Why we need to work for verification rather than throwing out the machines

The Facebook group Wisconsin Election Integrity has lately been dominated by two people: me and an advocate for hand-counted paper ballots. That school of election-integrity activists advocates for junking voting machines entirely. They believe verification would be as thoroughly fraudulent as they believe electronically counted elections to be and would therefore serve only to create even more voter complacency.

Advocating exclusive reliance onjunkedVotingMachines.jpg hand-counted paper ballots (HCPB), they demand that public, transparent hand counts be conducted in every polling place immediately after polls close. That speed is necessary to provide as little window of opportunity for tampering as possible. Electronic tabulators must not be used, they argue, because they count the votes inside a black box—that is, by one computer programmer acting alone in secret.

I usually try not to argue with HCPB-only advocates for two reasons. First, they are correct that their vision could produce accurate election results every time, and second, we all need to keep our eyes on the prize. We share the same goal—verified accurate election results—so this is an argument over means, not ends. As Bill Moyer explained in “Doing Democracy,” in every successful social reform movement, you can see people playing at least four different roles ('citizen,' 'rebel', 'change agent', and 'reformer'), and “dissension among them…reduces the movement’s power and effectiveness. Activists need to become allies with those playing other roles, since cooperation and mutual support will enhance the likelihood of success.”

Avoiding fights with people who are (or should be) big-picture allies is one thing. Refusing to explain yourself is another, and I think it’s past time that someone clearly laid out the rational case for demanding post-election verification of voting-machine output instead of demanding HCPB alone. So here goes.

1)      Verification of voting-machine output is hand-counting ballots, at least in those jurisdictions that use opscan machines, which is most of the big counties in Wisconsin. So this debate is actually a squabble over whether the hand count or the machine count should be accepted as definitive when they disagree, and to a lesser extent, how many of the votes we need to hand count to make sure the results are certified for the correct winner.

In Wisconsin at least, the HCPB advocates have already won the (non-existent) battle over which method produces more credible results, even if they don’t realize it. It’s not even up for debate. Statutes governing recounts already establish (and nearly every election official you can pin down agrees) that when an official hand count (not a citizens' audit) and machine count disagree, it’s the hand-counted result that prevails

Once you’ve established that transparent human-counted results are more credible than black-box machine-counted results, it becomes apparent that doing both is more secure than doing either one alone. Even if ballots are fully hand-counted before dawn on Wednesday, obvious security advantages are created by having voters insert their ballots on Tuesday into a machine that automatically preserves a digital image of that ballot.  And knowing that humans cannot ever be trusted implicitly, it is also obvious that you might as well get a preliminary electronic count when they cast their ballots, so that an election thief must jump over two hurdles (that is, to rig both an electronic system and a human one) instead of just one.

2)      Verification would catch unintended miscounts, because nobody would be trying to hide them. It’s possible that the majority of electronic miscounts are accidental rather than deliberate—due to human programming or set-up error or to randomly occurring mechanical or electrical issues. (I would say it's likely most are inadvertent, but cannot prove it because no one checks for the sorts of miscounts that hacking would create, so every miscount except the obvious inadvertent ones goes undetected.)

Since no one expects the accidental miscounts, no one would sabotage a verification to cover them up, and they would be detected readily, in much the same way that recounts discover errors more often than not. Those who reject routine verification because they believe clerks often want to cover up fraud are, in practice, proposing we allow all the accidental miscounts to remain undetected because we shouldn't trust the clerks to reveal the fraudulent ones. I am not willing to do that; I'll take what I can get when it comes to detecting and correcting miscounts.

3)      Verification would deter or detect fraud that the local officials were not in on.  National experts believe the most likely and most dangerous form of electronic election fraud is one in which a single programmer makes illegitimate changes to the vote-tabulating software at some time while it is not in the control of local elections officials.  John Washburn has also convinced me that it would be easy for someone to insert wireless communications capability into Wisconsin’s voting machines—again, before the hardware comes into the possession of any local elections official. This capability would go undetected by our election officials because they never look for it.

Wise criminals never involve anyone in fraud who does not need to be involved, and neither of these frauds needs to involve any local elections officials. So when elections are hacked, it's entirely possible the local elections clerks aren’t in on the scheme. Therefore, they have no reason to sabotage verification.

The only time a local elections clerk would want to sabotage a post-election verification would be when he or she is in on the fraud, and that clerk's resistance would then call attention to a problem, if the rest of the world knew what a decent post-election verification looked like. No one does now, because we don't do them. 

4)      Hand counts have their own problems. In the real world, it’s questionable whether relying entirely on hand counts would give us much more confidence in election results than relying entirely on machine counts. (To repeat: I think we need to do both full machine counts AND sufficient hand-counts to verify the outcomes.) Sloppy chain-of-custody practices come to light with every recount, and—I’ll be frank—I’ve never fully understood why people expect hand counts to be transparent enough. I’ve observed about six, and there wasn’t one of them where I could come close to seeing what the hand-counters saw except intermittently. The observers at Scotland’s referendum hand count were mightily frustrated by their inability to observe effectively, and what they did observe wasn’t reassuring.  

And then there’s the question of who, really, would come out to count or observe a hand count during the wee hours on Wednesday morning following Election Day? I have been greeted with “You’re the first citizen ever to show up to observe our public test!” at literally every single one of the dozen or so pre-election voting machine tests I’ve observed, so I can vouch for the fact that no one attends those--and they take only about two hours during the daytime! I’m all for making election work something like jury duty, but it would take years of work to get that legislation passed. That doesn't mean we shouldn't start now, but that brings me to…

5)      We could get verification in place by November 2016; we don’t have a prayer of HCPB by then. I won't stand in the way of those who want to work on getting the voting machines thrown in the junk yard. But there is no way on god’s green earth that we’re going to get state law changed to require HCPB between now and the critical 2016 elections.

How many more elections are we going to let be decided by unverified black-box output, when the most basic common sense tells us how incredibly careless that is? I cannot speak for other states, but if enough Wisconsin citizens pressured their county clerks and county boards of canvass hard enough between now and, say, January of next year--especially if even one county clerk is willing to show leadership and vision by adopting such practices--we could have fairly decent post-election verification policies in place in the critical counties in time to protect the 2016 elections from electronic miscounts. Truly, there is nothing in state law or GAB written policy that forbids it—it’s all a matter of the county clerks’ discretion and voters’ interest.

I could go on, but those are my best arguments.

Someday soon, I’ll get around to laying out the case for verification to its other sometimes-hostile allies: Those who argue that our energy should be directed at improving the technology rather than wasting time with a low-tech, unsexy chore like verifying output.


Showing 4 reactions

Please check your e-mail for a link to activate your account.
  • commented 2015-06-07 09:55:44 -0500
    Thanks, Brent. You and I have exchanged comments on this in other forums in the past, and I’m assuming we’re still in agreement: Open-source code would be an improvement. It would deter some fraud. It would also make most of the rest of deliberate electronic fraud detectable if anyone with expertise, credibility, and authority takes the time to verify the integrity of the software. If we install open-source software and no one with the authority to correct alterations ever checks, its advantage over proprietary software is moot.

    And even if our elections officials go to open-source software AND adopt the management review necessary to earn its benefits, they will still need routinely to compare the electronically tabulated totals to the evidence on the ballots, because examination of the open-source software before, during and after the election cannot detect mistakes in setting the machines up for the election (e.g., the missing votes in the Stoughton and Medford miscounts); mistakes in operating the machines (e.g., the type of miscounts described by the Michigan consultant in my earlier post titled “It happens all the time”), or miscounts by Election-Day machine malfunction (e.g., the counted votes in the Stoughton miscount, the Bronx overheating miscount.)

    There’s just no substitute: Whether we’re calculating vote totals, bank balances, the cost of a pile of groceries, or property tax bills, and no matter what system we use—proprietary, open-source, or homemade— whenever we use computers, we must either check the output to make sure it is correct or tolerate the risk of taking action on inaccurate output.
  • commented 2015-06-06 19:10:50 -0500
    There little chance to “verify” a voting machine that is running secret software as a ’bad actor " can manipulate the count and still show what appears s to be reconciliation . If by verifying you mean using open source systems.. then we are in agreement. remember the "first count " is the most most meaningful count.. as usually always accepted and certified
  • posted about this on Facebook 2015-06-02 15:19:55 -0500
    Why we need to work for verification of voting-machine output rather than for throwing out the machines.
  • posted about this on Facebook 2015-06-02 15:19:49 -0500
    Why we need to work for verification of voting-machine output rather than for throwing out the machines.