quote:COLUMBUS, Ohio (AP) -- An error with an electronic voting system gave President Bush 3,893 extra votes in suburban Columbus, elections officials said.
Franklin County's unofficial results had Bush receiving 4,258 votes to Democrat John Kerry's 260 votes in a precinct in Gahanna. Records show only 638 voters cast ballots in that precinct.
Bush actually received 365 votes in the precinct, Matthew Damschroder, director of the Franklin County Board of Elections, told The Columbus Dispatch.
State and county election officials did not immediately respond to requests by The Associated Press for more details about the voting system and its vendor, and whether the error, if repeated elsewhere in Ohio, could have affected the outcome.
Bush won the state by more than 136,000 votes, according to unofficial results, and Kerry conceded the election on Wednesday after acknowledging that 155,000 provisional ballots yet to be counted in Ohio would not change the result. (Full Ohio results)
The Secretary of State's Office said Friday it could not revise Bush's total until the county reported the error.
This problem was apparently found because of numbers that far exceeded the number of voters in the precinct. But if this is discovered to have occurred elsewhere after official vote counts in more populous precincts, combined with provisional ballots that probably favor Kerry... well who knows? Posts: 2432 | Registered: Feb 2001
| IP: Logged |
posted
That such a flaw is even possible on a voting machine speaks volumes of the incompetence or corruption of the company providing them.
Posts: 15770 | Registered: Dec 2001
| IP: Logged |
posted
How come ATM machines always work perfectly (by denying me money I do not have) and yet we can't get voting machines to work correctly?
Posts: 1592 | Registered: May 2000
| IP: Logged |
quote: How come ATM machines always work perfectly (by denying me money I do not have) and yet we can't get voting machines to work correctly?
actually they don't. I've known people that have had an ATM machine screw up and eat their card. I also saw an ATM eat a girl's deposit, and not count it as being deposited. I have also read several news stories about ATMs giving people extra cash...though the banks always check the records afterwards and track people down.
Posts: 1901 | Registered: May 2004
| IP: Logged |
posted
Let this be a lesson to you : the most hi-tech way is not always the best way.
[smug canadian attitude] In Canada, we have pieces of paper with the candidates' names, we mark an "X" in the circle opposite the name of the person we want to vote for, and drop it in the box. Then we count them by hand. No screw-ups, no recouts, no challenges. But, yanno... kinda primitive, eh? At least it works. [/smug canadian attitude]
posted
Same here in Australia - just cardboard boxes, pieces of paper and poll workers who at least give the appearance of non-partisanship.....
Posts: 2245 | Registered: Nov 1998
| IP: Logged |
posted
They have alot of people working on the counting. There are lots and lots of polling stations in every riding, each with a few hundred voters or so. At the end, the DRO (Deputy Returning Officer) and the Poll Clerk (that was me! ) count the ballots together, and double-check themselves and each other, and then report their numbers to the Supervisor, who then reports to whoever he reports to... It's really not that big a deal, you know. I mean, anyone can count.
Posts: 609 | Registered: Oct 2004
| IP: Logged |
posted
Yeah, but Canada uses metric ballots. They American system doesn't make any intuitive sense so I can see how people could get confused.
Posts: 3243 | Registered: Apr 2002
| IP: Logged |
posted
As one who works with numbers as my life's work, I have to say that I'm so totally offended by this that I believe the entire election should be scrapped and we should have to start over and keep doing it until we get it right!
I hereby claim the right to ignore the outcome of the 2004 election entirely. It is obviously flawed.
Anytime the count of MISSING values is HIGHER than the actual count, there is only one way for that to happen -- somebody is purposefully screwing things up.
Trust me on this. It will eventually come out that either: 1) A pollworker deliberately distorted the count, or, 2) The machines were deliberately programmed to miscount the tallies.
Maybe we should've taken Castro up on his offer to monitor our election.
Posts: 22497 | Registered: Sep 2000
| IP: Logged |
posted
As one who worked with computers as my previous life's work, there's a much less sinister possible explanation. In many applications, you don't want editing or deleting capability at all. If the logs are cumulative over the lifetime of the machine, and some installer ran test routines using the same on-screen button that ended up representing Bush, the repository holding those test results would have a much higher number than it should. And if someone read the result from the lifetime cumulation rather than the results for this election, it's possible the error occurred in this manner.
