Over the the last few weeks, we have learned that in the November Election, registrars have substituted a less effective form of pre-election testing that is less likely to catch errors in ballots or election equipment.
Pre-election testing in Connecticut used to involve feeding about twenty-five hand-voted test ballots of every type to be used by a scanner and checking the results. In general, pre-election testing is not a panacea – it cannot test every case, cannot detect every possible error, and cannot prevent clever hacks for recognizing the difference between a test and a real election. Yet, pre-election testing can detect many errors in ballot printing, memory card programming, or hardware problems.
For the November 2017 election, a new voting machine for those with disabilities was introduced statewide. One of its features is printing a vote on a standard ballot that can be scanned along with other ballots. It would have been advisable to have several test ballots of every type voted using the two voting methods designed for those with disabilities, a touch screen and an interface for sight impaired. That would be quite an undertaking.
Instead of a test of the user interface, officials used a special IVS test function which directly printed out test-ballots, bypassing the user interface. Then they used those test ballots to test their acceptance and results on the AccuVote-OS scanners. Sounds useful and helpful. It is. Yet, apparently, from our discussions that was the extent of testing or the majority of testing of the IVS and the AccuVote-OS in many towns. There are at least two problems:
First, such a test does not completely test the IVS. It certainly tests that the IVS understands the ballots, yet there is no guarantee that either of the interfaces would correctly display and record the votes for each candidate and contest on the ballot. Perhaps, for instance it displays or says the wrong names for State or local offices, such as State Representative, Registrar, or Probate Judge. It would still record the vote in a correct position on the ballot.
Second, such a test does not completely test the AccuVote-OS. The problem is that the IVS does not fill in the bubbles on the ballot in the way a voter is supposed to fill them in. For each vote, the IVS makes a black square about twice as wide as the wide dimension of the oval. Those votes should certainly be counted by the AccuVote-OS. However, their being counted is no guarantee that a voters proper vote would be counted. What if the location of the bubble was incorrectly programmed? A transposition, an incorrect number etc. could cause some bubbles, especially partially filled in bubbles to not be counted by the AccuVote-OS, while all the IVS voted “bubbles” would be counted.
We did not do a formal survey. We talked to several registrars and it seems they did little, if any, testing beyond the canned IVS test.
