13 August 2008

A really bad idea in the history of bad ideas

| johnboy
Join the conversation
34

The Canberra Times reports that electronic scanning is going to count the vital preferences on our ballots.

Having spent more years than I care to think of working with the results of electronic scanning I wish I could say I share their faith in machine ability to accurately read hand-writing.

Remember this is a fall-back after they arsed up electronic voting, which has now been withdrawn despite no public admission of what went wrong.

And here’s one for the conspiracy theorists. Hand-writing recognition anyone?

“Ah wait,” I hear you cry, “If the hand-writing recognition can identify voters surely it can at least count the votes right?”

Well if it can only correctly identify 20% of voters it makes a lousy vote counting system but still a massive potential intrusion and of great interest to political number crunchers.

I’m sure this is not happening now. But I’m in no way keen on opening this tempting path.

Anything that can be done often seems to be done eventually.

We can wait a few days for our election results, and we can afford to pay people to enter the data.

UPDATE: Thanks to Jonathan Reynolds who has chased up the EC and confirmed that the electronic voting will be continued, but still in an extremely limited implementation.

Join the conversation

34
All Comments
  • All Comments
  • Website Comments
LatestOldest
Jonathon Reynolds7:49 am 21 Aug 08

Found this: http://www.elections.act.gov.au/pdfs/scanningconsultationpaper.pdf

Still looking for more details on what they will actually be using for the election.

johnboy said :

From what they’ve been saying I think they have to be parity checking as each ballot is scanned rather than batch processing.

Frankly both options have serious consequences.

best to stick to manual counting.

otherwise the wrong lizard might get in…

From what they’ve been saying I think they have to be parity checking as each ballot is scanned rather than batch processing.

Frankly both options have serious consequences.

johnboy said :

Dalryk the software will have to store the image somewhere, (i’ll guess as a TIFF).

Who gets access to that data in the future is my concern. In the past there was physical security to that data and duplication would have been a massive undertaking.

I’m going to assume each ballot will be 5MB scanned.

Call it 200,000 ballots for the sake of simplicity and you’ve got a terrabyte of data. So an ~$1,000 storage array which tucks under the arm can carry it away. (feel free to correct my maths anyone)

Suddenly it is amenable to a level of analysis which it was not previously and I think we should maybe at least have a conversation about that before deciding to proceed.

Even if you lacked a handwriting sample to tack a ballot to a voter you could still, over multiple elections, build up a profile of individual voting patterns.

JB,

using OCR with storage facility will complicate this requirement. Using ICR, through a sheet feed scanner of 1000 pages, single sheet, with auto recognition and image purge functionality will mean that you will have

a) a pile of freshly scanned documents
b) a tally as defined by a scanner and the ICR software
c) opportunity for lots and lots of mistakes re marking, pressure of the voter (re handwriting, not the mob), and other forms of problems,
d) some poor soul who will need to crosscheck the results in a team, sorting through all the votes – much like a tally room.

we aren’t ready for automation yet. they need to find another way to present the vote to the public, perhaps via a touch screen system at a kiosk.

paper based is still too difficult for a computer program to represent efficiently.

Aurelius said :

PeterH, while ‘worst idea in the history of Stanhope Police State’ is a fairly contentious title, I think electronic ballots would pale next to “Oh, let’s not tell anyone the fire’s coming.”

but we aren’t talking about the other glaring omissions jon has made…

Mælinar - *spoiler alert* I've seen S04E1312:44 pm 13 Aug 08

I reckon ‘just tell them we’re going to spend several million dollars building a demonstration facility’ might be up there as well, given the topic is hard to swallow.

I predict that there will be no man hours saved, and if there is a recount it will probably have to be done manually anyway.

Typical ACT Govt, like the little brother, always going ‘hey, I can be the first to do . . . . [insert radical idea here] and show ’em we are the [choose from] coolest, most progressive, most human rights aware, most radical, most politically savvy !’

Why don’t we also have aromatherapy and foot massages by masked robots on the way in to the polling booth, to REALLY make voting day an exciting technical experience. Fair dinkum.

Dalryk the software will have to store the image somewhere, (i’ll guess as a TIFF).

Who gets access to that data in the future is my concern. In the past there was physical security to that data and duplication would have been a massive undertaking.

I’m going to assume each ballot will be 5MB scanned.

Call it 200,000 ballots for the sake of simplicity and you’ve got a terrabyte of data. So an ~$1,000 storage array which tucks under the arm can carry it away. (feel free to correct my maths anyone)

Suddenly it is amenable to a level of analysis which it was not previously and I think we should maybe at least have a conversation about that before deciding to proceed.

Even if you lacked a handwriting sample to tack a ballot to a voter you could still, over multiple elections, build up a profile of individual voting patterns.

