Saturday, October 28, 2006
Humboldt Logic and Accuracy Testing (Guest Blogged By Kathryn Hedges)
Kathryn Hedges is a long-time member of the Voter Confidence Committee who comments frequently here at WDNC and whose writing I have previously highlighted in the GuvWurld Blog. On Friday she attended the Three-Card Monte Dog and Pony Show, also known as the Logic and Accuracy (L&A) Testing conducted by the Humboldt County Elections Department on the County's voting equipment. What follows is her lengthy report with a few editorial comments inserted by me - Dave Berman
* * *
On Friday morning, October 27, 2006, I attended the Logic & Accuracy public "test" with Lindsey McWilliams. (He actually just demonstrated the process for one Accuvote optical scanner. It will take the rest of the week to test all the machines, and they don't even test all combinations of ballots possible at a polling place--just each memory card with at least one ballot.) Also attending were 3 poll workers and a reporter for the North Coast Journal (whose name I don't recall). A reporter for Channel 3 taped some video footage of poll tapes printing and interviewed Lindsey.
I arrived slightly late, but I hadn't missed much. Lindsey started by gathering us in a conference room to explain the different levels of elections in California (Federal, statewide, county, city), the types of races, and how these relate to the number of different ballots used in a given election--even at the same polling place. He went on to tell a story about a poll worker (who's now on the Arcata City Council) who inadvertently gave people ballots for the wrong precinct and is now immortalized in poll worker training.
**The L&A 5 Matrix**
Getting along to the actual testing, he explained the way the test ballots are filled out for what's called an "L&A 5" setup. The reporter, bless his heart, wanted Lindsey to explain how everything worked. Interestingly, Lindsey prefaced his explanation of how the L&A 5 test works with the statement "There is an element of faith in how it works..." How the test ballots are set up actually makes sense; what I see as a leap of faith is whether test mode actually reflects election mode. [Secret vote counting also amounts to faith-based voting. - DB]
So, what does this cryptic "L&A 5" mean? Well, it's no great leap that "L&A" is short for Logic & Accuracy. The "5" refers to the pattern of votes allocated to candidates in the sample deck of 30 ballots. Candidate A gets 1 vote, B gets 2 votes, and so on up to Candidate E getting 5 votes. So when the totals print, this should make a nice pattern of A = 1, B = 2, etc. "Should," as in "theoretically" is the operative word here.
**Visiting Election Central**
Finally it was time for Lindsey to lead us down to the basement. Along the way, he answered questions about how many observers were required for the testing by law (none) and how many usually show up (varies from none to more than 20 after the 2000 election, when a PoliSci class from College of the Redwoods made it a field trip).
In the testing area, we were surrounded by stacks of Hart Intercivic ESlates and shelves full of Diebold Accuvote terminals in laptop cases. The Accuvote terminal for the public testing sat on a conference table next to boxes of test ballots, with a phone cord dangling towards it from the ceiling beams. Off to one side, our account representative from Hart was busy doing stuff on a PC next to the GEMS server. Past us in the adjacent room, employees sorted absentee ballots.
**Detour for Absentee Ballots**
First, he explained the computer-assisted system for verifying signatures on absentee ballots and answered questions about absentee ballot rumors. In the last election, two elderly people accidentally voted early by absentee ballot, forgot they voted, and voted in person on provisional ballots. Lindsey found their original ballots in 5 minutes each, the system works so well. [VoteRemote is the Diebold signature verification system that requires no state or federal approval. It can be calibrated to require varying degrees of matching precision so that fewer or greater numbers of voters are disenfranchised. - DB]
**Accuvote and GEMS**
He asked us to randomly select a precinct for the test. The reporter picked one, and Lindsey opened the corresponding box of ballots. He turned on the Accuvote and invited us to get close enough to see the two-line LCD status display that walks users through the test. I didn't take exact notes of the process, but basically you start by running through a blank ballot and a ballot with all bubbles filled in to "train" the machine. He demonstrated that the scanner can read ballots inserted in any orientation--but it will reject one for the wrong precinct. Sometime along here, a reporter from Channel 3 came in and set up a camera.
