I thought I'd share a comment and response I got about trusting software to count votes. The comment was a very sensible one, though a mis-perception: that TTV is suggesting that software should be trust to count vote correctly. Not so! Here is the true but less simple story.
- Many election officials want to conduct elections with paper ballots.
- Most of those election officials want to count paper ballots using optical scanning and analysis software.
- Most of those election officials conduct statistical audits, in order to mitigate risk that the tabulation software malfunctioned in a way could have effected the election result.
In other words, the latter group of election officials don't trust the software to do the vote counting right, and use selective hand-count audit in addition to software counting.
- TTV development of scanning/tabulation software does not depend on the election officials' choices on how to conduct audits as part of an optical scan/tabulation method.
In other words, we don't make any assumptions about whether or how people trust software, and what additional non-technological steps they take to mitigate risk. To repeat what you may have heard me say before, we just make the technology; we don't tell the administrators how to deal with the risks that they manage, but we do listen to them to make sure that we're making technology that they can manage in the way that they want to. If their audit scheme can be improved by new features of the software, then we want to learn enough to provide features that are truly helpful.