I write software and work with big databases. A few years ago, I became very interested in election fairness and security. As a personal project, a few friends and I got together to study our electoral system through the data that defines it. Because of this involvement, I was asked to be one of the nine people who evaluated and scored the proposals for new Shelby County voting systems. The evaluation was a fair process, and the vendor that I believe the Election Commission chose is the one that I ranked highest on objective criteria.
I went into the process with varying levels of concern about each of the vendors based on derogatory reports I’d read, from election security advocate Jennifer Cohn and others, but ultimately I based my decision entirely on my understanding of the systems that I saw demonstrated and that I was able to examine over three days in February.
All of the systems we evaluated were acceptable, but not one of them was perfect. Any one of them would do the job for Shelby County, and each one had some advantages and some disadvantages over the others. We could all wish for some fourth alternative that combines the best features, but that doesn’t exist. Every software purchase decision is a compromise. Any of the systems proposed here is a big improvement over the current machines.
After the vendor demos, I studied the proposals and my notes and tried to balance the competing advantages and disadvantages. Ultimately, my score was based on the aspects of the system that I understand the best. I agonized over the decision. I was late turning in my scorecard. I hope that I made the right choices.
I was the only member of the group who is a practicing software developer. Two other evaluators, from Shelby County IT, work in networking and system administration. While I have some knowledge of voter registration databases and the data aspects of elections, I’ve never been a poll worker and my only experience with the polling place operations is as a voter. I based my score on my understanding of the functional and design qualities of the software components (and not on the portability of the carts or the cost of the paper, for example).
One question I asked each vendor was “If Shelby County chooses your system, will you make the source code available for code review and audit?” Each vendor answered by saying no previous customers had ever asked to review their source code. (That concerned me — this kind of diligence is appropriate for this essential election component.) Each vendor then went off to call someone and came back with a variation of “Yes, but you have to sign an NDA” or “Yes, but you will have to visit our office to see it.” The vendor I scored highest overall was also the most willing to allow a code review. I hope that the final contract will require that, and I would be delighted to join a team of developers from Shelby County to conduct that review
Each vendor submitted two options: one for Ballot Marking Devices (BMDs) and one for hand-marked paper ballots. The BMDs work similarly to the voting machines we have now, but they produce a sheet of paper printed with the voter’s choices. The voter then inserts the sheet into a scanner. It is scanned and falls into a locked bin. With the other option, hand-marked ballots, a printer produces a paper version of each voter’s ballot. The voter fills it out and inserts it in the scanner. If there are no stray marks, if the bubbles are all filled in neatly and the voter didn’t mistakenly vote for two candidates in the same race, then the scanner accepts it. If there are errors, the scanner kicks it out and the voter has to get a fresh ballot and destroy the old one.
The BMD option would eliminate the problem of mismarked ballots and lead to fewer errors. I think that’s why Linda Phillips prefers it. Jennifer Cohn has argued that BMDs introduce security risks because, while the voters’ choices are printed in text on the ballots they mark, the scanners actually read bar codes that encode the choices and are also printed on the ballots. I don’t think this is a problem. The paper ballots still show the choices in text and that can be audited and used in recounts. Either option will work and can be verified.
I prefer hand-marked ballots by a small margin. The problem of mismarked ballots can be mitigated with clear instructions and voter education, and the process is more transparent to the public. It’s not enough to have an honest system — it needs to be transparently honest. While I disagree with Linda Phillips’ preference for BMDs, I respect her arguments. In my experiences with her, she’s been intelligent, effective, and honest.
The evaluation phase of this purchase decision that I participated in was fair and thorough, but the secrecy of the process has been a problem. Controversy and lack of trust in the system can discourage people from participating. The new system should be good, and I hope many new voters will use it. Everyone should vote.
Doug Sims moved to Memphis in the fourth grade. He works remotely as a software engineer for a large media company. In non-pandemic years he plays Ultimate Frisbee.