Certainly it indicates incompetence. But until we know more, we can't be sure it's unrecoverable incompetence.
There's anecdotal evidence that people have been confused by this already, in Philadelphia, when Republican poll-watchers noticed the presence of hundreds of votes on the machines before the polls opened.
Also, note that the error was caught, likely because there's a hand-record of the number of voters in each precinct. This provides error detection potential in every single precinct.
posted
Just because there's a "potential" explanation involving simple incompetence, I am sure that is what this "error" will be ascribed to.
Posts: 22497 | Registered: Sep 2000
| IP: Logged |
posted
Mind you, I'm not buying into the conspiracy theory.
However, it was only "easily detectable" because that precinct had so very few voters. In a large precinct, this would have been [virtually] undetectable.
[ November 06, 2004, 11:56 AM: Message edited by: Sara Sasse ]
Posts: 2919 | Registered: Aug 2004
| IP: Logged |
quote:However, it was only "easily detectable" because that precinct had so very few voters. In a large precinct, this would have been [virtually] undetectable.
Not if the error only results in more votes than the number of people who voted - which my explanation would result in. In that case, every precinct where this error happened would show a mismatch between number of votes tallied and number of voters who voted.
I don't know that this is the explanation - I was responding to Bob's definitive statement with an equally (I'd say more) plausible theory.
We need to hear an explanation and have it checked, and whatever the error was we need to ensure it didn't happen elsewhere. But I'm not ready to conclude malfeasance yet.
Dagonee Edit: Bok. Nicely put - the idiocy thing, not the minion thing . Have you worked in government, too?
posted
Are all the precincts rigorous in checking the number of votes tallied agains the number of people who voted? Is this standard procedure for large precincts?
It may well be. (I really don't know.) But where I voted in 2000, my name was checked off a list by hand. So long as I was registered, I was permitted to proceed to the poll station. I would doubt that the Diebold machines check resistration separately.
So unless you were checking against the initial (hand-marked, in my case) list, you wouldn't be sure what the actual number of voters was. It seems pretty easy to make the machines adjust the number of voters if they were adjusting the actual votes.
It seems likely to me that this was picked up by the poll workers at that site because they knew there weren't that many people that went through the station -- 638 is far off from ~4500. I wonder if they used the checkoff list of registrants in order to get the actual number of voters, once they realized something was off.
4000 extra votes wouldn't have triggered intuitive mental red flags in the poll workers of a very large precinct.
[ November 06, 2004, 12:33 PM: Message edited by: Sara Sasse ]
Posts: 2919 | Registered: Aug 2004
| IP: Logged |
posted
Mind you, I too subscribe to the idiocy theory of anti-conspiracism. But this AP articel doesn't give enough detail for us to shrug it off so easily, in my opinion.
Posts: 2919 | Registered: Aug 2004
| IP: Logged |
quote:In an invitation to a Republican fund-raiser at his suburban Columbus mansion, O'Dell said he was "committed to helping Ohio deliver its electoral votes for the president next year."
The letter closely followed a visit by O'Dell to a fund-raising powwow at Bush's Crawford, Texas, ranch for six-figure fund-raisers known as "Pioneers and Rangers."
He said he regrets the wording in the letter.
Of course, there is a possible innocent interpretation to the words. It is suspect, though.
[Edit: Yeah, I know. But you see how it could happen, and given the context of the executive making such an asinine assertion to Bush's own fundraiser elite ... well, the usual idiocy isn't enough to assert. I think we'll all look for a recount.]
[ November 06, 2004, 12:22 PM: Message edited by: Sara Sasse ]
Posts: 2919 | Registered: Aug 2004
| IP: Logged |
posted
The thing is, the Diebold machines aren't actually bad because the CEO of the company is a die-hard Republican; no, the real problem is that they use a shoddy design (at least one version, that I don't know if it went out for widespread use, used a wide-open, security-wise, excel spreadsheet/CSV file to record the votes on a system that wasn't completely secure, in that if you could get into the intranet it was one, you could fairly easily change the file, WITHOUT AN AUDIT TRAIL).
Dag, actually, I was paraphrasing a more well known, and older, saying: "Never attribute to malice what you can as easily attribute to incompetence." Or something like that.
As for government work, nope (unless you consider summer jobs working for the DPW as a cemetery mower ), although I did work for IBM for a year, which is not unlike other large bureaucratic entities
posted
I find it interesting that all of the 3893 extra votes went to Bush. Why not more random? Bizarre.
