How they test punch card machines--THIS IS UNBELIEVABLE!!!

[dewaldd]

This is absolutely unbelievable….the computer can easily tell if it is counting test cards instead of regular ballots. I called up 4 Boards of Election in Ohio, and they all gave me the same story—it is based on recommendations from the manufacturer.


How they test that a machine is counting the punch cards correctly:

They have a "test deck" whose cards are punched out in a pattern that is completely different from actual ballots. One card is punched out on all the numbers ending in 1, two cards are punched out on all the numbers ending in 2, three cards are punched out on all the numbers ending in 3, and so on.

As anyone with a modicum of computer programming experience can tell you, it is very easy to include a small program to that says

IF test deck: count correctly
IF real ballot: do what you have to do.

The mind-boggling significance is this: because the test ballots are obviously different from regular ballots, we have NO ASSURANCE that the testing procedure is actually testing the program as it counts actual ballots. The “test” is completely meaningless. For a meaningful test, the test ballots would need to be indistinguishable from real ballots.

Since we are not allowed to see the code, we just don’t know if there is a difference in the program based on whether the ballots are test or real. Hopefully the certification process would notice something like this, but Bev Harris found many instances in which voting machines were using uncertified programs.

I don’t trust any results anymore. Everything needs to be hand-counted, as far as I am concerned. We need to push for complete hand counts on the basis of flawed testing procedures.

[okieinpain]

1. you are correct. very bad testing procedure.

[roseBudd]

2. Urban couties used ES&S to tabulate, suburban counties used

Triad Systems

[coreystone]

3. We have a machine in the USPS. It is called MERLIN!.....

This is really boring, but, please stay with me! :-)

We utilize this machine to verify "bulk mailings" where I am employed. For each 24 hour cycle, we MUST run a "test deck" comprised of a total of 11 cards. The first five are a rubber/plastic composite with various sizes and thicknesses; the second group of three, are paper replicates (different dimensioned rectangular pieces) of what would be typical larger pieces of mail; which we call "flats". The third group would, also be paper replicates of "letter" size mail (all equal in height and length).

Before we are allowed to perform a "verification" on the sample of mail which is presented, a "GO/NO GO" test must be performed to insure that the MERLIN is accurately reading, and weighing the 11 cards of the test deck. The software will analyze these 11 pieces to insure that the scales, camera, etc. are within a certain tolerance of error.

In the scenario of "TESTING" the MERLIN, the parameters of "tolerance" are predetermined by the software; which compares the results of the test (the scales, the camera, etc) in relation to the standard that these "physical" devices are measuring the attributes of the "TEST DECK". There is a somewhat constant variable which the software analyzes. If any of the four physical attributes of the MERLIN fail in verifying that constant, then we must have our maintenance staff make the proper adjustment to correct the physical aspects of the MERLIN which are necessary for a proper reading of our customer's mail. IT WILL NOT ALLOW US TO perform a verification until it PASSES a "GO/NO GO" TEST.

When running an "actual" verification, there are still problems which we encounter. Some due to the "software", and some due to several elements of the operation of the MERLIN itself.

For example, if we are verifying a sample of letters, sometimes the MERLIN erroneously flags an error in the "mail characteristics" of the "aspect ratio" of the mail. A letter must have an "aspect ratio" between 1.3 and 2.5. That means that length of the letter in relation to the height of the letter must fall within the "ratio". Example: A letter is 6 inches long, and 4 inches high; the "aspect ratio" is 1.5 Certainly falling between the requirement of 1.3 to 2.5.

Sometimes, the "belt" which feeds this mail into the MERLIN does run evenly. We then have to manually measure the mail to insure that it is being "INCORRECTLY" read by the MERLIN.

NOW! First of all, if the "physical feed" mechanism of the voting device is not properly calibrated, the ballot may not be being read accurately. This may or may not be a bias toward the selection of any candidate over another. It could also record A "NO VOTE"! Secondly, there may be times when another testing of the voting device would be in order, to insure that the "feeding" of the ballots was being properly synchronized with the optical scanning of the ballot. This could certainly be a more possible "reading" error of the ballot, dependent upon the voting traffic on any particular machine.