Then you print a zero tape that lists all the races with zero votes in each one. Finally you run the "L&A 5" deck of ballots through, which are pre-printed with TEST TEST TEST in red letters above and below the reading area and have votes bubbled in according to the pattern I described. Each observer got to feed it a few ballots. Then he closed out the test and printed a poll tape. The reporter got some footage of poll tape printing out, and Lindsey made a snarky comment about Some People who get all upset about interpreted code--"All the interpreted code does is generate this list." (And exactly why isn't that important?) I asked if that was the poll tape to be posted at the polling place and seemed to hit a sore spot with Lindsey. More on that below. [Actually, interpreter code scrambles the information that humans could otherwise read on a ballot, into AccuBasic, the secret and proprietary language that only Diebold can read. It is forbidden by federal regulations, which are required to be met as a condition of CA state certification. - DB]
The next step after an election is sending the results to the central GEMS tabulator, so he demonstrated how that works. Poll workers plug a normal phone cord into the Accuvote and go through the script to connect to the modem at the Elections Office. The Accuvote sends the data over an analog phone line to the AT&T switching station, where telco equipment converts the signal to the digital trunk line for the Elections Office. Inside the building, in a locked cabinet somewhere in Sempervirens Mental Health, this connects to the analog line going to the GEMS tabulator. He explained the different safeguards in this process. First, he said it would be impossible to decode the signal if it were intercepted after the digital interchange. Second, he said that if the Accuvote didn't communicate properly with the GEMS because someone intercepted the signal or tried to send false data (to make a long story short), this would generate error messages and people at both ends would know there was a problem. (Whether they'd know it was malicious rather than operator error or a blip in the phone line is something I didn't think to ask at the time.) [The Government Accountability Office (GAO) has a different view of the security of modems, citing a report by RABA Technologies that shows control of GEMS can be obtained via a modem hack. - DB]
Next, Lindsey printed out the GEMS totals and we had a couple of surprises--first of all, the 1-2-3-4-5 pattern doesn't come out so neatly because GEMS does not list the candidates in the same order as the ballots, or in the order of the winners (which would be the reverse order from the ballots). I didn't look at it well enough to figure out the pattern. We were too busy noticing that the maximum number of votes wasn't 5 per race--it was 7. After perusing the results a bit, Lindsey decided the printer had made up an L&A 7 instead of an L&A 5 without asking him, and said he'd complain to the printer.
While we waited for the Hart rep to finish setting up the ESlate for us, Lindsey chatted quite a bit about watching the displays update on GEMS as results come in, how the six modems rollover incoming data calls, and how he wished he had more analog phone lines to test them with. (The building is using nearly all the analog lines for fax machines.) Then the TV reporter wanted to interview Lindsey, so we waited a few minutes for that. The North Coast Journal reporter used that time to ask who we were and why we were there.
**The Infamous $800,000 ESlates**
Everyone was pretty curious about the Hart Intercivic ESlates; in fact, one of the poll workers has a disabled son who plans to use it. Lindsey demonstrated one and let us play with it a bit, then explained that he would conduct the actual testing later in the day--we could stay if we liked, but he hinted it would be a waste of time. The tests will be less extensive than for the Accuvote, since no results will be transmitted to GEMS from the ESlates. I managed to annoy him by mentioning the VotePAD after his explanation of how much trouble it is to transcribe votes from ESlates onto ballots for scanning in an Accuvote to get them into GEMS. More on this below.
Everyone else loved the ESlate, but I had a hard time remembering when to use the wheel to advance or back up and when to use the arrow buttons. I guess they are not meant for people with ADD.
Apparently, part of why the e-Slate is so expensive is that the units the voter uses are basically dumb terminals and must be connected to a central control (Judging Bench Controller, I think it's called) at the polling place. Ordinarily you'd have a bunch of terminals connected to the central unit, but here we have a 1:1 ratio.
