Technology and Elections

What follows here is really just a list of questions. Some are as old as voting, others are new. But even the old questions are becoming more urgent as technology gets into the picture.

And in case you're getting nervous, this column is not about politics, just about the technologies of elections (and I include in that the mathematics of elections). I am convinced that we as a society need to think hard about the mechanics of this definitive act of citizenship. Because the technology isn't always doing what we think it's doing. But when I say "we as a society", I really mean you. After all, who else is as capable as you of truly understanding the impact of technology?

I may not be talking about politics, but I am campaigning.

Recounts and Recalls

No, I'm not going to replay the Endless Election of 2000. But behind the jokes about hanging chads and butterfly ballots lie some legitimate questions about how to design a ballot fairly, and how to read one fairly.

The question of ballot design arose again this year when California election officials had to figure out how to present 135 candidates to voters without unfairly favoring one over the others. That's probably impossible; I certainly don't know how you would do it. The election officials didn't even try. Instead, they accepted that the ballot would be biased by a number of effects, such as the primacy effect that helps the person whose name appears first, so they just tried to give every candidate a fair chance at being first on the ballot.

They did this by the use of a random alphabet: RWQOJMVAHBSGZXNTCIEKUPDYFL. Names starting with R will come first, with the same alphabet being used to order R-names by their second letter, and so on. So a name that starts out "Ro..." looks good. Look for the primacy effect to help out tribal chairman David Laughing Horse Robinson and cigarette retailer Ned Roscoe.

It may not occur to the average voter, but you and I know that randomization is technology, and we immediately wonder: What was the algorithm? The answer is a familiar one: They drew lettered balls out of a Keno-style drum. Lottery tech is low-tech but not quite no-tech.

The butterfly ballot controversy seems no less silly today than it did back in 2000, but even Pat Buchanan admitted that he probably got votes that he didn't deserve because people were confused by the layout of that Florida ballot. Maybe quantitative information presentation expert Edward Tufte (The Visual Display of Quantitative Information) should hire out as a ballot-design consultant.

Then there's counting. Even if every voter pulls the intended lever or punches the intended chad or clicks the intended button, the counting mechanism will be less than perfect—whether it's done by machine or human. Depending on the design of the counting mechanism or process, there will be more type I errors or more type II errors—false negatives or false positives. If we haven't figured out how to set the sensitivity of spam filters to balance false positives and false negatives, how can we hope to do so with something as controversial and important as vote counting?

And the sensitivity question applies to assessing voting eligibility, too: In that same infamous Florida vote, a lot of African Americans allegedly got erroneously disenfranchised. My preference would be to set the sensitivity so that accidental disenfranchisement was as unlikely as possible, but that's probably a political opinion, so I'll keep it to myself.

Voting Machines and Machine Voting

The seemingly insurmountable problems of subjectivity and plain inefficiency in the human reading of paper ballots argues for at least some machine involvement. But how do you maintain a paper trail when there is no paper? Will voters trust electronic voting systems? Should they? Are trustworthy, totally verifiable electronic voting systems even possible?

A few questions that voters should raise: Suppose the machines don't always work right? (They won't.) Suppose they could be tweaked to introduce bias? (They can.) Suppose an election can be changed by flipping a single bit? (It can, and I don't mean the bit that Sandra Day O'Connor flipped in 2000.) Suppose there's a hard disk crash and the results of an entire election are wiped out? (It'll happen.)

Rich redundancy might be some comfort regarding that last concern, but the others are harder to address.

It's not that nobody's thinking about the problems, of course: Security and identity verification and reliability and privacy are mainstream computer science. They don't apply only to voting. But when someone steals your credit card, they only steal money.

Here's a scary quote from the Cleveland Plain Dealer that shows that we do need to worry about unscrupulous voting machine manipulation:

The head of a company vying to sell voting machines in Ohio told Republicans in a recent fund-raising letter that he is "committed to helping Ohio deliver its electoral votes to the president next year."

Then there's this from The Washington Post:

A touch-screen voting system that Maryland has just agreed to buy for $55 million and install in every precinct in the state is so flawed that a 15-year-old with a modicum of computer savvy could manipulate the system and change the outcome of an election...

