Ned Ludd Goes to the Polls

[Update: A thoughtful and thorough piece by Bruce Schneier, which covers a great deal of the same issues as below — and some more besides — though which seems to begin from the unsupported premise that technology in voting is in the end both inevitable and desirable, so long as its inherent problems are fixed. I can’t disagree more strongly, for the reasons I hope I make clear.]

Machines make the pencils — and, having been to the Cumberland Pencil Museum, I can attest to the adorably steam-age nature of those machines — but that’s where the technology stops. The rest is people. The pencils — wee stubby little things — are placed inside the rudimentary voting booths, where they wait patiently. Voter comes to vote. Voter gives their voting card — or their name and address — to a poll worker, who stamps a ballot paper to make it official, then gives it to voter. Voter toddles along to booth, picks up pencil, marks the cleanly, simply-designed ballot paper with the wee stubby little pencil. Voter folds ballot paper, takes it back to poll worker, and drops ballot paper into ballot box. After the close of polling, ballot box is taken to central counting location, where people, lots of people, tip out the ballot papers onto long tables and count them by hand. And then recount them by hand if necessary. Once the result is determined, local dignitary, surrounded on the stage by the candidates — the seven-foot-tall transsexual, the earnest Communist in the bad suit, the wild-eyed neo-fascist — puts on his very best Posh Voice and haltingly calls out the poll numbers. For the first few hours of the night, before the flood of results gets to be too much, he does this on live television; there’s a deal of civic pride to be had from being early to announce. It’s magnificent, epic theatre, simultaneously comedy, tragedy and farce. Media punditry then takes over to interpret and reflect, but the moment of truth is quite raw.

That’s what a general election in Britain is like. The whole thing is adorably low-tech. And it works. And — and here’s the crucial thing — the process of voting and vote-counting is never under any suspicion. That’s terribly important. Voting is like justice: it’s not enough that it be done fairly; it has to be seen to be done fairly. When trust in the voting process breaks down, trust in the democratic process breaks down, as we’ve seen in recent years.

As someone who writes software for a living, I shudder at the thought of introducing technology into the voting process. Software has bugs. Always has, always will. Even if one doesn’t buy into conspiracy theories about vote fixing, it’s just a fuck-up waiting to happen. Strike that. It’s not waiting to happen; it’s already happened, in Florida, 2000. Hanging chads changed the world.

And I don’t, don’t, head-banging-on-the-table don’t get why technology is considered for voting if it might in any way compromise the integrity of the process of voting, or the process of vote-counting. What’s to gain? Speed? Is that all? Once every four (or five) years, who cares about a few hours, or even a day or two, if it means that there’s confidence in the count. If people don’t have confidence in a vote count, then the rest of the democratic process gets all gummed up, as we’ve seen. Some things just aren’t possible without technology, so you bite the bullet, do your damnedest to make the technology robust and safe, and you try to live with the small risk. Voting’s not like that. It’s both simple enough not to need technology, and important enough that technology is mostly inappropriate.

I’ll be more specific. I can see three separate steps involved in voting:

1. Marking the ballot. The pencil’s finest moment. There is actually some room for the safe use of technology here, so long as three vital criteria are met: first, a marked ballot must be produced, whether by man or machine; second, the voter must be able to view the marked ballot, in order to see that their choices have been translated accurately onto the ballot; third, the voter must have confidence that the ballot will be counted according to their choices — the system fails if the voter believes with complete confidence that they’ve voted one way, but the counting system cannot support that confidence.

   These criteria don’t preclude the use of technology — punched-card machines, or even touch-screen machines — so long as the product of the technology is a marked ballot paper that the voter gets to see, hold in their hand, and carry across to the ballot box. The technology should, in other words, be no more than a high-tech substitute for the pencil. The criteria certainly don’t call for technology; they certainly do place limits on the proper use of technology where it is used in the act of voting. The key is not to put technology between the voter and the marked ballot paper. To have a machine assist in the marking of the ballot doesn’t increase the distance between voter and ballot, so long as the machine only helps to generate the marked ballot, its contribution stops there, and the voter can see, hold, and approve the ballot after its generation.