No, Bok, I agree that the basic problem is with the machines themselve. What puts a spotlight on it is the expressed intentions of the die-hard Republican executive -- not that he is Republican, per se.
You can see how the his comments make a halogen searchlight out of the interest, though.
[ November 06, 2004, 12:39 PM: Message edited by: Sara Sasse ]
Posts: 2919 | Registered: Aug 2004
| IP: Logged |
posted
When we wrote our cancer clinical trial application, we always "tested*" new features using the same set of criteria, so the incidence of adenocarcinoma of the lung was several thousand higher than it should have been. We had a special user ID we used that allowed us to exclude them, but if we didn't exclude that, the count was off.
Dagonee *By testing I'm not referring to true software testing - that had a nicely developed set of test data that spread test data out appropriately. Rather, I'm referring to running throught a clinical screening process because we wanted to discuss a particular screen, or see where an error was, or just get to a particular screen to make screenshots.
Posts: 26071 | Registered: Oct 2003
| IP: Logged |
posted
Let me get this straight. Voting machines allow the tallying staff access to more than one count? I go in with my password identifying myself as the supervisor of elections or designee and I have access to more than just the counts for this current election? I have access to what else?
A count of all votes ever cast in the system?
And to get to that count is just a button click selection just as if I were trying to get the counts of the most recent election?
Who wrote this software?
Does this make any sense as a way to run an election?
Also, if the machine is being "tested" why is it being tested with the name of an actual candidate? Wouldn't it make more sense to have the machine tested with dummy names (oooh, I have to let that one go...)?
Dag, I understand your supposition, but the methods to get to it, and the violations of standard programming practices (let alone the extraordinary steps one would want taken beyond simple standard safeguards if one is programming a voting application) seem to me to go beyond simple incompetence.
I'm not saying it COULDN'T be incompetence, but it is incompetence on a grander scale than I'm willing to admit.
It means: 1) the people who programmed the application were incompetent. 2) the people who used the machines to get the final tallies were incompetent and 3) we merely discovered THIS incompetence, and should be very worried about all the ways this could be messed with that involved less detectalbe forms of incompetence.
I am not even more in favor of throwing out at least Ohio's votes and starting over.
How many precincts are there in Ohio? How many votes "off" would each have to be before the state was incompetently given to Bush as a win?
Posts: 22497 | Registered: Sep 2000
| IP: Logged |
quote:Let me get this straight. Voting machines allow the tallying staff access to more than one count? I go in with my password identifying myself as the supervisor of elections or designee and I have access to more than just the counts for this current election? I have access to what else?
A count of all votes ever cast in the system?
And to get to that count is just a button click selection just as if I were trying to get the counts of the most recent election?
Who wrote this software?
Does this make any sense as a way to run an election?
Bob, a good software program for critical systems like voting will give the operator access to any information.
And putting roadblocks in the way of that information because someone might not want to see it in a certain situation is just bad user interface design.
posted
There was a more thorough article in one of our weekend papers here in Madison. It detailed how the problem occurred -- something about when the software was plugged into a laptop, the numbers were accidentally added (?). I just skimmed it. The interesting part to me was that it did not seem to be a Diebold-manufactured piece of equipment.
I will try to find the account.
Posts: 2919 | Registered: Aug 2004
| IP: Logged |
quote:And putting roadblocks in the way of that information because someone might not want to see it in a certain situation is just bad user interface design.
Hmm. In computer-based patient files a the hospital and clinics where I work, there are levels of access. I cannot access some information under my login -- whole areas are inaccessible, and unless I had a different person's login for somewhere in the hierarchy, I would be roadblocked from seeing that info.
Surely there is a hierarchy of operators with election software, no? So that access can be tracked and appropriately blocked or permitted?
Posts: 2919 | Registered: Aug 2004
| IP: Logged |
quote:Bob, a good software program for critical systems like voting will give the operator access to any information.
And putting roadblocks in the way of that information because someone might not want to see it in a certain situation is just bad user interface design.
So, you're telling me that the display for getting the vote for an election should be side-by-side with the display for getting the reference counts for every vote ever made with this system. So, one mouse click gets you the 2004 election results and the other gets you a diagnostic?
I'm sorry, but that's bad human factors.
If the system is to serve multiple purposes and types of users, the designers should at least have found a way to determine which type is clicking buttons that moment and given them a way to get at the information they need in an unambiguous way.
