Flutterby™! : Voting machine malfunctions 2008-10-19 12:56:14.516004+00

Voting machine malfunctions

2008-10-19 12:56:14.516004+00 by Dan Lyke 5 comments

Chris forwarded along this note about early West Virginia voters angry about malfunctions in electronic voting machines:

Virginia Matheney and Calvin Thomas said touch-screen machines in the county clerk's office in Ripley kept switching their votes from Democratic to Republican candidates.

I think it's Grey's Law that states "Any sufficiently advanced incompetence is indistinguishable from malice." I've seen what the touch screens on the credit card machines at the grocery store do to my signature. I notice that ATMs have gone back to buttons along the side. I've looked a little bit at touch screen technology. Doing touch screens right is hard, it's not a mature enough technology to be using in applications like this. If you must use technology in voting systems, use scannable paper ballots, and recall any voting officials who don't have something that leaves a solid physical audit trail.

"Never ascribe to malice, that which can be explained by incompetence." -- Napoleon Bonaparte "

A corollary might be: "You can almost always try to blame malice on incompetence."

I like to paraphrase J Edgar Hoover, and say "Never send a person where you can send an electron." But I also accept that there are somethings that are not likely to be, or should not be, automated within my lifetime. Voting is one of them.

I can't see using any electronic voting. It is subject to fraud at all levels, from the original collection of the data, and through tabulations at the precinct, county and state levels. Even scanning machines for counting votes have been reported to have been fraudulently programmed.

Electronic voting machines can never be completely verified. The complexity level is such that it is likely that the software ostensibly used can not be verified. Such solutions are almost always incomputable in finite time. Never mind the potential for software substitution, human abuse and simple blunder -- yes I will include the merely incompetent here.

The hardware itself must always be suspect. A cellphone in a voting machine could report vote counts and allow remote changes that would not raise statistical flags on vote counts. How hard would it be to hide the communications circuits for a cell phone in any type of electronic voting machine hardware. Less the batteries, touchpad, screen, audio, gps, camera -- there is not much physical space required. How difficult would it be to detect such a device spread out over all the electronics in the voting machine itself?

Canada uses paper ballots counted by human beings under observation. Canada, as spread out as it is, still manages to get an accurate count and tabulation within four hours of the close of the polls. We can't manage that with our current technology. Seems like voting is one of those things that are best left to humans.

#Comment Re: made: 2008-10-19 23:39:15.491664+00 by: TheSHAD0W [edit history]

I'm sending "This doesn't happen in America! Maybe in Ohio, but not in America!" off to my parents and sister in Ohio.

I've been reading Hubert Dreyfus' What Computers Still Can't Do. As an aging programmer, it's been a very interesting read. There seems to be an interesting parallel between the pursuit of AI and the attempt to automate voting. This comparison happens to be of interest to me because cartographic representation happens to be very similar to human reason and vote tabulation in its difficulty to automate.

Voting is a misleading problem because it seems like it's guided by simple heuristics:

The problem is the "Voter selects votes" bit. First, you have to properly identify a "voter". This is pretty challenging for a machine. See my previous rants about passwords and you'll get the picture. Second, making a selection has to be foolproof. No ambiguity. No software errors. No hanging chads. Finally, the vote has to be made somewhat persistent in a way that relates it to the voter (but in an anonymous context).

There has been great efforts in user interface design that could improve the second problem. Transaction processing systems are designed to manage the third problem. But we are still a long ways from the first problem - identification.

The real challenge to identification is privacy. In order to be identified accurately, you must provide quite a bit of information. I just got my new All- Government credential and while they didn't actually draw blood, they did just about everything up to that point. And this is simply a credential to be used portably in offices of the various branches of the federal government.

At the same time - while we gripe about and point out flaws in electronic voting machines - we don't really consider how secure traditional forms of voting are. I recently voted by mail. How hard would it be to counterfeit a mail-in ballot? How hard would it be to steal a bunch of mail-in ballots from a post office?