Via MIT’s Technology Review, an item on computer voting and the upcoming election.

There was a particularly stupid an ill-informed op-ed (warning: audio link) on PRI’s show Marketplace yesterday. Basically the commentator felt that since ATMs are so reliable, we should trust voting machines. This completely ignores that fact that ATM errors have multiple redundant means of catching errors, since they generate a paper trail at the time of the transaction, the customer has additional opportunities to catch errors when they receive their bank statement, and the bank has enormous incentives to ensure correct accounting if they want to stay in business. If there is a potential problem with an ATM it can be taken off line for a couple of days until it is fixed.

In the case of electronic voting machines, they are put to the test once every couple of years, set up by people with minimal training, there is no independent audit trail, and there is considerable incentive to falsify votes, knowing that if you are successful you or your allies will control the investigation into what happened. Only an independent voter-verifiable audit trail can make electronic voting credible. Unfortunately my state (MD) is dragging its feet on this issue despite a well organized effort to knock some sense into the heads of the Election Commission.

I blogged this topic earlier, and I’ll do it again before the election. This is the single most important technological issue facing the US. We have the potential to completely invalidate elections. Without trust in the electoral process government has no legitimacy, and people will be forced to accept disenfranchisement or resist with force. That may sound like hyperbole, but I suggest you think carefully about the likely reaction if there is a significant split between exit polls and reported (utterly unverifiable) election results in a hotly contested election. I don’t think rioting is at all unlikely, and public officials hanged from lamposts is a real possibility. It’s all well and good to joke about that being a good thing, but there’s no guarantee that the officials hanged are the guilty ones, or that large scale public disorder will in any way actually address the problem. Just ask Reginald Denny.

I spent four hours last night working with commonly used commercial software which crashed three times. It was MicroSoft Word, so there’s something of an expectation that it’s a P.O.S., but it’s at least as heavily tested as the Dielbold software that I’ll be using to cast my vote in November. My confidence in the system working as it should is not high.