Tuesday, April 29, 2008

Voting Machines That Can't Count

Apparently, the Sequoia voting machines used in the New Jersey primary can't count quite right. I don't understand how we entrust our elections to companies that don't have to disclose their software source code to be audited. Who knows what kind of mistakes the machine is making when it makes such a basic error? Plus, nobody will ever know but them?

Sequoia's explanation of the problem reveals a kind of idiocy to their software design. Why, when creating software that's sole job is to count, would you make the machine "best fit" any choice, whether by the poll worker or the voter. I mean what if a voted pressed a non-sensical button and the machine just arbitrarily picked a candidate on the person's behalf that it judged to be the "best fit?"

Additionally, apparently, they are blaming further discrepancies on people misreading 9s and 8s for each other. When your machine's sole job is to count votes and print out a verifiable record of those votes, shouldn't you make sure the printed out record is legible? You should choose higher quality output printing mechanisms than the extremely low resolution dot matrix printer that they seem to have chosen. On top of that, you should choose a font with unambiguous characters so that partially printed characters are still distinguishable in almost every case. When an 8 and a 9 are only visually distinct by one dot that the printer has a 10% chance of not drawing, then we have a problem.

While I'm not sure hand counting ballots is any better, having such blatant mistakes certainly doesn't make these election results any more trustworthy.

No comments: