Amid heightened concerns about the integrity of the voting process in the run-up to the 2020 presidential election, two election security experts recently quit Verified Voting, a respected election accountability group, in protest. They claim that it has been downplaying security risks in popular voting machines.
Richard DeMillo, a Georgia Tech professor who sat on Verified Voter’s advisory board, just left the group, soon after the departure of UC Berkeley statistics professor Philip Stark, a board member who sent a fiery letter of resignation on November 21st. Stark and DeMillo believe that Verified Voting has been giving election officials false confidence in their voting machines and providing cover for the companies that make and sell the machines.
The accountability group wields a lot of clout, since public officials rely on its recommendations when purchasing expensive voting systems. DeMillo, a professor at Georgia Tech University, has been deeply involved in trying to fix the voting systems in Georgia that saw widespread problems during the razor-thin gubernatorial election of 2018. And Stark designed a vote-verification tool that has been adopted by many states and endorsed by Verified Voting.
Part of the reason Verified Voting is such a trusted organization is that its members are respected scientists and researchers from academia. But both Stark and DeMillo believe that the leadership of the organization, including its president Marian Schneider, has its own agenda and has begun making public statements about elections and voting machines that aren’t backed up by science.
In his resignation letter, Stark accused the group of being on the “wrong side” by approving pricey new voting systems that replace hand-marked ballots with computer-printed ballot summary cards [BMD], the accuracy of which he questions since they depend on potentially insecure software.
“Our message to jurisdictions that buy poorly designed, insecure, universal-use BMD [ballot marking devices systems] should be, ‘We tried to warn you. You need a better voting system,'” Stark wrote. “Instead, we’re saying, ‘Don’t worry: VV will teach you to sprinkle magic RLA dust and fantasies about parallel testing on your untrustworthy election. All will be fine; you can use our authority and reputation to silence your critics.”
In his resignation letter, DeMillo charges that Verified Voting leadership has refused to discuss their concerns, saying that they’ve been “unwilling to have face to face conversations, even when we are in the same city and sometimes the same building.” He continues: “These apparent disconnects have been seized upon and exploited in Georgia and other states to weaken, not enhance, the cause of accurate and verifiable elections.”
At issue is a method of testing the integrity of election results called a “risk-limiting audit (RLA),” which Stark developed, and which Verified Voting has adopted as its “gold standard” method of verifying both elections and the voting machines used in them. An RLA involves manually examining a subset of the paper ballots generated in an election and running a statistical analysis on them to look for irregularities. An RLA can tell with 95% accuracy if the vote tallies are legitimate.
“Because there is software between the voter and the paper, what the paper shows might not be what the voter did or saw”
Since the election interference in 2016, many states and localities have been moving to voting machines called ballot marking devices that record the voter’s choices digitally and also print them on a paper ballot as a backup. And there’s the disconnect. The findings of the RLA depend totally on the assumption that the paper ballots accurately reflect the choices of the voter. Stark and DeMillo warn against making that assumption.
“Because there is software between the voter and the paper, what the paper shows might not be what the voter did or saw . . . on the device,” Stark told me. That’s because the software could be hacked and caused to create a false paper ballot. In close elections, it might take only a small number of these to change the result.
Verified Voting has been breezing over this truth, the two men charge, and has been giving its stamp of approval to both elections and voting systems based on audits that can’t completely rule out problems.
“We need to be paying a lot more attention to the physical security of the paper trail, and creating regulations, procedures, laws, etc. to ensure that the paper trail is trustworthy,” Stark told me. “An audit or a recount of an untrustworthy paper trail is just a distraction from election integrity.”
DeMillo states that one of the main reasons he became involved with Verified Voting was to help fix the widespread problems in the voting systems in Georgia where he lives and works. The state’s 2018 gubernatorial election was tarnished by reports of voting equipment bugs, long wait times, and revelations that the systems were outdated and vulnerable to hacks.
After the debacle, the state ordered 33,100 BMD voting machines made by Dominion Voting Systems, which counties are in the process of installing. The state held a pilot election in six counties November 5, all using the new BMD machines. A week after the election, Verified Voting participated in an RLA. Afterward, both Georgia Secretary of State Brad Raffensperger and Verified Voting president Marian Schneider made public statements suggesting that the RLA confirmed the election results, despite the fact that not all voters had checked that the paper ballots generated from the software correctly matched their choices.
“A risk-limiting audit is a thorough, transparent and structured process that provides solid evidence in the outcome of an election,” Schneider said at the time.
Stark has a different view of the thoroughness of the Georgia audit. “In Georgia and particular, the kind of audit that was done is called a ballot polling audit, and a ballot polling audit doesn’t even check the tabulation at all,” Stark told me. “It just checks whether there is a sufficiently large majority to report a winner and a sufficiently large sample that it’s implausible that somebody else won.”
Raffensperger’s office declined to comment on the record when contacted by Fast Company.
In a written statement, Verified Voting president Marian Schneider said:
“Given the current state of knowledge, Verified Voting opposes the purchase or use of ballot-marking devices for all in-person voters. Verified Voting will always be grateful for the years of service Philip Stark and Richard DeMillo gave to this organization and continue to give to these important issues, including Philip’s critical contributions on risk-limiting audits. With a serious issue like election security comes many passionate voices and differences of opinions. We welcome a robust debate and look forward to being in it for years to come.”
In another problematic election in Northampton County, Pennsylvania, widespread malfunctions were reported in the ES&S ExpressVote XL ballot marking devices used. The machines require a voter to feed a paper form into the machine, then make their choices on a touchscreen. After they push a “Vote” button, the voter is given a chance to examine their choices printed on the form, then push “Cast.”
“All the jurisdiction did was to recount the ballot marking device output and say ‘Ah, look, no problem; we now know who really won,'” Stark told me. “Except they have no idea if that printout actually matches what voters saw.” Stark says it shouldn’t be the responsibility of the voter to make sure that the voting system hasn’t malfunctioned and misrepresented their choices.
Correction: An earlier version of this story said that Verified Voting was involved in, and made statements about, a pilot RLA after the Northampton County, Pennsylvania election November 5. It was not. We regret the error.