Election officials in Myrtle Beach, SC, noted the absence of about 260 votes in the recent election there. Fortunately, the votes were found, and the reason for the error discovered -- and both before the election was certified. It's a good thing that Myrtle Beach election officials were making their lists and checking them twice, because it's quite possible for the omission to have been overlooked until after the election was final. However, the story is a good illustration of how some technology design decisions create the scope for operator error. The basic story is that in one polling place, a poll worker made a mistake in the process of extracting electronic vote totals from a voting machine, using a removable data storage cartridge in a way that's similar to how you might use a USB flash drive to copy some files off of your PC. The problem, however, it that these particular voting machines are designed to be picky -- if you don't use the right cartridge the right way, the data is not copied, and is omitted from the supposedly complete vote totals compiled later (Also, it is not obvious to exhausted poll volunteers when this mistake occurs.)

You might ask, why so picky? Wouldn't it be better to record vote totals in a way that didn't depend on a poll worker not making mistakes at the end of an 18+ hour day? Well of course, put that way, yes. But when you look at the actual details of the way the iVotronic voting machines work, you can see that from a techie perspective, the design is not crazy -- right up to the V-8 moment of realizing that a consequence is that operator error results in lost votes. Those details are not technical, quite instructive, and well explained by voting technology expert Doug Jones:

The electronic cartridge (PEB) is used to initialize the machine in the morning, enable the machine for voting before each voter, and close down the machine in the evening.

In some jurisdictions, a polling place has multiple PEBs: a master PEB for opening and closing the polls, and another used for routine voting. When you do this, only the master PEB has the zero report and the vote totals on it. If you've got several iVotronics at the polling place, and you use a different PEB to close the polls on some of them, the master PEB won't have all the totals; then, when you upload it, you'll be missing votes from some machines.

This is an excellent example of a procedural error of the type that the voting systems could help defend against, but don't. It would be possible to write specs that lead to automatic detection of machines believed to have been deployed for which no totals have been reported. Sadly, we haven't got such behavior in our system specs, and as a result, we chalk such problems up to human error and let the voting system off the hook.

As you can see, once again the devil is in the details -- and thanks to Doug for the infernal info.

-- EJS