More Twisted Logic: How Voting Machines Eat Pretzels
In a recent posting, I tried to give a flavor of the pretzel logic that derives from the crazy quilt of state-specific rules for counting ballots — with a particularly notorious case from straight-party voting. The next question is, given this crazy quilt, what is a voting system supposed to do? Before answering that today, let me recap what a voting system is, for some more recent readers. Loosely speaking, a voting system is a set of hardware and software that are used for casting and counting ballots, preparing for casting/counting, and completing the tabulation process after casting/counting.* Components can include paper ballot scanners, direct record devices, tabulating software, etc.
Now, back to the problem, and the specific case of straight-party voting, with a ballot marked for a straight-party vote of a particular party, and with an “emphasis” vote for that party’s candidate for a top-ticket contest, and also an “over-ride” vote for another party’s candidate in a down-ticket race on the other side of a 2-sided paper ballot. What’s a counting device to do? As we said before, there are various different state-specific rules about what to do.
I bring up this case to illustrate a couple key ideas about our development of the voting system components of the TTV system:
- the construction of ballot-counting software is much more that the exercise in the digital image processing functions that we’re working on now, and
- one size does not fit all.
As a result, we have several additional tasks for the voting system. We need to incorporate the notion of configuration data that encodes state-specific rules for interpreting marks on a ballot. And we need to implement an election-preparation process that to the greatest extent possible automates these choices, informs the election officials, but does not require humans to make explicit decisions which of several interpretation methods to use — because after all, this choice is not at the discretion of local election officials. (Note to election ninja readers: I wouldn’t be surprised to hear about a case where it is a local decision!). And if that sounds complicated enough, I won’t even go into even more squirrelly cases of ballot interpretation that involve disputes over whether a mark is “sufficient to be a legal mark.” One squirrel a day is enough!
* Details for the keeners: In the TTV system we’re building, we have paper ballots for manual ballot interaction, an accessible balloting device for computer-aided production of a marked ballot, and a precinct ballot counting device used for interpreting ballot marks and counting votes at the precinct level. There is a similar device for central processing of mail and provisional ballots, and another central device to tabulate results by consolidating these various counts. On the preparation end, there is a device builder that creates each election’s instance of the other devices. Rounding out the tail end of the process is the TTV Auditor, which consolidates audit logs of the activity of all the devices.