voting. The Washington Post alone has carried at least 300 [1]. Politicians at national, state, and local levels, ever eager to propose solutions before understanding the problems, introduced well more than a thousand voting-related bills before May of 2001 [2]. As early as a month after the election, at least three bills already had been introduced in Congress [3]; a month later the number exceeded a dozen [4].

The major voting-related bill that actually became law is the so-called Help America Vote Act, usually referred to as HAVA [5]. One of the features of HAVA is a requirement that polling places have at least one handicapped-accessible voting machine by the 2006 November elections. The definition of "accessible" is vague, but seems to imply that at least people in wheelchairs and blind people should be able to vote unassisted.

"Accessible" voting machines are available from many sources, but tend to be appreciably more expensive than "ordinary" voting machines. Since many state and county governments are facing budget crises, the legal requirement to upgrade voting technology seems to require painful decisions.

TRADEOFFS AND CONTROVERSY

Changes of almost any type generate resistance. Simple inertia can be a significant, if unstated, factor; however, reasons given for resisting voting technology changes include cost, accuracy, security, uncertainty, and even accessibility.

Cost is a major factor. If voting officials believe that current technology works "well enough" (and many do), why should they invest in something new? A change typically requires not only a major capital investment, but retraining of all involved, including the voters. For a "quick and dirty" estimate of costs, consider that Georgia, with a total population somewhat over eight million, spent fifty million dollars to upgrade the entire state to a common voting technology, or roughly $6 per resident [6]. Maryland is paying even more per resident for their upgrade.

Accuracy was perhaps the original key issue demanding change. The hanging chads in Florida were only part of it. Optical scanners had their own problems as some people did anything but follow the directions for marking their choices [7]. Surely, people thought, modern computerized systems would be more accurate. Then a Caltech/MIT group analyzed voting results and concluded that the electronic machines were no better than the notorious punch-card systems [8]. Other situations casting doubt on electronic system accuracy included races in which the results were quite surprising or even patently unbelievable [9]. Now many people are concerned that the electronic systems may be less accurate than older methods.

Security/trustworthiness is perhaps the most critical voting system issue at the moment. Major voting system manufacturers have had the means (proprietary voting machines used throughout a county or state), the motive (strong connections with specific candidates or parties [10]) and the opportunity (secret software, modified at the last minute [11]) to affect the outcome of elections. This does not necessarily mean that they have done it, or will do it; however, the possibility raises grave concerns. Dozens, perhaps hundreds of web sites and their backers articulate these concerns, document the suspicious facts, and/or push for a solution in the form of a "voter-verified paper trail". Among the better known are www.blackbokvoting.org (Bev Harris), www.notablesoftware.com (Rebecca Mercuri), and www.verifiedvoting.org (David Dill).

Uncertainty about accuracy, security, and trustworthiness of new voting systems, as well as uncertainty about the twists and turns of political processes are valid reasons to avoid as long as possible making big financial commitments. If a county or state invests in a voting system that later becomes illegal (perhaps because of a new legal requirement for a paper trail), it must either upgrade the existing systems for whatever the current supplier wants to charge, or discard the relatively new machines and go with a different supplier.

The accessibility issue, interestingly, finds some advocates for the disabled opposing some advocates for better security [12]. The latter insist that the only way electronic voting systems can be trusted is if they produce a voter-verified paper ballot that can be hand-counted. The former insist that existing electronic machines are perfectly satisfactory, and adding that paper verification step would violate the rights of blind people. This author insists that both are wrong.

As long as the technology allows a blind person to vote in private and review and change selections before casting them, it should not matter whether that same information appears only on a screen or also on paper. Sighted people who test machines can use all the outputs of the "blind mode" to attempt to assure that the machine operates properly for everybody. This country can't afford to leave major openings for voting fraud. As for the paper trail itself, that's merely a "paper patch" for a system that remains a black box. It may give a perception that the machines can be trusted, but paper records repeatedly have been involved in election fraud. There is a better way to achieve security.

A SOLUTION

In December of 2000 a radically different voting system, named SAVIOC, was conceived and prototyped. The driving factors behind the design were low cost, accuracy, security, and ease of use. Most voting systems require specialized, expensive hardware; SAVIOC is mostly software and runs adequately on ordinary old computers that are literally free. Most voting systems depend on secrecy for security; SAVIOC achieves security by complete openness. Everything is public knowledge except the selections of any individual voter. Most voting systems have failure modes that allow erroneous vote counts, and have demonstrated such failures [13]. Because of innovative design features, SAVIOC cannot misattribute votes or produce overcounts, and accidental loss of votes would require an event almost as unlikely as a meteorite or stray bullet hitting the diskette drive during voting. Most voting systems present choices in confusing two-dimensional arrays; SAVIOC is strictly linear - one pass through the ballot from top to bottom shows every legal choice. Furthermore, because of that linearity, the normal ballot-marking process can technically be accomplished using only two keys - the [down arrow] to scan down through the offices and candidates, and the [spacebar] to mark a candidate as selected (or unmark a previous choice). In practice, the [up arrow] is very useful for correcting overshoots and reviewing selections. In addition, the [Y] key (for Yes) is used to indicate the official start and official end of voting. (Write-ins, of course, could require any of the alphabetic keys.)

