DRE Junk: Human Error Rate vs. System Error Rate.

[Bill Bored]

I didn't want to put this in FogerRox's HAVA Sect. 301 Advisory thread
<http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=203x385691>
because it's a bit of a diversion, but it's something to think about:

Let's say the DRE is programmed to drop your Presidential selection if you make that before you vote Straight Party in states where this is allowed. (This happens to be the default behavior of Diebold Accuvote TS junk DREs, unless the PA Straight Party Voting Option is turned on.) So the voter selects a President, then selects either Democratic or Republican straight party for the rest of the ticket. The Presidential vote goes away. Is that a human error, or a machine error?

And if the Summary screen shows the undervote or switched vote for President, is it a human error even though the MACHINE dropped or switched the vote?

I'm thinking in terms of HAVA/EAC definitions here. It could be that all these Diebold machines can be decertified on this basis in a bunch of states that allow Straight Party voting. There were 17 of them in 2004.

[Amaryllis]

1. We need Land Shark for this question. Brilliant, Bill.

[Bill Bored]

4. He's a good man. Good heart. Strong leader.

Ask him to check his PMs too. I sent him one about something else.

Thanks!

[TruthIsAll]

2. Definitely not a human error; it's a programming trick to squash votes.

First rule of system design: Protect the user from making input mistakes. You do this by error-checking and prompting the user to re-enter his choice. Very elementary.

This may have been a deliberate scam to invalidate the presidential vote. At minimum, the user should be prompted to re-enter the presidential vote after choosing a straight ticket.

Check to see if this code was used in mostly democratic precincts where there was a strong tradition of straight party-line voting.

[Bill Bored]

3. No data on that but the code should have been used in 17 states.

AL, IA, IN, KY, MI, MO, NC, NH, NM, OK, PA, RI, SC, TX, UT, WI and WV.
PA should have had a different setting though: Republicans who chose Kerry and then their own straight party would have had their votes counted in PA by law.

So here's an idea:

If Repubs in PA found it easier to vote for Kerry than in other straight party voting states, there should be more Repubs for Kerry in the final exit poll for PA than in the other states because E-M would not have had to adjust that weighting. In the other states, if Diebold, you would expect to see a large discrepancy between Repubs for Kerry in the pre-adjusted poll compared to the final.

Do you happen to have horizontal Party ID weightings for the 50 states? would be interesting to see if PA or the other straight party voting states stand out.

[TruthIsAll]

5. I don't have that data. n/t

/

[Bill Bored]

6. How come? Was this question only asked in the national? nt

[FogerRox]

7. GREAT thread BILL--

There is a human factors report out there google it--
& The Shamos report that are pertinent to this topic.

Yes errors are created by the voting machines and then the Voter. BUT the user interface of voting machines are different and should be considered a 3rd category.

I'll check HAVA ---but I think there is some language in HAVA that suggests that the voting process must be somewhat easy---

It maybe that HAVA requires that a typical voter should be able to successfully navigate their way to casting a ballot. If the Interface sucks SO BAD that this doesnt happen--- one can make the case that that voting machine is not HAVA complaint--irregardless of its certification -- even 2002 NASED Certification.

Along the lines of one of Dr Shamos points in his report.

[FogerRox]

8. HAVA section 301----

SEC. 301. <<NOTE: 42 USC 15481.>> VOTING SYSTEMS STANDARDS.

Requirements.--Each voting system used in an election for Federal office shall meet the following requirements:
In general.--
Except as provided in subparagraph (B), the voting system (including any lever voting system, optical scanning voting system, or direct recording electronic system) shall--
i)permit the voter to verify (in a private and independent manner) the votes selected by the voter on the ballot before the ballot is cast and counted;
ii)provide the voter with the opportunity (in a private and independent manner) to change the ballot or correct any error before the ballot is cast and counted (including the opportunity to correct the error through the issuance of a replacement ballot if the voter was otherwise unable to change the ballot or correct any error); and
iii)if the voter selects votes for more than one candidate for a single office--
i)notify the voter that the voter has selected more than one candidate for a single office on the ballot;
ii)notify the voter before the ballot is cast and counted of the effect of casting multiple votes for the office; and
iii)provide the voter with the opportunity to correct the ballot before the ballot is cast and counted.