Stanford University computer-science professor David L. Dill is all over this stuff. His web site, http://www.verifiedvoting.org/, tracks developments, posts resolutions, and serves as a soapbox for voicing concerns about the dangers of moving too quickly to electronic voting technologies. Dill says: "We just aren't there yet." I'm inclined to agree, but you should decide for yourself.

The Death of Privacy And the Secret Ballot

The secret ballot is a wonderful invention. It makes vote buying difficult; to the extent that the ballot is actually secret, it makes vote buying impossible. The ability to certifiably monitor actual votes is a requirement of an effective vote-buying plan. But are we about to lose the secret ballot? With 3G phones and miniature cameras, how do you ensure that there's only one person in the voting booth? And if the voting is not done in a booth?

The Pentagon seems to think that its SERVE system, with its secure military servers, digital certificates, and dual-key encryption schemes, will ensure trustworthy e-voting for military personnel overseas. It's possible that an electronic voting system could provide service personnel more assurance of privacy than they have had in the past, when they essentially handed their envelopes to their bosses and hoped they wouldn't peek. But the assertion that SERVE "would function pretty much like most common e-commerce web sites" is not reassuring.

So how workable is a voting system using digital certificates and encryption? Setting aside the "peripheral" issues, like spoofing and the ballot-box privacy issue, are there practical voting protocols? Well, one such scheme that has been put forward is the FOO protocol, named after its developers, Fujioka, Okamoto, and Ohta. The FOO protocol provides provable security provided that the cryptographic functions that it uses are unbreakable. However, it wouldn't work in a real election. To maintain a secure election, the FOO protocol requires that people who opted not to vote must check to make sure no vote was counted for them. But FOO's not new; I'm sure there are better protocols now.

It can be argued that if we really eliminate all privacy, vote buying will again become impossible. Mind reading may never be a reality, but technologies for reading mood and reaction to stimuli and for filtering out lies and pretense are getting more powerful. Real, unfakable opinion-reading technology would make it impossible for anyone to force another to vote against their true desire, because your desire would BE your vote. After all these years immersed in technology, I generally figure that the solution to any technological problem is better technology, but opinion-reading is a technology we don't have yet.

Rousing Us Rabble

One fact that emerged in the runup to the California voting this fall was that Arnold Schwarzenegger voted about as often as the average American. Whether that helps or hurts him is unclear, but it is generally assumed that it isn't good for democracy when citizens don't bother to vote. On that assumption, the biggest problem with voting in America in recent years has been that so few Americans do it.

Does the Internet have the capability to help increase political involvement? Does connectivity trump money? Or does it just help raise it?

Political strategists are all analyzing the efforts of Howard Dean's campaign manager Joe Trippi right now, because the Dean campaign was able to zip from nowhere to the front of the pack in a few months, and because it was raising money at a rate comparable to what Bill Clinton was able to raise. It was also getting people politically involved in impressive numbers.

Trippi has said that the tough lesson for him came when he decided to embrace the Internet. Running campaigns, he said, is traditionally about central control, and the Internet doesn't do central control well. He said he had to force himself to let go, to allow the campaign to decentralize. It appears that the right balance between central direction and decentralization helped produce impressive results for Trippi's candidate. Trippi became a blogger, and the MeetUp web service became a major organizing tool of the campaign. Other candidates jumped on the blogwagon, and strategists are studying Trippi's techniques for their applicability in the 2004 general election. The Internet has changed electioneering, but we may have to wait to see what it all means.

The Primary Is Not Secondary

Dean was also helped by the MoveOn "primary," an online Presidential preference vote held by the MoveOn.org organization. This was not really a primary. Some states have primary elections to select electors for party conventions, some have caucuses. The general idea is the same: Candidates run for the right to be candidates. Frequently, the positions that get you a win in the primary will come back and bite you in the general election. This is why politicians have such a problem with the idea that they should be expected to be consistent.

The consequence of the primary process is that voters in a few states determine who will be the candidate of each party in the next election. As a result, a voter in New Hampshire has a lot more political influence than one in California. I talked with Alane about how late-primary voters are trying to sway voters in early-primary states. At monthly meetings around the country this summer, Dean supporters were handwriting letters to registered New Hampshire Democrats, encouraging them to give consideration to Dean. It was a low-tech attempt to offset a mathematical imbalance in voter influence.

