Paper trails, Voter IDs, and election machines

I wanted to call attention to another fine article in the "Freedom to Tinker" blog, this one by Dan Wallach.

The issue is related to how voter registration systems should work, and whether state-issued photo ID should be required for registration and for authentication at polls. Dan explains the connection between this issue, government funding for conversion to e-scanned paper ballots, and more.

In addition, I wanted to point out that the confusion and political footbal about voter ID is one example of many reasons why it is so hard for for-profit vendors to create good e-voting products. In this case, there is a lot of confusion, differing requirments in each state, and apparently a good chance of changes in requirements at state or federal levels. It's just one example of how hard it is to nail down a good description of a business process that one wants to partly automate with computing. Voting is like that all over. No wonder it's so hard build a good voting system - it is hard to specify one. (Or at least hard if you follow the traditional tech product vendor model.)

In this case, though, the "business requirements" for voting systems are particularly obscured by politics. The voter ID idea is popular with some people who believe that there is a significant amount of voter fraud, specifically, a person going to the polls, claiming to be another person that didn't turn out to vote, and using that person's vote. In fact, that is hard to prove because of ... no surprise ... e-voting systems! The current technology makes it very hard to collect and publish info about who actually voted, and to detect cases where the record is that a person voted, but the person denies that -- presumably they were defrauded of their vote.

And of course with no market requirement to add features and to use standard public data interchange formats, vendors have no motivation to spend on product development that isn't required. As a result, this potentially valuable info is locked up in proprietary data formats inside closed system products.

Of course, some readers will know the recipe for fixing all that, so I won't belabor the point. ;-)

EJS