***News and Gossip***
**Poll Tape Posting Not Required, Says Lindsey**
Probably the biggest news of the day was that Lindsey said the poll tapes will NOT be posted. When the Accuvote printed out a poll tape for the test, I asked if this was like the poll tape to be posted at the polling place after the results are tallied. He got fairly annoyed and claimed this only applied to a system that has since been retired. His other excuse was that when they used to do this, people would steal the tapes, then the next people who tried to check them after that would get upset they weren't there. Conclusion: Don't bother posting poll tapes at all. (So why can't you tape them inside a window? I'm not familiar with all the polling places in Humboldt; is Arcata City Hall the only location with glass by the entrance?) [The CA Election Code is HERE. Scroll toward the bottom and see section 19384 requiring poll tapes be posted. - DB]
**HAVA and ESlates: Disabled Votes Counted Last**
Although HAVA decreed that disabled voters must be able to cast votes without assistance at any polling place, Humboldt County was backed into a corner and bought a system that's incompatible with their Diebold Accuvote and GEMS systems. So disabled votes will be transcribed at the Elections Office onto scannable ballots to get them into the system. Although I agree with Lindsey this is more honest than manually editing the GEMS totals to add the ESlate totals, it also means that votes cast by disabled voters on ESlates will not be counted along with other votes at the precincts. I'm not sure how long it will take officials to get around to transcribing these ballots from the paper tape in the ESlates. They have a system where poll workers will mark ESlate equipment that was actually used, so those units will be checked first. Luckily, they'll check all 58 units (in case poll workers forgot to mark the newfangled box after a long day), but it will take a while to get through the ESlates. This didn't seem like a top priority, given how few votes they expect. [How can it be considered an independently cast vote if that vote CANNOT be counted in the official tally without the intervention of multiple other people to do the ballot duplication? - DB]
Based on the experience of other counties introducing accessible voting machines, Lindsey and the Hart rep expect less than 10 votes in Humboldt County will be cast on the $800,000 equipment. Only 58 units will actually be rolled out this time. Lindsey said he was glad it's unlikely they'll be used much, for two reasons.
His first reason is the hassle factor of transcribing the votes onto ballots usable by the Accuvote. (To give Lindsey credit where due, the reporter made a big deal about "Why can't you just add the total from the ESlates to the GEMS total?" and Lindsey said it was technically possible but he wanted to have all votes in the GEMS backed up by scanned ballots.) There's already a process described in the Elections Code for transcribing spoiled ballots (usually absentee ballots with coffee stains or holes) they'll use to make sure the votes cast on ESlates are recorded correctly. Lindsey admitted that transcription is tedious enough that the process routinely catches 10-15 errors per 100 ballots transcribed.
His other reason he's glad only a few people will use the ESlate is that he says that implementing it in such a rush is "like being on an old wooden roller coaster, and it shakes and rattles," as he told the reporter from the North County Journal. Doesn't that just build your confidence in the system--and the administrators?
**Lindsey Loves DREs**
As you may recall (or not), the ESlate was a last-minute replacement for the VotePAD preferred by Lindsey's boss, Carolyn Crnich, the Humboldt County Recorder. After this gripe session about how much hassle and expense the ESlates are, I commented that it was too bad the VotePAD hadn't been certified so we could use the same ballots for disabled voters. Well, that hit a sore spot. Lindsey started ranting about how inferior VotePAD is compared to the ESlate, and contradicted just about everything I'd read about the device or heard from Carolyn about it.
After I got home, I called Ellen Thiesen at VotePAD to confirm some of his statements that didn't jibe with my recollection of her device. Among other things, she said it is NOT true you have to back up the whole tape and start over if you want to review/back up on the ballot, as Lindsey stated. The rep from Hart claimed he had been in Yolo County when they used it in an election and it was a big mess. Ellen said it has never been used in an election in California, so that statement is obviously false. (And what would the Hart rep be doing there if his equipment wasn't selected? Even if he got confused and meant the testing in Sacramento, what's he doing there either?)
Lindsey said he tried to use a VotePAD with his eyes closed and it was confusing. (Therefore, he assumes it would also be confusing to blind people who are accustomed to working by touch.) He described his attempt to verify his selections by running the vibrating wand down the page, and it had buzzed every time he crossed a mark or scratch. Ellen said he was using it wrong--you are supposed to touch the wand to the holes over the ballot bubbles. He also said it was only useful for blind voters, and of course Ellen said he was absolutely wrong about this too. She said Carolyn could tell that despite the badly-designed usability testing it was a good solution. Unfortunately, I'd managed to elicit this rubbish from Lindsey in front of the press. I gave Ellen the phone number and URL for the North Coast Journal.
Lindsey admitted he'd had calls from unhappy disabled voters who said they did not want to use a DRE. They will be allowed to vote on paper, and poll workers are allowed to assist. Oh, and he told the reporter about the blind people who were so happy to vote on touch-screens in the 2000 elections. That's what convinced him we had to go electronic for accessibility. Nothing like a good sob story. (I wondered if those videos were made before or after Diebold started funding organizations for the blind/disabled.)
