Harvie Branscomb – electionquality.com – 2/17/2019
Georgia State Rep. Barry Fleming states that “Using the new touchscreen-marked paper ballots solves all of the problems associated with hand-marked paper ballots.”
No it doesn’t. And it doesn’t provide the many advantages that hand-marked paper ballots offer.
Rep. Fleming is taking a very simplistic view that ballot marking computers remove uncertainty and exception from human attempts to vote. He is concerned about election judges interpreting marks on ballots (such as with magnifying glass) and prefers computers. Yes, his preferred electronic system silently and conveniently produces a single unarguable vote count number like the existing soon-to-be-banished DRE does. He can legislate that the number is official but that doesn’t mean it is accurate.
A well-designed hand audit of paper ballots that all voters either hand-marked or carefully examined, understood and corrected would find and correct enough outcome-changing inaccuracies. But Rep. Fleming is suggesting hand-marking is outdated and error prone. His straw man about hanging chads isn’t even remotely relevant.
A much anticipated and much needed feature of Georgia’s proposed electronic vote capture devices for all to use is voter verification of whatever the machine prints on paper. Experts warn that verification doesn’t actually happen – voters rarely even glance at the paper. How could Georgia allow a crucial distinction between an unverifiable past and a verified future to be just wishful thinking?
The proposed rerun of touchscreen technology at the polls perpetuates the problems that exist with current DRE touchscreens. The biggest problem is that they are computers – custom built computers. Most of us know about problems with computers – and particularly ones that are rarely tested and rarely used: they cost a lot to buy, cost a lot to maintain, require spares for breakdowns, and will inevitably be in too short supply in some polling places causing lines. Ballot marking computers will be difficult to test and may require calibration to avoid consistent errors. As with DREs, they will be misunderstood by some voters and probably even scare some of the more timid folks away.
And if something is wrong with software in one machine that happens to help some candidate to win, it is likely to infect many or all machines.
Worse yet, skeptical voters will be lulled into thinking they have verified their vote while watching the screen. In fact the vote to be counted is encoded into an unreadable row of lines on a computer image – containing one or more barcodes – that is printed only at the end of the voting session. An average voter probably won’t even look at that paper and if so wouldn’t recognize all of what is written there – even the part that is human readable. Only the barcode is counted, not the human readable text.
Hand-marked ballots are common. In Colorado, not unlike many states, 95% of voters hand-mark ballots. Colorado has recently been called the safest state to vote in. In Colorado we simply have pairs of election judges check enough exceptional physical paper ballots to make sure the inevitable mistakes made by scanning machines don’t affect outcomes. We call those checks human adjudication for the tabulator followed by a risk limiting audit. Those steps resolve any concern that the convenient flexibility of hand marks on paper will cause an incorrect outcome. In real elections there are not many hand marks that a machine cannot interpret and almost none that election judges have trouble interpreting.
Why are those steps used for hand-marked paper? Because human marks on paper are created by natural variations in the human condition – lack of muscular coordination, failing eyesight, unwillingness to read instructions, mistakes made while marking, etc. Sound familiar?
Voters using machines to mark ballots have the same foibles but technology is cleverly designed to leave no trace of those problems. Are those natural and unavoidable human variations actually remedied by the ballot marking device? Certainly not all. We cannot know because the evidence is disappeared – no room was made for it in the electronic memory that prints a bunch of bar codes on paper and many lines of unfamiliar shorthand like “voted yes on amendment 23a.” The voter may have difficulty associating that shorthand with the explanatory text of the question they saw onscreen. In any case, the machine’s version of the countable vote is hidden within a barcode. A vote as if carved in stone in an unrecognizable language covering up any original nuance or mistake by the voter.
Rep. Fleming writes that the new system will remove “the opportunity for human error” when in fact it removes the opportunity to detect, audit and correct for human error. It avoids a simple means to accommodate for human and machine weakness and forces voters to express their vote only through the machines’ limited and not so trustworthy software interface.
To make the new system actually “verified” and different from the unverifiable DRE about to be abandoned, you may be surprised to learn that after machine-marking, it becomes the voter’s unsuspected and undesired but vital responsibility to make sure the machine printed the vote correctly so it can be counted correctly. Neither the system nor Rep. Fleming reveal how important that verification is or how difficult it might be. Since we cannot force voters to take that responsibility, verification is not effective. Any voter that attempts to verify also shoulders the inconvenient task to ask officials to take corrective action if exceptions are actually noticed. Does that sound like Georgia is replacing unverifiable with verified elections? Not to me.
The chads in Florida have been ridiculed as if paper was at fault. Instead the local officials were at fault for failing to empty the chad trays. Meanwhile the paper actually did a heroic job of capturing evidence of voter intent while the machine failed. That is called fail-safety – a beautiful thing. Fail-safety for electronic ballot marking computers is to stop using them and substitute hand-mark paper.
