[Date Prev][Date Next] [Chronological] [Thread] [Top]

Re: Logging Scanner Jams (was: Audit trail improvement suggestions)



Dear Devil,
----- Original Message -----
From: Ian Piper
Sent: Friday, December 10, 1999 10:26 AM
Subject: Re: Logging Scanner Jams (was: Audit trail improvement suggestions)

Excuse me while I play the devil's advocate here. 
 
Is this "jam event record" something we want to have displayed or recorded at all?
Yes
 
If there is a problem with the count totals between the AccuVote display and the ballots in the Ballot Box, why do you want to so quickly use it as an excuse? 
Jams are the only excuse we've got for these inconsistencies.  Remember that that the totals are being compared not just to the number of ballots in a ballot box, but the number of signatures on the roster as well - plus the ballots are numbered on the stub.  They know how many voters were given ballots and voted without too much doubt.
 
What would you do if there was no jam event recorded and yet there was still a mismatch between ballots in the box and the counter on the AccuVote?
Depends on whether its high or low.  Realize that most sites don't do anything if the count is less than 3 or so off.  They mostly look at the ones that are off by 3+ or so.  Again, its not just the Counter on the AccuVote, its the # of ballots, the AccuVote report, and the signatures on the roster.
 
What would you get from a jam event recordThat a jam happened? 
You get an explanation that human error was the factor and explanation for the discrepancy rather than "the AccuVote isn't counting" right.
 
But who was at fault: the poll worker or the machine? 
The Pollworker.  That's the whole point.  It gives you a chance to show that X number of times there was a potential that the pollworker didn't handle it correctly.
 
How do you know if the mismatch was caused by that jam.
Again, that's the whole point, you don't know, you never know what really happened, but if we can show that there were instances where the pollworker could have mishandled the ballot, then they do less "fingerpointing" at the AccuVote itself.
 
Would they call into question the results of an election if a unit showed multiple jams yet the ballots totals matched the counters?
No, they'd just assume the pollworker handled the jams correctly, or they would recount the precinct.
 
I understand it would be great to have the detail for troubleshooting, but do you think they would find the fault due to the poll worker when they have a machine that could be perceived as "at fault" instead? 
Of course.  The "culture" out there believes that pollworkers can barely get their shoes on in the morning.  It's assumed that the pollworker is at fault.  No administrator wants to believe its the AccuVotes fault.
 
I think less info is better under these circumstances.
Dear Devil, you may want to attend more elections and be the one trying to explain the reasons for strange AccuVote behavior.  What the folks in the field - Sorry, what I'm saying - is that "jams are our friends" when it comes to explaining anomolies. 
 
What kind of interpretation would someone make when seeing that detail? 
Example
"Oh, look here!  The paper tape shows it jammed during the day.  The machine must be responsible for the error in the totals."
Quite the opposite in the real world.  Its "Oh, look this explains that the pollworker didn't read the display!  Those pollworkers never listen to us in training!"
 
 
If we provide this "jam event record" as an excuse, wouldn't it throw doubt onto the integrity of the machine and its programming. 
See above.
 
I believe it is better to leave answers for these unsubstantiated anomalies to the discretion of the election supervisor.
That's why you're the Devil's very own advocate in McKinney!  But we like you anyway.
 
Outside of the "O" level ROMS, when a ballot is processed into a Ballot Box by the AccuVote, it is counted.  If a reader takes a ballot, it accurately scans that ballot and deposits it into the Ballot Box. 
Now you're damaging your own credibility by making these sweeping statements.
 
If there is any problem with the reader (i.e., hung or failed Self Test), it will not process the ballot into the Ballot Box.
So even the Devil's team has strong degrees of faith.  I come more from the Kierkegaard side of the family, faith is a daily affair, and depends on the circumstance and the ROM level.  I can point to any number of "can't happen" events that turned out to be true (ballots being put in the emergency bin, and ending up in the write-in bin, passed ballots, etc.)
 
