[EM] CORRECTING Black box voting repost re how HAVA imploded

[Brian Olson]

Here's a data point on scanners:
http://catalog2.panasonic.com/webapp/wcs/stores/servlet/ModelDetail?displayTab=O&storeId=11201&catalogId=13051&itemId=93888&catGroupId=13856&surfModel=KV-S3105C
100 page scans per minute with 1000 page feeder, $1300. Definitely not 
bad.

Then it's just a simple matter of writing good OCR software to read the 
ballots. I could do that, though it's not as simple as getting a human to 
read the ballot and punch numbers on a keyboard. Oh, but here's a group of 
people trying to do basically all this already:
http://www.openvotingconsortium.org/

Optical scan equipment is in wide use from many vendors including the 
infamous Diebold.


Rivest pointed out in his discussion of 3-ballot that posting actual 
images of ballots is bad because it allows for covert channels of 
information which destroy secrecy. Some thug can instruct you to make such 
and such mark astray on your ballot, then look for it to make sure you 
voted the way he told you to (or paid you to).


The primary good cause used as justification for HAVA, the Help Americans 
Vote Act, is to allow disabled (blind, quadruplegic, etc) people to vote 
with only the help of the machine. I think the tradeoffs and costs of that 
good cause as its being implemented are winding up being really bad. 
Perhaps the evil lead the good down the road to hell by paving it with 
good intentions.

Brian Olson
http://bolson.org/

On Thu, 1 Feb 2007, Abd ul-Rahman Lomax wrote:

> I've made the following proposal for public elections.
>
> The ballot is designed to be scannable with reasonable accuracy. Like the 
> multiple-choice tests that have been optically scored since long before PCs.
>
> Each ballot would be coded, in batches as large as a precinct, after being 
> cast, to allow identification of each ballot by precinct and a unique number. 
> The code could be added with a sticker, but it might be better to use a stamp 
> or a printer to add the code. The batches would not be coded excapt in 
> batches large enough, and by appropriate procedures, to make identification 
> of a voter with a ballot impossible.
>
> Ballot images would be public information and would be openly accessible.
>
> There would be an official scan, the raw images from this would be made 
> available. However, election observers could also scan the ballots or at 
> least obtain images of them. (After the ballots are numbered.)
>
> Consider the implications: optical recognition technology is sufficiently 
> ubiquitous that a decent automated analysis of ballots could be made 
> independently by many different means. However, it would also be quite 
> practical for visual counting to be done *of the images*, by anyone; the 
> results could be tabulated in a way that makes controversies about the 
> meaning of a ballot reviewable; this is one reason why the ballots would be 
> coded, so that results of various analyses could quickly and easily be 
> compared.
>
> The official results would include not only the total counts, but a 
> ballot-by-ballot determination of what each vote was that was counted. And 
> each ballot that was invalidated, for whatever reason.
>
> Essentially, the counting process would take place in public, not only 
> watchable by a few observers, but by anyone interested, at leisure.
>
> The "scanners" could simply be fax machines. No special equipment would be 
> necessary. The *analysis* could require special or customized software, but 
> it is quite possible that there are off-the-shelf solutions, that could 
> convert a ballot into a standard text file.
>
> I believe that ballots appropriate for this are already being used in places.
>
> The "official" scanning and results need not require any new equipment. I've 
> never understood the fuss over new election equipment as being anything other 
> than a boondoggle. Quite simply, hand counting is not difficult enough, and 
> counting is needed infrequently enough, that the kind of spending which has 
> been poured into election equipment is little short of insane. How much per 
> vote is being spent, and how much would it cost to count ballots by hand?
>
> There are standard keypunch procedures, developed years ago, for obtaining 
> high accuracy by using multiple operators and comparing the results. With 
> coded ballots, noncontroversial ballots could quickly be identified and 
> attention focused only on controversial ballots, which would normally be a 
> relatively small percentage. What happened in Florida in 2000 was that *every 
> ballot* was counted by a whole group of people, one ballot at a time. I don't 
> believe they did independent batching. Highly inefficient and not verifiable.
>