For (Digital) Poll Books -- Custody Matters!

Today, I am presenting at the annual Elections Verification Conference in Atlanta, GA and my panel is discussing the good, the bad, and the ugly about the digital poll book (often referred to as the “e-pollbook”).  For our casual readers, the digital poll book or “DPB” is—as you might assume—a digital relative of the paper poll book… that pile of print-out containing the names of registered voters for a given precinct wherein they are registered to vote. For our domain savvy reader, the issues to be discussed today are on the application, sometimes overloaded application, of DPBs and their related issues of reliability, security and verifiability.  So as I head into this, I wanted to echo some thoughts here about DPBs as we are addressing them at the TrustTheVote Project.

We've been hearing much lately about State and local election officials' appetite (or infatuation) for digital poll books.  We've been discussing various models and requirements (or objectives), while developing the core of the TrustTheVote Digital Poll Book.  But in several of these discussions, we’ve noticed that only two out of three basic purposes of poll books of any type (paper or digital, online or offline) seem to be well understood.  And we think the gap shows why physical custody is so important—especially so for digital poll books.

The first two obvious purposes of a poll book are to [1] check in a voter as a prerequisite to obtaining a ballot, and [2] to prevent a voter from having a second go at checking-in and obtaining a ballot.  That's fine for meeting the "Eligibility" and "Non-duplication" requirements for in-person voting.

But then there is the increasingly popular absentee voting, where the role of poll books seems less well understood.  In our humble opinion, those in-person polling-place poll books are also critical for absentee and provisional voting.  Bear in mind, those "delayed-cast" ballots can't be evaluated until after the post-election poll-book-intake process is complete.

To explain why, let's consider one fairly typical approach to absentee evaluation.  The poll book intake process results in an update to the voter record of every voter who voted in person.  Then, the voter record system is used as one part of absentee and provisional ballot processing.  Before each ballot may be separated from its affidavit, the reviewer must check the voter identity on the affidavit, and then find the corresponding voter record.  If the voter record indicates that the voter cast their ballot in person, then the absentee or provisional ballot must not be counted.

So far, that's a story about poll books that should be fairly well understood, but there is an interesting twist when if comes to digital poll books (DPB).

The general principle for DPB operation is that it should follow the process used with paper poll books (though other useful features may be added).  With paper poll books, both the medium (paper) and the message (who voted) are inseparable, and remain in the custody of election staff (LEOs and volunteers) throughout the entire life cycle of the poll book.

With the DPB, however, things are trickier. The medium (e.g., a tablet computer) and the message (the data that's managed by the tablet, and that represents who voted) can be separated, although it should not.

Why not? Well, we can hope that the medium remains in the appropriate physical custody, just as paper poll books do. But if the message (the data) leaves the tablet, and/or becomes accessible to others, then we have potential problems with accuracy of the message.  It's essential that the DPB data remain under the control of election staff, and that the data gathered during the DPB intake process is exactly the data that election staff recorded in the polling place.  Otherwise, double voting may be possible, or some valid absentee or provisional ballots may be erroneously rejected.  Similarly, the poll book data used in the polling place must be exactly as previously prepared, or legitimate voters might be barred.

That's why digital poll books must be carefully designed for use by election staff in a way that doesn't endanger the integrity of the data.  And this is an example of the devil in the details that's so common for innovative election technology.

Those devilish details derail some nifty ideas, like one we heard of recently: a simple and inexpensive iPad app that provides the digital poll book UI based on poll book data downloaded (via 4G wireless network) from “cloud storage” where an election official previously put it in a simple CSV file; and where the end-of-day poll book data was put back into the cloud storage for later download by election officials.

Marvelous simplicity, right?  Oh hec, I'm sure some grant-funded project could build that right away.  But turns out that is wholly unacceptable in terms of chain of custody of data that accurate vote counts depend on.  You wouldn't put the actual vote data in the cloud that way, and poll book data is no less critical to election integrity.

A Side Note:  This is also an example of the challenge we often face from well-intentioned innovators of the digital democracy movement who insist that we’re making a mountain out of a molehill in our efforts.  They argue that this stuff is way easier and ripe for all of the “kewl” digital innovations at our fingertips today.  Sure, there are plenty of very well designed innovations and combinations of ubiquitous technology that have driven the social web and now the emerging utility web.  And we’re leveraging and designing around elements that make sense here—for instance the powerful new touch interfaces driving today’s mobile digital devices.  But there is far more to it, than a sexy interface with a 4G connection.  Oops, I digress to a tangential gripe.

This nifty example of well-intentioned innovation illustrates why the majority of technology work in a digital poll book solution is actually in [1] the data integration (to and from the voter record system); [2] the data management (to and from each individual digital poll book), and [3] the data integrity (maintaining the same control present in paper poll books).

Without a doubt, the voter's user experience, as well as the election poll worker or official’s user experience, is very important (note pic above)—and we're gathering plenty of requirements and feedback based on our current work.  But before the TTV Digital Poll Book is fully baked, we need to do equal justice to those devilish details, in ways that meet the varying requirements of various States and localities.

Thoughts? Your ball (er, ballot?) GAM | out