Play out an election night scenario in your head with these kind of questions arising.  Now imagine there is nobody there from Global to explain the information provided.
It's my dream come true.
 
Let's not consider every user's request as an absolute.  Perhaps the user is just as much in the dark about the ramifications of that request as we are.
The problem here is that I agree with the user, but not for the same reasons.  Jams are our friends guys.  They give us all kinds of room to explain things.  The problem comes more from the scenario where a Global person says, "Well, there were probably jams" and the pollworker can't remember whether there were or not.  Repeat after me, "J-A-M-S  A-R-E  G-O-O-D".
Ian
 
----- Original Message -----
Sent: Thursday, December 09, 1999 8:26 PM
Subject: Re: Logging Scanner Jams (was: Audit trail improvement suggestions)

I think the statistical approach would be adequate, although the log with a
date and time stamp would be the preference.

I think that saving these for the 2.0 release is OK in the sense that we
might be able to "sell" the upgrade, vs having it be required.  However, it
might mean the AV would have to be recertified in many states.  Not good.
Other thoughts?
----- Original Message -----
From: Guy Lancaster <glanca@dieboldes.com>
To: Kerry Martin <kerry@dieboldes.com>
Cc: <rcr@dieboldes.com>
Sent: Thursday, December 09, 1999 8:40 AM
Subject: Re: Logging Scanner Jams (was: Audit trail improvement suggestions)


>   Some form of logging scanner jams has been brought up
> before and I think that I spoke with Steve about a month ago
> about it.  This may also be related to Jane's RCR: Print
> Vote Centers Activity of Oct 27, 1999.
>
>   The problem, as Monica stated, is that the officials need
> statistical support when trying to resolve discrepencies
> between the number of ballots in the ballot box and the
> number recorded by the Accu-Vote.  I can see three possible
> ways of providing this, each with pros and cons.
>
> 1) Add a stats counter for the number of jams experienced
> since the counters were last cleared.  While we're at it, we
> could have separate counters for counted and returned ballot
> jams.  This would be my preference since it provides the
> required statistics while consuming the minimum of
> resources. However it requires a memory card format change
> and so would only be available in major new releases such as
> the upcoming version 2.0.  These statistics would appear on
> the audit reports and could be uploaded for reporting by
> GEMS.
>
> 2) Print a log entry on the internal printer for each jam.
> This could be selectively enabled and disabled from GEMS'
> Accu-Vote parameters and would not affect the memory cards.
> Downsides include having to prevent the tape from jamming
> while it's locked in the printer compartment and having to
> count the entries on the tape to gather the statistics.  The
> upside is that it would provide a timestamp on each entry
> which could help with resolving the problems.
>
> 3) Record an entry in the audit log for each jam.  This one
> scares me a little because we have limited space on the
> memory card for the audit log and currently we allow for
> about 400 entries.  The purpose of the log is to record
> critical operations such as inserting the memory card,
> powering on the unit, the starting and ending of counting,
> printing reports, and a few others.  Although possible, it
> is very unlikely that we would overflow the audit log with
> these entries.  Events like jams are far less predictable
> and a problem with a machine could cause a large number of
> jams.  Should the log overflow, the oldest entries are lost.
>
>   Some would suggest that we increase the size of the audit
> log or dynamically allow it to grow to use available memory
> card space.  The latter is technically difficult especially
> if we want to record events starting from the beginning of
> the download operation as we do currently.  Allocating more
> space for log entries comes at the expense of less space for
> ballot information and counters which will affect those
> counting absentee ballots.  It seems less of an issue for
> 128K memory cards than for 32K but once we decide to log
> jams, we have to allow for it in all systems.
>
>   My suggestion is to add support for optionally printing a
> jam log message in the next 1.94 release.  For version 2, we
> would continue to support this plus add statistics counters
> for jams.  Unless people feel both that it is critical that
> we record the time of each jam and that managing the printer
> tape is not acceptable, I would avoid adding this to the
> audit log.
>
>            Guy
>
>