I wholeheartedly agree with you "dewaldd", that the "test deck" should definitely be, totally "replicated", in every way the "actual ballot". There are too many possibilities for error, intentional, or not.

Thanx for allowing me to relate my experience with this sort of physical "recording" as I know it to be. I hope that some of the more diligent members may be able to help connect the dots better than I!

:-)

[Hobbes199]

5. Maybe that's the whole point.

An incumbent will usually be listed first on the ballots. Maybe the machines tend to favor the incumbent by either giving the first entries more votes, or by giving subsequent entries fewer votes. No hacking necessary, just use this machine and make sure your name is listed first.

[coreystone]

6. Don't know, can't say....and doubt if the manufacturers would be ..

honest concerning the reliability of their diagnostic tools. I know that HQ of the USPS isn't, and they didn't even make the stupid machine. Hardware and software was completly "subcontracted" to the private sector. I talked to the individual in Washington responsible for initiating changes in some of the "software" glitches on these machines. She became very indignant and self-righteous about her role in correcting obvious problems that we were experiencing. Government bureaucracy!!!!!!!

:crazy:

[Ms. Toad]

14. ballot positions are rotated

by law, so the Bush was no more likely to be first on the presidential ballot than the Green Party candidate (in order to counteract people's tendency to vote for the first position if they don't have their minds firmly made up.

If there are 4 candidates, approximately 25% of the precincts will have Candidate A first, 25% will have B, 25% will have C, and 25% will have D. There were so many gubernatorial candidates in the recall election that they ran out of polling place groups - a few of the candidates never got to first place, but that is very rare. Normally it evens out pretty well.

[Fortunato]

7. thanks for the testing description.

No, it's not boring at all.

It's interesting to hear what testing procedures are in place for other systems. Obviously your tests reflect the gamut of materials the machine is expected to sort through on the field.

It's pretty clear that the ballot testing does not simulate real ballots.

I'm a tech writer -- so you could consider me an amateur "programmer" (i.e., I can read the stuff and understand it without actually knowing the language in question), and am acquainted with normal testing processes as part of my work, since we write software for clients.

I find it utterly pathetic that corporations have much more rigorous testing in place for non-critical apps, but we have entrusted our voting processes -- which essentially determines how billions and trillions of dollars of taxpayer money get spent (aside from the "leadership" issues) -- to what amounts to untested software that an eight-grader would know how to test effectively.

It's insane.

[coreystone]

8. 25 YEARS ago I could write Basic, Fortran, + COBOL off the top of head...

but, I never continued into "C++"! The linguistic format and syntax of a particular "programming language" and the "logic" of a rather rudimentary "flowchart" CAN be two different animals, BUT, the LOGIC still prevails!

I am a Psychology major and have academic background in that field, as well as an academic background in Educational research methodologies. I have found many of the principles to be rather similar. It doesn't take a rocket scientist to realize that unless the "test conditions" accurately replicate the real life usage of the voting machines which are being utilized, then the "reliability" is only restricted to the "deck of cards" having been run through the machine at that moment of time in which they were tested!!!

:-)

[Hobbes199]

4. I keep remembering back to taking standardized tests...

And there was always an oval that you _could_ mark which would tell the machine that it was the answer sheet to be used to check all the subsequent cards.

I wonder how many little secrets abound in these machines. Maybe a sheet can be run through to give weights to the different options, or remove an option completely — after all the machine itself doesn't know it's being used for election purposes.

[Eloriel]

9. Yeah. I've had a very uneasy feeling about these machines

as well -- kinda along the lines we've been ignoring them at our peril. Thanks.

[GetTheRightVote]

10. Terminator - Machines will rule the world one day

Why is this inclusion of computer in our world unexpected but we
sould never trust those who run them or create the logic within them.

[GayGuyinCalifornia]

11. Who owns the machines that counts the punch cards?

[Starlight]

12. I'd like to know this too. n/t

[ignatzmouse]

13. Ohio Punch Card Maker

Triad Governmental Systems, Inc.

[Bozos for Bush]

15. Kick for turkey!