Recently the voting machines down in Florida seemed to be causing some problems. Last year I had the opportunity as part of World Usability Day to play around with two of these DieBold voting machines. These machines had some major problems in the UI design. These problems alone could easily cause someone to be confused and inadvertently vote for the wrong person but the issue that really stood out was the touchscreen.
The issue with the touchscreen was that it required to be calibrated. If the system was calibrated correctly the location of your touch would match the location pressed on screen. But if the screen was not calibrated you could be way off. This was a known issue two years ago, I'm really surprised it hasn't been fixed.
Unlike a traditional mouse the touchscreen doesn't display any-visual feedback of where the computer believes your finger is. The result is that the computer cursor can be far from the actual physical location.
This is a great example of why real world testing and usability is needed. Given a simple task 50 people who are asked to vote for A and 50 people who are asked to vote for B you should have a task success rate of 99% or higher and you should be able to verify that 100% of the results are correct.
Although it's possible that there is a voting conspiracy I believe it's far more likely that the voting machines have bad UI and even worse touch screens.