I am not a programmer, and I don't play one on television. But I've made my living for nine years repairing computers, managing computers, and supporting computer users. So I believe I know a little something about how computers work, and what this means for using them as voting machines.

I won't try to engage in journalism here: what follows is about issues, not technical details. If you need exact details, you can Google them right up. But watch your sources!

Any election is a dialog between competing candidates and those who would like to elect them. All elections have a certain margin of error, which isn't significant unless the election is very close. Then questions of error become very significant indeed.

In 2000, a badly designed ballot brought to light some uncertainties of our punch-card voting system. Voters learned about hanging chads and many technical aspects of recounting votes. Transaction-handling equipment companies saw an opportunity to use public distrust of the voting process to market a whole new class of voting machines. One of these machines was made by the Ohio-based Diebold corporation. I'll use them as an example because their system has become controversial.

"I am committed to helping Ohio deliver its electoral votes to the President next year." - Wally O'Dell, CEO Diebold Corp.

Diebold developed a new touch-screen voting system that was said to overcome the shortcomings of the old system. It would be easy to use, and it would record the vote certainly one way or the other.

The problem with the system is that it's basically a computer - the same kind of thing you have on your desk. I won't recount here the history of the development process, but leaked memos and notes revealed that the system was vulnerable to tampering. And the system would not give a paper record of the vote. Diebold said there were technical problems with printing any paper record (even though all of Diebold's ATM's do just that.)

Another objection to a paper receipt is that it could become part of a vote-selling scheme. But as we'll see later, that's an easy problem to fix.

In 2004, many elections will be using the new machines, which Diebold says are safe from tampering and far more accurate than than traditional methods. They still won't be making a paper record.

But I maintain that a paper record won't guarantee accuracy as long as the vote is recorded electronically. It could even help cover up election fraud.

There were two main voting-booth issues in the 2000 election:

1) The system was confusing: It was not totally clear in some places where you punch to vote for a certain individual. In West Palm Beach, Florida, the ballot was designed so confusingly that even Pat Buchannan, that bastion of fairness, admitted that the votes that went to him should have gone to Al Gore. There were enough of those votes in that county alone to have thrown the Florida electoral college votes to Gore.

Many commentators said that the voters were "dumb" or "senile" to have made such a mistake but the simple fact is, if a lot of users are confused by a system, then it's a confusing system.

This could be fixed with a touch-screen voting machine. You can show a picture of each candidate, along with the candidate's name. If it is properly designed errors of choice could be almost eliminated.

2) The vote was not always recorded clearly: In many cases, the intent of the voter was not clear because the punch did not go all the way through the card. Many voters simply didn't punch hard enough, or the punch track was worn so the stylus wasn't properly lined up. The breakaway holes or "chads" were sometimes dimpled or hanging in such a way it was hard to tell what the voter meant to do.

This ambiguity is what made recounting the votes so difficult.

Again, this could be fixed with a computer. Computers are digital, not analog. A data bit is either -0- or -1-. Once the vote is recorded, it will be unambiguously for Smith or Jones.

So electronic voting machines are great, right?
Not so fast. First, as I mentioned above, it's pretty simple to get a computer to do one thing and say another.

The real problem is electronic tallying. It is quite simple for a computer to say it did "X" when it really did "Y" because the screen, the printer, and the external connection are different output devices. The computer needs only say x to one and y to the other.

For example, you could vote for Jones, and the computer screen would say, "You just voted for Jones." The receipt printer would print out, "You just voted for Jones." But the vote actually transmitted would be for Smith.

Wouldn't that be noticeable if a precinct only recorded votes for one side?
Not if the computer randomly shifted votes from Smith to Jones. It wouldn't have to do that for every vote to throw the precinct.

Wouldn't it be possible to detect the tampering?
Yes and no. If you write a script that tells the computer to alter the votes, then clean itself up and leave the system in correct working order afterwards, it might only be detectable in a computer forensics lab. And, perhaps not even then. Generally, "bits" can be changed without leaving any trace.

Assuming, of course, you even suspected there was anything wrong.

But suppose you could tell that a machine had been tampered? That's a long way from accurately retrieving the data that the voters had input originally. A "recount" would consist of asking the computer - "How many votes?" and it would say the same thing again. You'd be left with unusable data. You would have to throw out the votes - a rather pyrrhic victory.

What's the solution?
Very simple: a paper ballot. Not a paper record of an electronic ballot to be counted electronically, but a paper ballot that goes into a ballot box and is counted.

I understand that in Canada there is a well-designed paper ballot about 5 inches wide, on which the candidate's names are printed. Voters choose their candidate by making a mark in a circle next to the name. The ballots are then counted at the voting station under the supervision of representatives from the parties of the election, and the vote tally is then reported to the central counting location.

To keep the counts manageable, each voting station has about 500 ballots to count. If there's any question about the total, the ballots can be counted again... which is relatively simple to do with physical ballots.

One election fault that Canada shares with the United States is that precincts on the East coast may be reported in the news before precincts on the West coast have voted. (In my opinion this should be fixed: no precincts should be reported to the media until everyone has voted.)