Dedicated people like the memorable Judge Robert Rosenberg of Broward County used a large magnifying glass to interpret evidence of voter intent that only existed because of paper. That’s what we want to happen in a close election. His photo was made world-famous as an example of a bad system but could have been celebrated as a human and paper system recovering from mismanagement. The spin on that photo has done inestimable harm to election recounts and audits. As Rep. Fleming admits, that story led straight to Georgia’s (undesirable) DREs. It is about to serve the same counterproductive purpose again to bring in BMDs , undesirable when used by all. If SB316 is adopted, Georgia won’t need the likes of Judge Rosenberg to perform audits or recounts of perfect ballots printed by software. That is not a selling point from my citizen perspective.
The inevitable rough edges of the voter experience can be revisited during an audit or recount of hand-marked ballots to obtain the most accurate vote count – something that the machines cannot provide – with or without electronic marking. These citizen auditors deserve sincere thanks as first responders who use trustworthy tools such as, occasionally, the magnifying glass. If electronic markers are in use, evidence of original voter intent, expressed either with ease or with difficulty, will be known only through perfect-looking interpretations printed by software developers, known to match the voter intent only if the voter is both prepared and willing to check.
Machines are known to make consistent mistakes in interpreting human expressions – no matter how expensive and well designed they might be. And voters, including those who are probably disabled but not ready to admit it are known to have problems with machines. Are you one of those voters? Do you know someone who is? You might want to ask for a paper ballot in hand and sooner rather than later as the Georgia Legislature makes its decision soon with HB-316. For some details look at my article on the bill: bit.ly/GA-316 .
Harvie Branscomb
I’ve been a credentialed watcher of elections in Colorado since before 2000 – during which time they evolved from hand-marked paper to DRE and dramatically back again. I have served as a citizen election official, a party official, a SOS-appointed committee member all on numerous occasions. I am a software developer. I post my comments on my website: electionquality.com
The original article referred to above is linked here:
New Voting Machines for Georgia | InsiderAdvantageGeorgia | State Rep. Barry Fleming
http://insideradvantage.com/2019/02/15/new-voting-machines-for-georgia/
Monthly Archives: February 2019
Colorado leads on election tabulation audits
But what about transparency for audits so that the public can verify the audit, and go further to verify the election based on access to cast vote records and ballot images, and by visiting the appropriate office, the paper ballot sheets themselves? Yes the risk limiting audit does examine a sampled randomly selected set of paper ballots and compare them to the cast vote records. Cast vote records are the digital recording of the vote after the paper ballot sheet has been scanned, turned into a ballot image, and processed for interpretation by machine, and sometimes by human election judges until the voter intent, already presumably verified by the voter is now recorded in both human and machine readable format. That is the format of the cast vote record or CVR. The sum of the cast vote records is the election vote totals for each contest and this is used to determine the winner. Linked here is a briefing I wrote that talks about the need for anonymity of each ballot sheet and each row in a CVR file, and how to get there and what to do if we haven’t got there yet. Once almost all the ballot sheets are anonymous there should be easy access to check these records yourself over the internet – not quite as good as seeing the paper held in front of you, but close.
Colorado’s Audit center has records of which ballots were selected for audit in each county. During the RLA it should be possible to visit that county and see citizen election judges appointed by the parties review the contents of each randomly selected ballot sheet and input a hand interpretation that is transmitted to the Secretary of State office. There these interpretations are compared to the machine interpretation from the voting system and any discrepancies are flagged for research and further auditing. Ideally all the ballot sheets selected for audit will be anonymous, just as every ballot sheet is supposed to be. So you as a member of the public, part of a political campaign, press etc. will be able to watch the manual interpretation process. The link below is to a paper that discusses what Colorado could do to achieve that kind of anonymity for its ballot sheets. Eventually one hopes that Colorado will provide easy access to the anonymous voted ballot copies in image form and CVRs too, by following some of the suggestions in this paper that could lead to adequate anonymity and deliberate delivery of the election evidence to the public: strategies-for-achieving-anonymity-of-voted-ballot-sheets-images-and-cast-vote-records-for-colorado
As of 2012, Colorado is holding back all the voted ballot versions including those that are electronic at county clerk offices until it is too late for citizen involvement to provide any corrections – that means the end of the deadline for the request of a recount. However, certain interested parties (typically political parties) may obtain special access during a recount. After the deadline for a request for recount, or a recount is over, anyone can receive potentially redacted versions of voted ballots using the Colorado Open Records Request (CORA) law. These would be digital pictorial ballot images or cast vote records (CVRs) produced by the tabulator.
Unfortunately the CORA law actually sets up an adversary relationship between requester and custodian, and custodians tend to take the maximum time to respond and use the maximum available flexibility in producing the records if they actually do. CORA also specifies a criterion for anonymity that the records must meet and leaves much up to the custodian in that regard. I hope that in future Colorado’s Title One, the election statute for Colorado’s legislative, state and county district elections will some day include deliberate transparency for voted ballots and other election records such that access is easy and does not require a conversation, a visit or a payment to each of 64 counties.