Friday, November 9, 2012

The Trouble with Tribbles, er, Electronic Voting (DRE)....

Freedom to Tinker is a blog started by Ed Felten at Princeton University that examine and critique, in many ways, technical mythologies.  Of relevance, Prof. Felten and his blog has aggressively critiqued electronic voting machines.  Freedom to Tinker has become a very well respected academic blog and has multiple contributors.

The latest post is by Prof. Jeremy Epstein.  He spent election day fielding calls concerning election troubles, and what do you know, the vast majority of them came from Virginia, and a vast majority of those came from electronic voting machines.  To be clear, these were not troubles suggesting voting fraud or voting suppression.  Rather these are troubles suggesting why it is time to end the experiment with electronic voting (DRE or Direct Recording Electronic). 

Here is what Prof. Epstein observed:
 Hard to solve problems included:
  • Long lines. In most cases these seemed to be caused by people waiting for machines, not problems with checking voters in, but it’s sometimes hard to tell. These were most pronounced in locations with DREs [Direct Recording Electronic] – if you’ve got paper ballots, you just hand out more pencils. In some cases, pollworkers were turning voters away, telling them to come back later – for whatever reason, they weren’t using the emergency paper ballots that were intended for this case.
  • Machines that won’t boot or crash. In most cases these were DREs, in which case they led to long lines. If it was an optical scan, see below under “easy to solve problems”.
  • Problems with the magnified version of ballots. In Norfolk and Virginia Beach, there were repeated reports that the magnified version only showed Romney on the first screen, and to get other presidential candidates the voter had to advance to a separate screen. I was quite surprised by this – I guess Logic and Accuracy (L&A) testing performed before the election doesn’t include the magnified versions in those localities?
  • Ballot setup problems – in particular, in Covington VA, the DRE offered a choice of Obama as the Republican nominee or Romney as the Democratic nominee. After coverage on the local TV station, the locality shut off the DREs and went to paper ballots, which had the candidates associated with the proper parties. The part that scared me was a quote from the local election official that “All votes that were cast Tuesday morning will be counted properly.” I don’t know what “properly” means in a case like this. (See http://www.wdbj7.com/news/wdbj7-story-grayson-voting-11612,0,5763810.story) Again, shouldn’t this be found by L&A testing?
  • Ballot setup problems – reports from across the state that people weren’t getting the presidential race, and it was going straight to the Senate race. These are unfortunately hard to reproduce. When I asked voters about the confirmation screen, they generally said there hadn’t been one.
  • Ballot display problems – several reports that the ballot was displaying party name but not candidate name. It’s possible, of course, but hard to reproduce. Only on DREs, of course.
  • Pollworker errors in helping voters. We got a few reports that voters complained of problems (e.g., “no presidential race”) and called over a pollworker, who pushed the “cast vote” (or equivalent), and then refused to let them vote again. (Which was, unfortunately, probably the right answer.)
Easy to solve problems included:
  • Voters confused why ballots were being put into the side pocket of the scanner and concerned that the ballots wouldn’t be counted (obviously because the scanner wasn’t working, and they would be counted by hand or machine at the end of the day).
  • Not enough paper ballots for the number of voters. This was especially true in places that had DREs, but the DREs didn’t work (or the lines were long), so they were using the “emergency” ballots. In Virginia, the chief pollworker in a precinct can make photocopies of the ballot with minimal approvals. Unfortunately, this wasn’t well understood by pollworkers who didn’t take advantage of it.
  • Using markers instead of pens or pencils. This can be a problem if the marker bleeds through on a double-sided ballot, as the scanner may pick up the reverse-side mark as intentional.
  • An optical scan machine bouncing ballots as overvoted (too many votes for one or more candidates) when voters insist they hadn’t. Hard to say if this was a calibration issue, a smudge on a reader, or an actual voter mistake. In any case, the solution is simple – just hand count the ballots.
Epstein says:
"Moving to optical scan will solve the vast majority, but given tight budgets, it’s going to be a while in Virginia."
I saw some optical scanning machines at my place of voting and was thrilled.   There are more problems with electronic voting than just what was experienced tuesday - but what was experienced tuesday demonstrates the vulnerabilities of this type of voting.  The litany of articles at Freedom to Tinker critiquing the problems with electronic voting can by found under the tag Voting Machines.

No comments:

Post a Comment