And, if that same user needs access to the less likely query, then he or she can reasonably asked to go through a verification screen before getting it.
That's SOP for any system, but would seem even more important for one where getting the appropriate count quickly and without confusion is important.
Of course, we don't know how this particular system was designed, but the design you seem to be advocating is not one I'd want to see out there as a human factors person.
(ps, I have studied this stuff from the HF perspective, by the way...)
PPS: Like what Sara's pointing to. Levels of access are crucial to proper use of most systems that hold onto any sensitive information, or require security from corrupting the data.
posted
I've made a lot of systems with a lot of queries. Number one complaint I've gotten from users - "why do I have to use so many clicks to get X?"
"Because you don't use X as often as Y" just doesn't fly with them.
Especially when dealing with retreival operations that don't alter data, excessive verification is wasteful.
Levels of access don't apply: This is reading information. The person responsible for retrieving the number of votes MUST be able to retrieve all information, precisely so they can investigate numbers that seem off.
posted
Dagonee, the fact this error was possible in your hypothesized system is argument enough for greater attention to human factors considerations.
When it matters most is when you have to stick to sound design principles.
And frankly, do you think that users of a voting system that is designed to report the tally for a specific election are going to have that much heartburn over having to verify their selection or go though one more button click to get to the diagnostic data?
And, if so, then you should have another way to display those data so that it is clear when the user is looking at the diagnostic display and when they are looking at actual vote counts.
Developers can't just throw up their hands and say "well the users want it this way" when they know that responsible use of the system might require that users might more cues than just the words on the screen.
Especially since many developers are fairly inept at putting discriptive labels for onscreen buttons.
It's also a mistake to rely on the training of users as a fallback.
Again, as the importance of the system increases and the sensitivity of the information it holds increases, the need for safeguards in the system goes up.
It's pretty clear that HF considerations were given short schrift in this system design if what you say happened did in fact happen. The proof is simply that it was so easy for the person to make this error.
posted
I think it was a Danaher machine, not Diebold. It was a touchscreen model.
Madison's current edition of the paper does not have the story online, so I am trying to find it elsewhere.
Posts: 2919 | Registered: Aug 2004
| IP: Logged |
posted
No, it's not. The mere fact that an error happens is not proof that the system was not designed correctly.
First, you'd have to show that any kind of verification would have helped. Oh, you say, simply add a pop-up screen explaining that this query is one they might not want? Users DON'T READ THOSE SCREENS. having extra warnings reduces the effectiveness of warnings that are critical, such as ones that alter or destroy data.
Move the link to this function elsewhere? Out of sight, out of mind. Doing so would mean this link doesn't get used. And the link not getting used could mean important verification procedures are skipped.
Face it, a complex system is going to allow user error. The processes built up around the system need to take this into account.
And, in case you missed the point, the UNOFFICIAL count was wrong. The error was caught before it was made official.
In other words, the process worked. As I said before, the the fact this error happened means we taker the time to check this out. For all we know, such a check IS part of the process.
posted
By Sunday or Monday, this article should be available online. I can't google it elsewhere, but I think it was off the wires (not a local reproter). Pretty comprehensive. I will keep an eye out.
Posts: 2919 | Registered: Aug 2004
| IP: Logged |
posted
Dag... we're arguing about a hypothetical system at this point.
There seems to be the possibility of human error here that was foreseeable and therefore it's also possible to make the error less likely.
The system you described does nothing of the sort.
Blaming the human is fine, but it's also important for system designers to pay attention to the points at which use of their system poses the greatest risks for human error and attempt to do something about it.
The side by side choices that you talked about (one button gives election results, the other gives a diagnostic screen) shows no evidence of even thinking about the possibility of user error.
Maybe you had something else in mind?
Posts: 22497 | Registered: Sep 2000
| IP: Logged |
posted
What I had in mind is that the process worked, and you're declaring the system designers incompetent without having nearly enough information about the system requirements.
I'm posing hypotheticals as counterexamples to your rather sweeping declarations on the subject.
posted
Actually, I was going to go back and change my original post to allow for human error as the culprit, but I decided that then it would make your post look like you were arguing about something I'd already acknowledged. So I left it.
Truth be told, the most likely scenario IS human error.
But, I don't understand how people who know the importance of the vote in Ohio could go ahead and make that error. It just smacks of something more than incompetence.