The use of a standard computer means that anyone who can read the screen and press a few keys (perhaps even using a mouth stick or head stick) can vote unassisted. This makes the SAVIOC voting system accessible to people with most types of physical disabilities. In 2002 it was recognized that the same system could be made accessible to the blind, with some additions to the software. The inherent linear presentation of the ballot uses one screen page per office. An option was added to request tones that indicate the current page number and candidate number, and whether that candidate is selected. By referring to a copy of the ballot in some other form (Braille, large print, voice-on-tape, etc), one can use the tone clues to vote without seeing the screen. Two legally blind people and one totally blind person have used SAVIOC successfully.

SAVIOC is ultra-secure because everything (operating system, programs, and ballots) fits on a single diskette and everything on the diskette can be verified mathematically [14]. Ballots, screen displays, and stored results are all plain-ASCII characters. Tone clues are relatively easy to generate in such an environment, and never disagree with what appears on the screen. Security and accuracy should not be sacrificed for fancy fonts and speech, when simpler approaches work well. If some compact DOS-based software could be identified that could pronounce text messages, that might be incorporated into a future version.

All of the very limited blind-testing done so far has used the computer's built-in speaker. Headphones or earphones are desirable because someone else listening carefully to the sound clues from an open speaker could decipher the blind person's choices. In that case, the vote would not be secret. Laptops are much more likely to be equipped with a small earphone jack and volume control. (The author's ThinkPad is so equipped.) The disadvantage of a laptop is that the power switch and diskette drive are immediately adjacent to the keyboard. Custom shielding for each style of laptop would be necessary to avoid accidental or malicious shutdown.

A better approach is simply to install a speaker jack on any machine to be used for blind voting. This can be done for a few dollars worth of parts and in less than an hour of time. A similarly simple extension (only slightly more expensive and time-consuming) can add volume control and even a switch to toggle between headphones and the computer's built-in speaker. Another option is to add a small loudspeaker to the extension, giving volume control on this open speaker. (The volume of built-in speakers varies significantly from one computer to another, and placement of the system unit also affects volume for the user.)

ADDITIONAL SOUNDS AND FEATURES

It was mentioned above that the sound clues indicate page number, line number, and whether the candidate on that line is selected. They actually do much more. There are distinctive sounds for scrolling up or down within a page and between pages, for overvotes and undervotes, and for reaching the end of the ballot.

Once any sound clue has been requested, all automatic clues (such as stepping up and down within the ballot) remain active until deliberately turned off or until the current voter finishes. While in this sound-clue mode, write-ins are handled a special way. During typing of the write-in, the name being typed is displayed in extra-large characters (only eight fit across the screen) for the benefit of the legally blind. When that doesn't suffice, a write-in still can be verified. When sound clues are requested for that write-in line, they first indicate the page number, then the line number, then that this line is selected. After that, the write-in is spelled out one character at a time using the "alphabet song" that most people learn before first grade. The notes of the song are played, stopping on the note of the current character. Morse code would be a lot faster, but few people know Morse code anymore.

As people become used to the sound clues, they may get impatient with them. For that reason, a wide range of speeds is available. The [F1] key requests tone sequences at the slowest rate. The [F2] through [F8] keys provide progressively faster rates, so one can pick the most comfortable speed. Any of these keys turns on all of the "blind mode" features and produces an immediate indication of current status. The [F9] key reverts to the default mode, turning off all special sound clues and the extra-large write-in characters.

EVEN GREATER ACCESSIBILITY

An ordinary discarded computer with a simple headphone modification provides all the hardware necessary for a voting system usable by the vast majority of voters. It is not sufficient for those lacking even head motion (no direct keyboard capability), or those who are both deaf and blind (tactile sensing only). It is not yet clear how far counties will go in providing for extreme disabilities, but the author has begun investigating the possibilities.

Because SAVIOC is based on standard PC hardware, it can take advantage of commercially-available accessories. One of the simplest changes, helpful to some, would be replacing the standard keyboard with one that boasts inch-square keys and a simplified layout. (Such keyboards are available, for instance, from Fentek Industries in Cottonwood, AZ: www.fentek-ind.com.) There also are interfaces available that can accept any type of contact closure (sip-and-puff, foot switch, etc.) and convert it into a keyboard input. (One such is named DarciToo, available through www.westest.com.) The "keywedge" technology used by DarciToo to mimic keyboard input is available independently. (Altek, in England, is a supplier.) Using such a device, a sophisticated computer system could be used to interpret eye blinks or perhaps even brain waves and convert them to simple keypresses sent to the SAVIOC computer.