(Many thanks to the author of Paper Vote Canada for specifics! See also: The User-Friendly Guide to Voting in Canada.)

Just like a liberal - to want to do things the Canadian way!
I'd be interested to hear any suggestions as to how that system would work in the favor of one party or the other. Seems to me it works in favor of the voter.

But paper is so... last century!
Paper ballots have been around for a long time for a good reason: they're reliable, and make a record of the voter's intent that (if handled appropriately) is pretty hard to tamper with. The vote takes the form of a physical object, not a "bit" inside a black box.

Great system. But what about the advantages of the touch-screen?
Again, simple: just forget about tallying electronic votes. Use the computer to print out a clearly printed ballot that can be read both by machines and by a person. The voter looks at the ballot and verifies that the vote was cast for who they wanted. Then they exit the voting booth, fold the ballot, and put it in the ballot box with the election monitor watching.

In this example, the computer does not have to be honest. It is simply a means of printing a very clear ballot. If the ballot indicates a vote that is not in accordance with the voter's wishes, he can have the official mark it as a spoiled ballot and it will be placed in the spoiled ballot container. The machine will be checked and the voter can try again, until that ballot matches the voter's choice. The correctly printed ballot can then be put into the ballot box.

If you're thinking, "Just eliminate the computer and have paper ballots like Canada," I could live with that, too. Computers would just speed up and (hopefully) clarify the process of marking the paper ballots.

Won't that compromise voter privacy?
No. The ballot can have a fold-crease on it and voters know how to fold a ballot before exiting the booth.

Vote-buying won't be possible. The voter can get a receipt saying "I voted" but the actual ballot with the candidate's names clearly printed on it is in the box - an actual physical box, not a "black box."

This system combines several advantages:

• The computer does not have to be "honest"
• It makes choosing the candidate easier
• It creates an unambiguous ballot
• Voting can still go on even if the power goes out (filling out ballot tickets by hand just like in old-fashioned elections)
• The ballots are guarded in a physical box, not by mysterious software
• It allows a recount of physical ballots. The ballot could be readable both by humans and machines. At the bottom of each ballot would be a barcode that encodes the voter's choices. It would be simple to feed the ballot through a reader and compare its output to the visible choice.

Computers are supposed to carry out their instructions, but sometimes the instructions fail and sometimes they're tampered with. At times computers behave in unpredictable ways for lots of reasons. Try asking 100 office workers if they'd trust their computers with the outcome of an election... and be prepared to duck!

But voting machines will be in sturdy cases and look pretty indestructable, like ATM machines. It will create a false impression of security. People might not understand that at its heart this is just another computer, which should not be trusted with the outcome of an election.

Accurately recording votes is a solved problem, and has been for a long time. Ask the same 100 people to write a word down on a piece of paper. Then show them the paper and ask them, "did you just write that word?" People are right to trust paper, because it is very reliable. It's the technology we need for our elections.

And if it seems too old-fashioned, just slide your finger along the edge. That's "cutting-edge" technology!

- George Wiman

Appendix: notes on design of screen interface for voting machine:

The touch-screen system is supposed to make choosing the candidate clearer than with the butterfly ballot. And it will, if it is correctly designed - but here are some considerations the designers may not have taken into account:

• Will each choice be isolated and presented to the voter for confirmation? Upon touching a candidate's line, the screen should then show only that line with the candidate's picture, and the question: "You have chosen to vote for Jones. Is this correct?" and then present the choice of "Yes - confirm this vote," and "No, let me make the choice again." Long experience with computer user interfaces has shown this to be necessary.
• What is the scan rate of the computer monitor? A scan rate anywhere near 60 or 90 hz causes noticeable flickering under flourescent lights, which makes choosing a specific line more difficult for people with slightly compromised vision.
• How will glare be handled? Is the curtain in the voting booth white on the inside? If it is, it will glare on the screen. The inside of the curtain needs to be a dark color.
• If the screen is wheelchair accessible, it will be too low for most people to use. A chair will have to be provided in the booth. If it is high enough for most people to use, it will be too high for people who use wheelchairs. A monitor that swivels around a horizontal axle might answer this problem.
• How many candidates can be shown on each screen? In elections with a lot of candidates (such as city council elections) each line gets correspondingly less space in the low-resolution medium of the computer screen.
• How will the voter make the transition to ending the voting session? Not all voters will be "computer literate," which means "distrustful of interfaces and prone to reading every last thing on the screen before touching anything." (This same approach would have enabled voters to beat the confusion of the West Palm Beach butterfly ballot, by the way. But the voters weren't expecting a trick ballot.

Gee, kind of makes you want to print up clear voter ballots on paper and just skip the computer altogether, doesn't it? - George Wiman

(There were other problems with the 2000 election, such as voters being turned away because of record-keeping errors and such (in some cases the errors were rather suspicious). But those problems were beyond the scope of this essay.)

___________________________________________________________
[Home] ... [Site Index] ©1997-2003 George Wiman, All Rights Reserved
This page created in Macromedia Dreamweaver. Please email george@wiman.us if you have any problems viewing this page.