But, you're right, it could just be garden-variety incompetence.
And it probably happens a lot more than we know. It's just that this time the whole country was watching...
Posts: 22497 | Registered: Sep 2000
| IP: Logged |
posted
I'm actually interested in this phenomenon, having encountered it before.
When someone makes an error which is caught by the normal error-catching process, doesn't that validate the system? Does it mean it's an error? How far does it have to get before we should worry about it?
Obviously, the more important the process, the more we worry, and this is an important process. It also matters how close the error came to the end of the safety gauntlet.
In medicine, if the pharmacist picks up the wrong pill bottle, dispense the pill, but then realizes it's the wrong dose and corrects it, has he made an error?
If the nurse notices when she picks it up, we'd feel safe saying the pharmacist made an error. But the system as a whole has not, right?
If the bedside nurse notices the mistake when she's about to hand the pills to the patient, the system still caught the error.
If the patient is about to swallow the pill, and notices it's bigger and asks about it, we'd probably say the system made an error.
Each of these requires a response of some kind, but how which ones indicate systemic problems and which ones indicate errors?
Dagonee *This is entirely hypothetical at this point - I'm interested in process error management.
Posts: 26071 | Registered: Oct 2003
| IP: Logged |
posted
Man.. they'd better go double check the results everywhere else where electronic voting was used. If its that wide spread then lots of little 'glitches' like this could quickly add up to reveal that Kerry actually won...
I dunno how widespread Electronic voting is though. If its only in one or two places then its probably not that big a deal.
Its incredibly suspicious that all those extra votes ended up Bush.
Posts: 3295 | Registered: Jun 2004
| IP: Logged |
posted
How do they know all the extras are bush's? how do they know that the machines didn't just multiply all the canidates number of votes by something?
Posts: 2489 | Registered: Jan 2002
| IP: Logged |
posted
That was the "total vote", which the article makes clear they reconstructed correctly by looking at the individual totals on each machine and adding them together.
Posts: 15770 | Registered: Dec 2001
| IP: Logged |
There are people who study human error and system error.
The usual thing that one hears is that there should be no possibility of a single point failure bringing down the entire system. O-rings in the Challenger disaster were a classic example of single-point failure on the engineering (design) side of things.
Looking at that same thing from a Human Error perspective, however, there were multiple points at which it could've been corrected. It just wasn't.
Same with the Columbia. A wrong decision was made based on some very confusing presentations by Boeing which, after the fact, were shown to be irrelevant to the actual situation with the foam and the wing on the actual shuttle.
What good systems do is incorporate design and procedures into an overall safety system. And you're right, the larger the risk, the more important design factors and process/procedure are.
In the pill example, someone designed labels to be distinctive. Pills have varying shapes, sizes and colors. There are multiple cues in the environment that would assist people all through the process "chain" to discover an error.
Training is important too, of course.
Whenever it stops prior to the "bad consequence" then the system worked.
You have to decide what the points of the bad consequence are.
1) Stopping short of bad publicity might be one goal. 2) Stopping short of anyone noticing at all might be another goal. 3) Stopping short of killing people, etc...
So, sure, nobody died because a county released bad numbers in Ohio. They corrected it. But it caused bad publicity and raised some concerns. So, the system worked, but so far down the chain that it still has SOME bad consequences. It didn't call the entire election into question. Or maybe it did. Bad enough of a consequence.
In your hypothetical, the system "works" as long as the error is caught before the patient got the wrong meds.
But the rubrik is to catch the errors as early as possible. And if you catch it at the end of your process, that's an indication that you need more error checks earlier on. That's not a signal to be happy it worked, but a warning sign that some errors are probably getting through and you have no way of knowing how many!
You never want errors caught at the last step in your error check for exactly this reason -- you can't tell how many are getting through undetected.
Edit: Which, come to think of it, seems to apply quite closely to this vote machine case.
Printer-friendly view of this topic
![[Wink]](wink.gif)
![[Evil Laugh]](graemlins/evillaugh.gif)
![[Dont Know]](graemlins/dontknow.gif)
![[Smile]](smile.gif)
) count the ballots together, and double-check themselves and each other, and then report their numbers to the Supervisor, who then reports to whoever he reports to... It's really not that big a deal, you know. I mean, anyone can count.
![[Big Grin]](biggrin.gif)
. Have you worked in government, too?