Braille "displays" already exist. (Suppliers include Fentek, Freedom Scientific, ALVA and others.) Linking one to a SAVIOC system could give a completely tactile interface. Since all SAVIOC screens are plain text, it is not conceptually difficult. (Far more significant changes have been made to the software before.) SAVIOC could send current screen content to the Braille display through a serial port. Information returned via that serial link could substitute for actions on the SAVIOC keyboard. This option is not cheap! Braille displays can cost in the $10,000 range if they display a full 80-character line.

CONCLUSION

The SAVIOC voting system is usable by almost anyone who can press a few (visually and tactilely) well-marked keys on a standard computer keyboard. It lends itself to adaptations for those with more limiting disabilities. It is arguably more secure, more accurate, and more trustworthy than any other voting technology in existence. It costs so little that a county switching from some other type of voting technology is almost guaranteed savings the first year. Fully functional software and instructions can be downloaded by anyone from its web site: http://www.savioc.com.

SAVIOC is accessible (to nearly all), affordable (even free, for testing), and available (today, to anyone who wants to try it). As of this writing, it is not used for official voting anywhere.

REFERENCES

[1] A search of Washington Post Archives for {"voting machine" or "voting technology"} returned roughly 100 hits for each recent year. (Search done 2004 June 13.)

[2] 2001 Overview of Election Reform Activity in the States. (2001, May 9). Retrieved 2004, June 13, from the National Conference of State Legislatures web site: http://www.ncsl.org/programs/legman/elect/taskfc/appE.htm

[3] The Associated Press. (2000, December 5). Lawmakers introduce voting system reform bills. Retrieved 2004, June 13, from the CNN web site:
http://www.cnn.com/2000/ALLPOLITICS/stories/12/05/voting.reform.ap/

[4] The Associated Press. (2001, January 30). Congress Considers Voting Changes; Lawmakers Introduce Election Reform Bill. Retrieved 2004, June 13, from the ABC News web site: http://abcnews.go.com/sections/politics/DailyNews/electionreform_010130.html

[5] Help America Vote Act of 2002, (2002, October). Retrieved 2004, May 21 from the Federal Election Commission web site: http://www.fec.gov/hava/hava.htm

[6] Pettys, Dick. (2002 September 17). Some fear Georgia's new voting machines will be replay of Florida problems. An AP article retrieved 2004, June 1, from the Kiosk.Com web site: http://www.kioskcom.com/articles_detail.php?ident=1530

[7] A Florida voting official, quoted in The Miami Herald, 2001 April 6, referring to ballots uncounted by optical-scan systems:
"Some [used] pencils instead of the special, black marking pens ....
"Elsewhere, voters used pens instead of pencils that were supplied ....
"Some used their own pens filled with red or blue ink ....
"Some scrawled X's through bubbles instead of filling them in.
"Some drew their own bubbles in the wrong places.
"Some people just like doing their own thing."

[8] A Preliminary Assessment of the Reliability of Existing Voting Equipment, (2001, February 1). Retrieved 2004, May 21, from the CalTech web site: http://www.vote.caltech.edu/Reports/report1.pdf

[9] Missing Votes Break Florida Law! (2004, January 7). The Miami Herald. Retrieved 2004, July 17, from the Verified Voting Foundation web site: http://www.verifiedvoting.org/
The article reports that on January 6, 134 voters signed in for a special (one contest) election in Broward County, but no votes were recorded on the electronic voting machine.

[10] Fitrakis, Bob & Wasserman, H. (2004, March 5). Diebold's Political Machine. Retrieved 2004. June 1, from the Mother Jones web site:
http://www.motherjones.com/commentary/columns/2004/03/03_200.html

[11] Zetter, K. (2003 November 6). Suspect Code Used in State Votes. Wired News. Retrieved 2004, June 1, by the Lycos search engine:
http://news.lycos.com/news/story.asp?section=MyLycos&storyId=797139

[12] Zetter, K. (2003 December 16). Group Seeks E-Voting Standards. Wired News. Retrieved 2004, August 8, from http://www.wired.com/news/evote/0,2645,61617,00.html

[13] Fact: Electronic Voting Machines Have Miscounted Votes. (2004). Retrieved 2004, July 17, from the Verified Voting Foundation web site: www.verifiedvoting.org/article.asp?id=997

[14] Jones, P.E. & EastLake, D. (2001 September). US Secure Hash Algorithm 1. Retrieved 2004, July 12, from the Internet Engineering Task Force web site: http://www.ietf.org/rfc/rfc3174.
The code subsequently was modified to enhance usability and SAVIOC-specific verification.