Video the vote recorded this video in West Virginia of a county clerk showing problems with an ES&S voting machine. He first shows what happens when the machine is not calibrated, then calibrates it and we see it mess up again (it turns out the error after calibration was just confusion on the part of the clerk — which points to another usability problem, in my opinion).
Wired has more context and a full description of the problems: Video: Recalibrated Machine in W. Virginia Appears to Record Vote Inaccurately.
The county clerk here seems oddly okay with the problems. He's okay with having to recalibrate the machine every day and he actually seems a bit defensive about the machine, which goes to show that people will learn to put up with anything. He is also defensive because he wants to disprove a voter's claim that the machine misrecorded an Obama vote as a McCain vote — his proof is that, when "out of calibration," the machine will misrecord an Obama vote as a Chuck Baldwin vote! (Chuck Baldwin is the Constitution Party's candidate… I hadn't heard of him either.) And as the second video (below) makes clear, the exact error is somewhat random — in this case it is a function of the artificial miscalibration he performed for demonstration.
Some thoughts to tie this into HCI/usability a bit:
- people will try to form a consistent mental model for your system, even if it's broken,
- design and usability need to take account of the whole ecosystem (did ES&S really expect for these machines to need recalibration by poll workers so often? do those poll workers know how to do it?),
- fix the bugs first. These systems really don't look ready for production (or even for usability testing).
Update: Video the vote has now posted the complete, unedited video.