Read carefully--- this deals with what should be considered the
USER INTERFACE

And if the interface sucks, the above parts of Section 301 dont happen.
Heres an example: "provide the voter with the opportunity to change the ballot"

If the interface sucks--- does that limit the voters "opportunity to change the ballot?

I would think so.

[Bill Bored]

13. What about undervotes?

I thought it was in there but perhaps it's just in the proposed amendments by Holt et al or a state law I've read? I'll check later. It seems that there is no explicit warning required to tell the voter that he/she didn't vote in a particular race or ballot question.

[Nederland]

16. Not a programming error

Its bad requirements. I'm sure the programmers made the system behave exactly the way they were told to.

[FogerRox]

9. Kick ---------------- for the team & Bill-

[Land Shark]

10. the voter's task is to vote

and the voter above has voted correctly, so it is a clear machine error.

Especially in a "voter intent" state, the job of elections is to divine that intent in any case where intent is ambiguous and to respect that intent whenever clear. States that are not "voter intent" states vary based on the specific laws and rules there.

Note that statutes are now beginning (and in some states have long provided) that a "responsibility" of the voter is to "correctly follow the procedures for voting" or something similar. This provides the basis for dramatically expanding the category of "human error" vis a vis machine error, and provides a claimed justification for denying someone their vote (the voter "can't follow the rules").

However, the summary screen, as your question implicitly points out, provides at least a fig leaf basis for blaming the voter for missing the machine's error, and voting officials have seized on it to shift blame to the voters. (Blame the voter is the theme, usually). The claim will be made that in all known instances, the voter successfully switched the vote back, because those who didn't SEE it of course didn't either report it OR correct it! Thus, the summary screen provides, under cover of a "service", the ability to shift blame to the voter in many instances of actual machine error or fraud. This would be one reason why a person bent on fraud would allow or tolerate visual evidence of that fraud: the votes were there on the screen and confirmed by the voter, and the voter is considered legally incompetent to impeach the vote once cast, and factually incompetent for "failing" to review the summary screen.

Others have correctly pointed out that under proper rules of programming as well as human factors approaches, error rates in the use of software are SOFTWARE problems, not human or user problems. At a sufficiently very low level these may be tolerated, but voting is a fundamentally simple process and there should be no reason for the rates of error actually observed (even in those cases where errors are admitted by officials and vendors, not to mention the others).

[FogerRox]

11. Exactly so Land Shark

"the voter above has voted correctly, so it is a clear machine error. "

Dr Shamos made a similar case in his report on the decertification of the Unilect DRE in PA.
If the user interface of machine #1 is easily understood-- & machine #2 isnt as well under stood, and machine #2 has a higher undervote rate--- than obviously it is the fault of the voting machine.

When undervote %s breakdown by machine type or breakdown by a particular vendors machine-- its clear-- its the machine-- not the voters fault.

[Bill Bored]

12. VVPATs are akin to summary screens.

One argument against VVPATs is that they duplicate the function of the Summary Screen and discrepancies between the Summary Screen and the vote totals have not been found (although I'm not sure how hard anyone has actually looked!).

The VVPAT might be easier to read than a Summary Screen, esp. those screens that require scrolling or paging. This is where a full-face ballot law can be helpful, although some advocates for disabled voters don't like them.

I think you have hit on a key issue though. If there is law, HAVA or otherwise, that requires the machines to be designed in such a way as to minimize voter error (not to mention ballot definition programming error!), it would be helpful. Meanwhile, it's up to the Land Sharks and the Courts, but anything that uses HAVA against itself could be a good strategy.

[eridani]

14. I wonder if any of these "blame the voter" assholes--

--would tolerate a word processing program without an "undo" option. Maybe the program designers could just explain to them that if they are too damned stupid to avoid mistakes, then they just don't deserve the privilege of using a word processor.

[FogerRox]

15. I had to use the UNDO on my laptop last night---