November 3rd, 2004

How to run a verifiable election

I'm looking at the results and the concession, and maybe Bush got more votes than Kerry and maybe he didn't. There's no way to tell. Kerry would probably have been a less bad President than Bush, but it's hard to be absolutely certain of that, either. I don't know whether were enough states with close results that there should have been recounts.

Anyway, what my mind's been on (aside from a sense of suspension because ghu knows what happens next) is how to bulletproof an election, starting from the assumptions that we want to keep secret ballots, there's a strong incentive to cheat, and there's no way to trust anything that's only in a computer.

First off, you've got to have material ballots, presumably paper. It probably makes sense to have a computer interface for the voters which prints out the ballot so the voter can check it. The ballot (with the filled out part not visible) goes into a transparent box outside the voting booth. Humans and video cameras keep eyes on the boxes. (There was an NPR piece about an African country that uses transparent ballot boxes to prevent ballot box stuffing. They also have post-election parties to celebrate voting.)

After the election, the boxes are given a good shake, and then the ballots are piled together and shuffled again. Human and video eyes stay on them through the whole process. Then the ballots are opened and put up on a public wall and the wall is broadcast on the net so that anyone can count the votes.

The ballots were shaken and shuffled because otherwise it might be feasible to tell who'd voted for what.

Afaik, this system offers very little chance of cheating with the count. Voter intimidation is still a possibility, as is physical destruction of the ballots before they're opened. So might be some cleverness with disappearing ink, I suppose, which would be an argument for people to fill out their ballots with their own pens.

Doing it this way wouldn't be cheap, but I don't think it would be horribly expensive as large projects go, either.

I can't figure out how to make voting by mail and/or email similarly bulletproof, but I'd be interested in suggestions.

Thoughts? Suggestions?