« 13 July 2004 - 3:41 pm | Main | 9 August 2004 - 2:22 pm »

Krugman mentioned it and metafilter linked to it. You too should take a few minutes to read about why you shouldn't necessarily trust touch-screen voting.

The key is for these machines to leave some kind of paper trail. I'm not a big fan of printing out receipts for the voters, but an internal audit tape seems like a really good idea. ATMs have been using them for years -- that's why you'll hear the clatter of a printer when you withdraw money even if you've asked not to be given a receipt. The machine is making a hard-copy of your transaction. And that's exactly what needs to happen with these fancy new touch-screen voting machines.

One of the main benefits of these new machines is faster vote tabulation, but for the first few elections I want someone to be comparing the audit record to the electronically reported results. I may sound a bit like a troglodyte when I say something like that, but I earnestly believe in real-world testing -- its the only way we'll find out whether these machines work as advertised.

Posted by on 27 July 2004 at 10:25 PM

You don't sound like a troglodyte at all, quite the opposite. I forget where I heard the analogy between software and engineering and bologna, but it is quite apt. Learning about how it's made makes you very suspicious of the final product. Those who place great trust software to solve problems like voting come across as those who are still coming to terms with technology. All software has bugs, and the more widely it's deployed, the more often those bugs are going to be exposed. For accounting, payroll, customer records, etc, the risks aren't that big. Voting is different.

The benefit of faster vote tallying doesn't seem like that much of a benefit, particularly when weighed against increased equipment costs and the risk of an untested, fragile technology. David, you've worked polls before, would it really be that much of a help to you?

Posted by Charlie on 28 July 2004 - 1:39 AM

To be honest, I don't really faster vote tallying is really all that important. With everyone so up in arms about these new-fangled voting machines it is important to make one point that no one seems to be considering: counting votes via computer is nothing new. We use computers to read and tally votes that come from scantron systems and the infamous punch card ballots. So a rogue programmer can just as easily change the code on the punch-card ballot readers as he can any other computer-based vote tallying system. The important difference is that all of these other systems produce a paper record that can be manually counted if you doubt the automated system.

Now, the deeper you dig into this issue the murkier it becomes. Because, if I'm a rogue programmer and I'm stealing votes from one candidate, if I'm good I'm going to make an effort to change what is printed out on the paper record. So we probably need to have the voter view the paper record before their votes are accepted. But we have to be careful that we don't let them leave a secure area with the vote receipt, because that opens up the door to a different kind of voter fraud (selling votes).

So, I guess what I'm trying to say is that newer is not always better, and this seems to be a case where technology doesn't offer us much.

Posted by david on 28 July 2004 - 8:58 PM

There's tension beween somebody being able to prove the world how he/she voted, so that miscounting can be proven, and "selling" one's vote. If I can prove I voted a certain way then I can sell my vote.

Mechanical voting methods also have problems. Nothing is perfect, so we have to figure out what's the best way to receive and tally votes. Public scrutiny of the code used for voting systems is necessary.

Posted by rkc on 29 July 2004 - 6:37 AM