Pre-election tests in Miami

From: Douglas W. Jones <jones_at_cs_dot_uiowa_dot_edu>
Date: Mon Aug 16 2004 - 10:52:28 CDT

I am back from Florida, after spending a very intense 3 days of work
on pre-election testing in Miami. I need to do some digesting of
the experience, but here are some brief notes:

1) If elections are like a war, pre-election testing in a county the
size of Miami-Dade is a battle in that war. It is intense, even
frightening at times.

2) The audit bugs that led to this spring's controversy about the
iVotronic are put to rest. We observed that the low battery condition
did not corrupt the data on the compact flash card and was properly
reported in the event log.

3) The Unity election management system contains sanity checks on the
data from the precinct that are designed to force human intervention
if the results from a precinct are troublesome. Apparently, this is
a relatively new feature, and unfortunately, during testing, this
caused some grief because our tests included more ballots
on some machines than there were registered voters in the precincts
for which those machines were configured. This delayed the test
into the early morning hours while people scrambled to find out how
to release the hold on the data from those precincts. Releasing such
a hold should be very public and demand explanation, but it shouldn't
be as difficult as it seemed.

4) Hurricanes should not be scheduled during pre-election testing.
They add to tension of an already tense process. Planning for testing
when you're in the path of a massive hurricane is nerve-wracking (we
were in, but toward the east edge of the probable paths predicted on
Wednesday, and just outside that path on Thursday Noon, the deadline
for making the go-no-go decision. Thursday night, Charlie drenched
Miami with one rain band, with tornado warnings, but aside from that,
all we got were sustained winds all Friday, nothing to write home
about.

5) Testing DRE voting machines is really difficult. Even with two
people entering each test ballot, one entering the votes, the other
verifying the votes, mistakes get made. Failure to vote a blank
ballot and voting on the wrong ballot without noticing it were the
most common problems. (The latter, because there were nonpartisan
ballot issues on the primary ballot, so if your test ballot only had
votes in the nonpartisan issues, it was easy to accidentally skip
the partisan races without noticing that they were for the wrong
party.) Working out good procedures for dealing with such errors in
following the test script us a real challenge.

6) On the whole, I'd say Miami did better with these tests than many
jurisdictions, but they didn't perform the tests as well as they'd
intended (state intervention and test planning both contributed to
this), and they could (and should) plan for better testing in November.

As I said to start with, I need to do quite a bit more work on this
before I have anything resembling a final report on these tests, and
of course, the report goes to Miami first!

                        Doug Jones
                        jones@cs.uiowa.edu
==================================================================
= The content of this message, with the exception of any external
= quotations under fair use, are released to the Public Domain
==================================================================
Received on Tue Aug 31 23:17:12 2004

This archive was generated by hypermail 2.1.8 : Tue Aug 31 2004 - 23:17:22 CDT