**Absentee Ballot Chatter**
One of the poll workers asked about missing absentee ballots. Lindsey confirmed that he's had a lot of calls, particularly from Blue Lake, from people who haven't received theirs yet. He is following up on this, but he's puzzled because those were some of the first sent out and his office has already received about 25% back already. (So what happened to the missing ballots? Does someone think they could get away with using them to stuff the ballot box? Does Blue Lake have one of those postmen who just gets tired of delivering mail and trashes it instead?)
The poll workers also heard that some of the absentee ballots had been misprinted and had to be replaced. Lindsey said it was actually just the sample ballot in the voter information packet, but that this had been garbled and spread around before his office could announce the correct information.
**Gossip Corner**
I tend to believe him on those items because he was open about what was going on. I've noticed he gets defensive when he's making s*** up--and he starts making snarky comments about his nemesis, Dave Berman of the Voter Confidence Committee. [Moi? Just because I'm the only one in town who has publicly said it is obvious that Lindsey McWilliams should resign in disgrace, all of a sudden I'm his nemesis? - DB]
For example, I'm not sure if Lindsey was responding to my query about poll tapes or the reporter's question about data transmission security, but he threw out this comment: "Dave Berman has made a conscientious effort not to educate himself about what we do here." When discussing the ESlate and electronic voting, he said Dave kept bugging him about hand-counting ballots, but that it was too slow and everyone knows it's inaccurate. A couple of other times he made passing references, but I didn't make notes. I thought this was rather unprofessional, to say the least. I'm sure he doesn't think he's doing anything wrong, because he made those comments to the reporter... you'd think he wants to be quoted or something. [If you are describing public servants who believe they're acting in the best interest of the community while actually sabotaging Democracy, you might be describing a Manchurian. - DB]
**Two Sides to the Story**
I wish I'd noted the reporter's name. He asked who I was and why I was there, and in front of Lindsey and everyone, I just said I was a concerned citizen and curious about how things worked. (Which is true, but not the whole story.) Then I realized I was missing an opportunity, and caught him as the meeting broke up. I told him I had heard about the testing from Dave Berman's mailing list, and I wasn't sure why Dave hadn't been able to make it, but yes, I was definitely not confident in the outcome of our elections. [Confident? How could you be? There is no basis for confidence. - DB]
I also said that Lindsey was wrong about hand-counting being slow and inaccurate. Canada and the EU hand-count ballots quickly and accurately--if voting machines are so great, why don't they use them? He hadn't heard these countries don't use voting machines. He insisted that it would be easy for hand-counters to collude to report more votes for a certain candidate. I tried to describe the safeguards against it, but his mind was made up.
I agreed that the L&A test *seemed* to work, and that it had caught the unexpected ballots since the test was changed without notice. My argument was that how do we know that election mode does the same thing? And how do we know the machines won't "hiccup" under field conditions? He wasn't familiar with the technical problems in other states or with the Bilbray race this June, since he's apparently too busy reporting to follow other news.
I didn't think of pointing out that Lindsey's boss Carolyn was a VotePAD fan and had taken the time to learn more about it than Lindsey had, even though it's his job to do more than poke at it randomly for a few minutes.
Yeah, I can see why Lindsey's irritated with Dave--he's not accustomed to being held accountable.
It will be interesting to see what comes out in the news.
**Picky Details**
I don't know if this information is helpful, but I noted that the version of Accuvote software printed on the poll tape was 1.94.6. I wonder if this is the currently certified version, and if there are any Known Issues with the software? [The main issue, above all others, is that the votes are counted with secret programming code. - DB]
***Conclusion***
Well, that visit this morning was interesting in ways I didn't expect. Although I wish the Accuvote had malfunctioned--that would have been very fun to watch. More precisely, it would have been fun to watch Lindsey cope with it.
Permalink:
http://wedonotconsent.blogspot.com/2006/10/humboldt-logic-and-accuracy-testing.html
Read or Post a Comment
Kathryn,
Thanks for this tremendous report, and astute follow-up comments. I like the Simpsons reference.
To clarify just one thing, it seems in your first comment above you are addressing the privacy aspect of the disabled voter's vote being transcribed. You may or may not accept Lindsey's explanation. Either way, that's not what I was addressing in my comment. Voting privately and independently are the two HAVA provisions for which eSlate was supposedly purchased. My point is that casting a ballot that requires multiple other people to be officially counted is not voting independently.
Curses. Foiled again.
Dave
Permalink to comment | Top of Page | WDNC Main Page