Slashdot Mirror
Maryland Tests Voting Machine, Declares Success
Dachannien writes "Capital News Service reports that the Maryland State Board of Elections has staged a test of its Diebold touch-screen voting machines in an effort to demonstrate their security and accuracy. A machine randomly selected from Maryland's voting machine warehouse was tested in a mock vote against two human vote-counting counterparts, and after counting fifty votes, the human vote counters had made several errors versus zero for the voting machine. But is this a legitimate test of the concerns of voting machine activists, or does it merely support a logical fallacy?"
Were they Diebold employees or something? That's simply not credible.
after counting fifty votes, the human vote counters had made several errors versus zero for the voting machine.
.MDB files. Election Day hasn't even arrived yet and already people have been busy introducing systematic error into the pool of registered voters. Even if the 2004 election involves pretty blinking lights, and is the most precise ever, it will undoubtedly be a less accurate measurement of the desires of the electorate than the election we had in 2000. This is what Stalin meant when he said that those who cast the votes decide nothing, and those who count the votes determine everything.
Did the humans make random errors (degrading precision), or errors tending to favor one candidate (worse, degrading accuracy)?
This looks like a test of precision, not accuracy, and Maryland is confusing the two. I wrote a long post a few days ago in another story that's relevant to this one, and makes the point I'd want to make here, so I'll be a lamer and paste it.
* * *
The punch card system proved itself to be a very accurate method of vote counting, even under the extreme condition of a tie- to a precision of several hundred votes. Much attention was paid to the relatively few cards that had chads hanging, but the vast majority of the cards were quite unambiguous in their representation of the voter's intent. Unfortunately they occurred in equal numbers for both candidates. The entire system was at least as auditable as any vote counting system can possibly be.
People don't understand the difference between precision and accuracy. Precision means that, given a measurable X, your measurements are sharply defined. But that is not the same as accuracy- which implies that the measurements actually reflect the true value of X, and not the influences of other sources of systematic error- like air resistance, or the thermal expansion of the ruler you're using, or the political affiliation of the manufacturer of your measuring equipment. A measurement is only accurate if sources of systematic error have been minimized. Sources of random error- like hanging chads- merely degrade precision.
The outcry for computerized voting that followed the 2000 election- to "bring our elections into the 21st century" and similar nonsense- was most unfortunate. We are making the transition from an accurate but slightly imprecise system to a new system that promises only extreme precision with no guarantees of accuracy. What is worse, we are about to trade susceptibility to random error for something far worse- susceptibility to systematic error- which is fundamentally different from a human perspective since it introduces a huge motive for people to screw with the accuracy of the electoral process.
The 2000 election had its share of systematic error. There was that butterfly ballot, which confused both Gore and Bush voters alike, but had the effect of transforming Bush votes into Bush votes and Gore votes into Buchanan votes. There was the Florida felon purge, which knocked thousands of blacks but only dozens of Cubans off the rolls. The 2000 election is still bitterly disputed, but very few people still complain about the hanging chads, which were sources of random error with relatively nonpartisan effects. The sources of systematic error had a much more corrosive effect- they cast doubt on the very legitimacy of the outcome, since they gave the election the appearance of having been stolen.
I have no doubt that we have an ultraprecise election ahead of us- computers are good at being deterministic, after all- but as far as accuracy goes- we'll see. There are many who would love to insert some systematic error into those Access
Yes, I'm serious. There's a software setting on the touchscreen to do one or the other.
But that's OK, 'cuz the software in there is "certified" and subjected to code analysis by a test lab, right?
Oooops. Diebold withheld thousands of lines of custom code in the voting terminals from review by declaring it "Commercial Off The Shelf" (COTS) software. Under FEC rules, "COTS" doesn't need serious scrutiny...but Windows CE at the terminal is NOT "COTS" despite Diebold's assertions otherwise - WinCE is a "software kit" that needs to be "finished" (mostly core drivers like video, etc) by the hardware manufacturer.
Diebold.
So hide a couple hundred lines of code somewhere that checks for the "election mode" versus "test mode" flag...
For more on this WinCE issue including Diebold internal EMails discussing it:
http://www.equalccw.com/sscomments2.html
That's not even getting into how screwed up the central tabulator software is:
http://www.equalccw.com/deandemo.html
Ain't no WAY you can trust a Diebold system. Period.
Jim March
Member, Board of Directors, www.blackboxvoting.org
The failure modes, however, should be considered using a metric weighted for both likelihood and impact of failure. A miscount is random; vulnerability to it is roughly equally likely to be for or against the candidate who would otherwise win. Sabotage on the other hand, where deliberate vote tampering occurs to a directed purpose, is highly likely to be for an undeserving candidate-- almost by definition. As such, I'd consider vulnerability to tampering a bigger problem than an equal probablility of a vote miscount. And the most effective way to reduce vulnerability to tampering is to increase transparancy to (multipartisan) human observers-- which the proposed electronic methods need to work on.
I believe it is possible for electronic voting to allow for an improvement in both usability and security in the electoral process over current methods. I do NOT believe it represents that improvement yet.
//Information does not want to be free; it wants to breed.