   The criteria also call for a simple, clearly-designed ballot paper. Voter confidence that their choices have been translated accurately onto the ballot diminishes if the ballot is unclear or complicated. Technology can also hurt this confidence. A ballot paper that’s designed to be counted by a machine requires significant design decisions which are not necessarily consistent with the ballot paper being clearly human-readable.

2. Counting the ballot. People and machines have very different skills. Reading marked ballots quickly is a machine skill. Reading marked ballots accurately is a human skill. Assuming that the voter’s choices have indeed been translated accurately onto the ballot, there isn’t a machine on the planet that’s better at interpreting those choices in order to count them than a human counter, nor could there be. Technology-based counting can at best approach the accuracy of a human counter to within some acceptable tolerance. But, nuh-uh, rewind, rewind, and scrub ‘acceptable tolerance’. What’s acceptable? Why should it be acceptable? Why compromise at all?

3. Recounting the ballot. I hope it’s clear that there’s no such thing as a viable recount in the situation where there are no paper ballots. What would one do? Run the same counting program over the same data? The very concept of a recount requires that something be done differently: perhaps a new team of human counters, or at the very least a new count by the same team with a renewed focus, reflecting the fact that a recount was necessary. Technology’s great strength isn’t accuracy. It’s a tireless consistency and fidelity. What it screwed up once, you can rely on it to keep on screwing up in exactly the same way, with a completely straight face.

None of this implies any partisanship or wrongdoing in the use of technology in the voting process. It’s not necessary to propose partisanship or wrongdoing to shoot big holes in the integrity of a vote that relies on technology. Software screws up. Often software screws up in horribly subtle ways that both defy explanation, and — scarily — defy detection. Ask yourself how you’d know whether a vote count provided by a machine was accurate? You wouldn’t, of course. You’d just have to trust it.

Which is not to say that people don’t also make mistakes. Of course they do. But here’s the thing: in the process of counting votes, what would a genuine human mistake consist of? What would its consequences be? A vote in the wrong pile? A handful of votes? That’s because the count of each vote is entirely separate from all of the others. Mistakes don’t propagate, and there’s no reason to imagine systemic errors which might prefer one candidate over another. There are no higher-level system problems which might cause a vote counter to screw up on a larger scale.

Technology’s not like that. The relationship between the size of a mistake in software, for example, and the size of the consequences of that mistake, is not proportional. A single misplaced punctuation mark might bring down the Space Shuttle. Likewise, a single mistake might potentially create an error of almost any size in a vote count. Unless the discrepancy was plainly anomalous, it could easily escape detection.

Of course, even though it’s not necessary to propose partisanship or wrongdoing, it’s not unreasonable to also propose both of those, at least in principle. Software can be hacked. The use of technology implies a developer of that technology, which implies contracts of significant value and importance. That amount of power in the hands of a commercial concern over the voting process, and therefore the democratic process, ought to be a worry for everyone.

And, asks the Ned Ludd in me: Why is this even an issue? What’s the problem that technology is intended to solve? Is a hand-count too slow? Okay, get more people. Or just fucking wait a bit longer. It seems to work just fine in Britain. Are ballots too complicated to be completed by hand, or counted by hand? Okay, simplify them. Separate off Federal-level votes from State-level votes if need be.

I wrote a few days ago about how meta the US election process is, how discussion of policy disappears beneath media spin, conspiracy theories, dirty tricks, fatuous American Idol polls. It’s increasingly the case that a simmering dissatisfaction with the integrity of the voting system itself is an issue — though after Florida, ‘simmering’ might be way too bland. This is poisonous stuff. Democracy just doesn’t work if people don’t think the leaders it gives them were fairly elected, and that’s more and more an issue in the US. Concerns about partisan manipulation of voter registration and vote counting abound. The increasing use of technology within US voting is only a small part of this, but it is a part. The voices decrying that seem to be far too quiet, and far too few.