This article originally appeared November 3, 2010 on the blog of John Washburn, a software quality professional since 1994 and a software developer from 1985 to 1994. Mr. Washburn has extensive experience in most areas of software quality including test planning, test execution, test tracking, defect analysis, root cause analysis, software configuration management, tool integration and test automation. He is fluent in 4 programming languages: Java, C/C++, PERL, and Visual Basic. He is also a competent programmer in nearly a dozen additional languages.
I was able this evening to prove conclusively that the voting system in Village of Germantown, District 1 is programmed incorrectly. This also proves the pre-election functional testing done prior to the election (aka logic and accuracy testing) pursuant to WI Stats. §5.84, failed to detect this faulty programming. The Wisconsin State Election Board (now known as the GAB) has had since September of 2005 testing guidelines designed to aid clerks to effectively test voting machines. If the GAB had given the guidelines to the Village Clerk and she had designed her test ballots according to those guidelines, then she would have discovered this faulty programming prior to the election.
Until I get the original ballot as part an open records request, here is a sample ballot marked the same way. This ballot has a mark for Republican party preference voting and marks for other ballot lines. According to the statute on elector intent, [WI Stat. §7.50], this ballot is a legally marked ballot were the marks unambiguously designate to which candidate a vote should accrue.
The Diebold/Premier/Dominion AccuVote OS v1.96.6 infrared scanner erroneously rejected the ballot as over-voted. The chief inspector offered to hit the override button, but doing so would have switched my Democratic vote to Republican and my write-in vote to Republican as well.
The trade-secreted, vender-owned programming/configuration found on the removable memory pack erroneously rejected this ballot and that is illegal. Here is why:
Ø In Wisconsin, marks for pre-printed candidates trump marks indicating party preference voting [WI Stat. §7.50(2)(a)]
Ø In Wisconsin, marks for write-in candidates trump marks indicating party preference voting [WI Stats. §7.50(2)(a)]
Ø In Wisconsin, marks for write-in candidates trump marks for pre-printed candidates [WI Stats §7.50(2)(d)]
My ballot was legally marked with votes for the following candidates:
Scott Walker / Rebecca Kleefish
J.B. Van Hollen
Todd P. Kolosso
I was not very surprised the programming/configuration within the removable memory card of the AccuVote OS v1.96.6 infrared scanner broke Wisconsin election law. I marked my ballot in such a way as most likely to discover the programming/configuration error(s) introduced by slothful, lazy, and untested programming. Sadly, I was correct.
I have been to dozens of voting system test sessions and have never seen any of this faux “testing” actually test the voting system software correctly. This is the professional opinion of a software tester testing software since 1994.
The defective programming/configuration was present during an actual election and was NOT caught by the pre-election functional testing.
WI Stats. §5.84 reads [emphasis mine]:
Where any municipality employs an electronic voting system which utilizes automatic tabulating equipment, either at the polling place or at a central counting location, the municipal clerk shall, on any day not more than 10 days prior to the election day on which the equipment is to be utilized, have the equipment tested to ascertain that it will correctly count the votes cast for all offices and on all measures. Public notice of the time and place of the test shall be given by the clerk at least 48 hours prior to the test by publication of a class 1 notice under ch. 985 in one or more newspapers published within the municipality if a newspaper is published therein, otherwise in a newspaper of general circulation therein. The test shall be open to the public. The test shall be conducted by processing a pre-audited group of ballots so marked as to record a predetermined number of valid votes for each candidate and on each referendum. The test shall include for each office one or more ballots which have votes in excess of the number allowed by law and, for a partisan primary election, one or more ballots which have votes cast for candidates of more than one recognized political party, in order to test the ability of the automatic tabulating equipment to reject such votes. If any error is detected, the municipal clerk shall ascertain the cause and correct the error. The clerk shall make an errorless count before the automatic tabulating equipment is approved by the clerk for use in the election.
Because the faulty programming/configuration found on the removable memory card used by the AccuVote OS, the machine failed (during an actual election) to correctly count the votes on a legally marked ballot for the six races shown. In a sane, law abiding world the ballots in the Village of Germantown District 1 would have to be hand counted. The presumption of correctness in WI Stats. 7.51(2)(h) is fatally pierced because an “error in the record is clearly apparent.” That would inconvenience election officials though and the GAB will not allow that.
I could have made a different ballot which probably would have been accepted by the trade secreted, vendor-owned programming/configuration on the removable memory card. But, why should I change my legally marked ballot to accommodate defective voting software? Since, the pre-election, §5.84 testing clearly missed the defect whereby my legally marked ballots and the unambiguous votes thereon was erroneously rejected. What other defects the “testing” miss?
If this legally marked ballot could not be processed correctly, what confidence should I have that a second ballot marked in a way that the ballot is sucked into the front of the machine and deposited in the lock box below the machine was processed and counted my ballots correctly as traveled through the machine?
If that which I can see is clearly faulting and manifestly illegal. Why believe the hidden portions of the system are working correctly or within the Law? In a rational world the answer you shouldn’t.
In Election World though the answer is you must always trust what is hidden. If the public can’t see it, then it isn’t a problem.
It easy to not see something if you studiously avoid looking.
P.S. This is one ballot of 2961 cast. This error rate (>1 in 3000) is very, very much higher than the maximum allowable error rate of 1 in 500,000 permitted under by the Help America Vote Act (HAVA).
See all of John Washburn’s blog entries here. http://washburnsworld.blogspot.com/