Florida recounts are excluding millions of voters. At this writing, there are two recounts underway in Florida: not the Senator and Governor contest recounts, but two different recount processes for both; two different types of counties; and two different types of voters. One kind of recount has been covered well in the media: a 0.5% margin and machine re-scans of paper ballots, and a 0.25% margin and hand counts of paper ballots. The other kind of recount is, at worst, a sham, and at best consists of efforts of local election officials doing what they can for voters not fortunate enough to cast a paper ballot.
Let’s start with the two types of counties: those that still use paperless voting machines like Miami-Dade with nearly 3 million of Florida’s 16+ million people; and counties like Escambia, Broward, and Okaloosa where on Election Day voters mark paper ballots that are counted by optical scanners and software.
Voter Protection: Humans Fill Software Gaps
For those all-paper-ballot counties, the recount process sounds like a good way to protect voters from machine errors. All the paper ballots are counted normally, but if the margin in a contest is less than 0.25%, it is time to bring in the people to do the fine tuning, and find votes that the software might have missed. Let’s use the Senate contest of Scott vs. Nelson to illustrate.
One kind of missed vote is an over-vote. For one paper ballot, the software may have noticed an emphatic and clear mark for Scott, but also some ink that appears to be a mark for Nelson. In that case, the software will record an over-vote, a case where more than one mark was detected, and as a result, no vote is recorded for any candidate. But in a human review, the ballot counter may consult Florida election law, scrutinize the ballot, and decide that the ink near Nelson’s name doesn’t rise to the level of a legal mark. The result: a vote for Scott where the software recorded no vote.
Another kind of missed vote is an under-vote. The software looked for a mark next to Scott and next to Nelson, but found none. A human review might agree that there is no indication of a vote for Scott, but a wobbly circle about the bubble for Nelson does match what state election law says is a mark that qualifies as a vote. The result: a vote for Nelson where the software recorded no vote.
This sounds great! We use software to do a first-approximation of what human counters would do, applied to a large number of ballots, but if the count is close, we bring in people to apply their brain-power to every ballot, to make sure that every voter’s every vote was properly recorded. Voters are protected from software misinterpretation of ballot marks that, in aggregate, could award a contest to the wrong candidate.
Unequal Protection: Digital “Re-Count”
But in the counties with election-day paperless voting, the story is not so encouraging. In Miami-Dade for example, a voter stands in front of a 12-year old machine, manufactured from 16-year old parts, quaintly called an “iVotronic”. They look at a screen version of the same things that a paper ballot voter saw, indicate their choices, confirm their choices, and — poof! Their ballot disappears in a puff of logic, and their votes are added to vote totals stored as data in this antique computer.
The same misfortune happens to other voters using that same machine, and eventually all those vote tallies are copied from the iVotronic to a central system, combined with other iVotronic votes, and also the votes recorded by scanning paper absentee ballots.
At recount-time, the absentee paper ballots go through the same process of re-scan and hand-count. But those paperless voters? What protection do they get? None. The iVotronic may have incorrectly recorded a voter’s choices, but we’ll never know. For these voters, there is no human correction of software errors.
Digital Re-Count: Anything Learned?
But more than unequal protection, the paperless voting machine “recount” process is even more weird. Humans re-do the process of getting the vote tally data off of the machines, and adding up all the tallies. We’d expect the totals to be the same. But what if they are different? Would that mean that some votes were erased or some added? Or perhaps the initial data off-loading process had some human error or technical glitch, and the new, different totals are the correct ones. How could we know which is the case?
The answer is: we can’t. Too many paperless voting machines, too many people with access to them, no way for a county election official to swear in Court that every single one was completely controlled with no chance for any tampering, not even the completely invisible kind with a person carrying one of the machines, while having a good sized magnet in a pocket to scramble some of the data.
The End Result: Faith-Based Paperless Recounts
When the recounts are done, our hats off to the hard-working local election officials who will have dutifully ground through the process, and refined the vote totals to correct for perhaps a few thousand cases of software misinterpretations that might have affected the election result. And off to the side, millions of paperless votes with no way knowing if the software recorded the votes correctly, or if the digitally stored votes were protected from modification.
As a result, at the beginning of recount process, our confidence in the election results was only as good as our confidence in antique paperless voting machines’ correct operation for millions of voters. After the recount, and after the human correction of hundreds or perhaps thousands of votes, we still have the same millions of paperless votes, and the same confidence that the election results are only as good as our confidence in the paperless voting machines.
From there, we will still have to take the election results on faith, and add some additional faith that Florida will finally jettison all the paperless voting machines, and use a simpler, more cost-effective, easier-to-protect method of counting paper ballots and routinely having humans cross-check the totals in case of computer error.
— John Sebes