Then comes the election, and there is a similar problem. What if you know that your state heavily favors your preferred candidate. Then it really doesn't matter if you stay home or vote, does it? Although if everyone thought that way, it would. There's probably a name for that paradox.

The strategy for dealing with that problem is vote swapping. A lot of people in the 2000 Presidential election swapped a Nader vote in a solidly Democratic state for a Gore vote in a state that was up for grabs. Or claimed they did. There was some discussion about whether this was legal, but apparently it is. Whether it's enforceable is another question.

These primary and general election strategies are plays in a huge multiplayer game, and it is not necessarily the case that what makes sense for the individual makes sense for the whole. There is evidence that people regularly employ strategies in voting. One study of the 1992 U.S. Presidential election estimated that 35 percent of Perot supporters voted strategically rather than voting their real preference. Do you have to understand game theory to develop a voting strategy to, in turn, make your vote count?

One Person, One Vote?

Game theory may not help.

Is one person, one vote, plurality rules, really the most democratic voting scheme? There are alternative voting schemes, including: Unanimity, True majority, Plurality, Plurality with runoff, Vote for m-out-of-n (often used in city council or other local elections), Fractional voting (I give .6 of a vote to A and .4 to B).

Then there's the California conditional ballot (I vote "no" on the recall but here's my choice for a replacement if the recall goes through. Hmm. How richly Boolean can ballots be?)

The Marquis de Condorcet observed that it is possible for individual voting preferences to be transitive (if I prefer A to B and B to C, then I must prefer A to C), but for the collective preferences not to be transitive. This is called the "Voting Paradox," and it means that election results can be really screwy.

Arrow's Paradox, the devilish discovery of economist Kenneth Arrow, says that if you want to construct a voting scheme that is unrestricted, monotonic, nondictatorial, reflects citizen sovereignty, and is not dependent on irrelevancies, you're out of luck. I won't try to give Arrow's technical definitions of these constraints, but they're not unreasonable: Any voting scheme that doesn't have them is something we'd reject. In a stronger version of Arrow's Paradox, monotonicity is replaced by Pareto efficiency: If every individual prefers a certain option to another, then so must the resulting societal preference order. Not possible.

The Condorcet method is a nice voting scheme that produces very sensible results, if it produces any results at all. For each pair of candidates, you determine which candidate is preferred by each voter. If there is a candidate who wins every such comparison, there's your winner. If there is no such universal winner, there is no Condorcet winner.

Why not just run all elections as Condorcet elections, and declare "none of the above" or "no decision" if there is no Condorcet outcome? Because that's an outcome, and it may be one that everyone hates.

Then there is the Electoral College. It is not a political assertion to point out that a plurality of Americans rejected George W. Bush's bid for the Presidency in 2000. The pros and cons of direct democracy have been debated for long before America was a country, and is debated every time some crazy ballot initiative gets passed in California. But it's taking on much greater significance with the possibilities presented by the Internet.

Of course, the very idea of a multiple-choice ballot is a limitation on citizen control of their government—or at least a relegation of the control to some other mechanism. Maybe it would be best to just pass out blank sheets of paper to every citizen at regular intervals and ask us to write down any issues we have.

Well, these are some of the questions. I'm expecting you to come up with the answers. If I were providing the answers, they'd probably look like what you find if you search the Internet for "Austin Powers: Government and the Net."

Correction

I let Martin Campbell-Kelly off too easily in my September 2003 coverage of his book From Airline Reservations to Sonic the Hedgehog, as numerous readers pointed out. "[H]ad Microsoft ever produced a Pascal system," he says, "it would surely have eclipsed [Borland's] Turbo Pascal." But Microsoft did produce a Pascal system, Microsoft Pascal, which was eclipsed by Turbo Pascal. Later, I believe, Microsoft followed Borland's example and developed a bargain-priced IDE-centric product called Quick Pascal. And that was also clobbered by Borland. Apart from that, I think Campbell-Kelly was right on the money. I only wish my corrections could be as pithy as Wired magazine's delightful "Correction: Wonder Woman is not a native of Atlanta."

DDJ