Slate's Mickey Kaus complains that a computerized voting machine malfunction in North Carolina is causing the state to have re-vote for agricultural commissioner. Apparently, some of the machines "lost" more votes than the margin of victory for that office.

The Instapundit's response to this is: "I told you so!" Reynolds references an article he wrote for TechCentralStation that contains the following statement.

Paper ballots are easy to understand - just put an "X" in the box next to the appropriate candidate's name. I don't find voting machines especially hard to understand, but I do always have to read the instructions on the ones I use, and I'm a law professor who works as a sound engineer on the side. So others may find them more confusing than I do. Everyone, on the other hand, can make an "X."

Since most paper ballots actually have circles that you fill in, I'm reasonably confident that Reynolds would argue that filling in a bubble is the same level of simplicity as making an "X."

But while the North Carolina fiasco is a cautionary tale for the use of electronic voting machines, the Washington State governor's race should be one for paper ballots.

After Republican Dino Rossi won the initial count and a machine recount, a hand re-recount introduced human bias into the equation and produced a narrow win for Democrat Christine Gregoire.

For example, this is a vote for Gregoire:

This "voter" didn't make Reynolds' proverbial "X." There's some mark on the page's margin, but not in the bubble -- not in any bubble on that ballot. It looks like a blank ballot that someone scribbled on the margin.

Reynolds says that paper ballots are naturally fraud-resistant.

Paper ballots are surprisingly resistant to fraud. Actually, it shouldn't be that surprising. A paper ballot encodes lots of useful information besides the obvious. Not only is the information about the vote contained in the form, but also information about the voter. Different colors of ink, different styles of handwriting, etc., make each ballot different. Erasing the original votes is likely to leave a detectable residue. Creating all new ballots with fraudulent votes requires substantial variation among them or the fakery is much more obvious; that's hard work. And destroying the original ballots in order to replace them with fraudulent ones isn't that easy - there's a lot of paper to be disposed of, and shredding it, or burning it, or hiding it is comparatively easy to detect.

The image above answers Reynolds' contention that making a fraudulent ballot is difficult. Reynolds' defense of paper ballots also assumes that election officials are interested in detecting fraud -- they aren't. The mantra is "make sure every vote is counted," not "prevent vote fraud."

Besides, as Stefan Sharansky has pointed out, it's not necessary to replace, hide or destroy ballots when you're trying to commit vote fraud. There are 3,359 more ballots counted in King County than there were voters to cast them.

If paper ballots are the simplest system, then we should require "simple" competence from voters. Make your "X" in the proper place.

Even though they have their own problems, electronic voting machines also have advantages -- there is only one way to determine voter intent -- the actual vote. There are no marks in the margins to interpret, machines won't allow overvotes and will remind you if you've chosen not to vote in a certain race.

We need to be wary or any voting system which allows political partisans to start determining "voter intent." The best system is probably one which combines an electronic system with a redundant paper trail -- one under glass where human hands can't start "interpreting" the results.