PeterH, while ‘worst idea in the history of Stanhope Police State’ is a fairly contentious title, I think electronic ballots would pale next to “Oh, let’s not tell anyone the fire’s coming.”

dalryk said :

I feel the title of this post is unwarranted.

Number recognition is pretty simple from a computer scanning point of view. And pretty reliable. Potentially *more* reliable than hand data-entry, which is the alternative (since you pretty much have to use a computer to tally the results in our complex system).

So, bad idea? not so much.

And as for handwriting recognition from a bunch of numbers? A, the software isn’t designed to do that. B, even a human handwriting expert couldn’t do it. And C, why in the hell would anyone outside of a police-state want to do it?

Worst idea in the history of bad ideas? maybe something like ‘Communism’, or ‘using mercury to treat headaches’ might better fit the bill.

what about:
Worst idea in the history of the current ACT Government?
or
Worst idea in the history of canberra?
or
Worst idea in the history of stanhope police state?

I feel the title of this post is unwarranted.

Number recognition is pretty simple from a computer scanning point of view. And pretty reliable. Potentially *more* reliable than hand data-entry, which is the alternative (since you pretty much have to use a computer to tally the results in our complex system).

So, bad idea? not so much.

And as for handwriting recognition from a bunch of numbers? A, the software isn’t designed to do that. B, even a human handwriting expert couldn’t do it. And C, why in the hell would anyone outside of a police-state want to do it?

Worst idea in the history of bad ideas? maybe something like ‘Communism’, or ‘using mercury to treat headaches’ might better fit the bill.

Jonathon Reynolds said :

Computer Says no….

ROFL!

priceless…

Jonathon Reynolds said :

With electronic counting there is at least a fall back position whereby a manual count of the physical ballot papers can still be undertaken.

Agreed JR.

But I still prefer scrutinised double manual entry with cross checking. To “Computer says…”

johnboy said :

To put it more technically.

As a preliminary sorting method OCR is probably very efficient for the postal service, but they have multiple levels of checking thereafter right down to the postie sticking it in your letterbox.

What’s being proposed here on face value is a much greater reliance on accuracy, and for something rather more important than a parcel taking an extra day to get to its destination.

the accuracy would be down to the type of scan. maybe the voting public need to get rid of this idea, removing the glorious leader into the bargain.

Jonathon Reynolds11:28 am 13 Aug 08

@johnboy:

With electronic voting there is no audit trail and no fallback position as there is no proof other than blind trust that the votes being counted were actually the actual votes that were cast.

With electronic counting there is at least a fall back position whereby a manual count of the physical ballot papers can still be undertaken.

To put it more technically.

As a preliminary sorting method OCR is probably very efficient for the postal service, but they have multiple levels of checking thereafter right down to the postie sticking it in your letterbox.

What’s being proposed here on face value is a much greater reliance on accuracy, and for something rather more important than a parcel taking an extra day to get to its destination.

Auspost can afford to get 30% failure on postcodes and still reap a benefit from automated scanning.

“There’s no 22 Acacia Avenue in the 2614 postcode? Hmm, maybe that’s a 7. Etc.

Also four digits with limited permutations are orders of magnitudes less complex than an ACT ballot.

I’m pretty sceptical of this stuff too (there’s no way I’ll be voting electronically for the forseeable future), but the handwriting recognition problem is considerably reduced when it’s just numbers you’re talking about – Australia Post has been doing this with the “postcode squares” for more than a decade now. Anyone know what their failure rate is like? Presumably there is still a role for the scrutineers to check OCR’d ballots, too?

Jonathon Reynolds10:58 am 13 Aug 08

@Johnboy:

The CT article appears contradictory..

In the 2004 election, 28,169 votes were recorded electronically. This year Mr Green is expecting 30,000.
Twenty machines at five pre-polling stations in Belconnen, Gungahlin, Civic, Woden and Tuggeranong will record electronic votes from September 29.
”Last time we put all the pre-polling votes on to the website at 10 [minutes] past six on election night.”
Mr Green said 16 of the 17 winning candidates were correctly identified.
There will be no electronic voting on polling day.

When I rang the electoral commission this morning, and yes I am anally retentive and perverse enough to do such a thing to confirm details before posting a comment, I was told categorically that the voting machines would be used in five pre-poll stations then also used on election day.

If we are referring to ABC 666, I heard that too. I have more faith in a system that provides the mechanism to be able to go back and manually count the hand scrawled votes than that of a totally electronic system. A candidate/party involved in the election can always request a recount which can be highly manually (human eye visually) scrutinised.

Harold Hird countenanced that option in 2001:
http://www.canberratimes.com.au/news/local/news/general/hird-to-ask-court-for-recount/342173.aspx

Jonathon Reynolds said :

Oh and in regard to your conspiracy theory… I’d make sure you are wearing thick rubber gloves when you mark your ballot paper… you don’t want those nasty political parties working out who you are and how you voted. It is all too easy to use fingerprints from the ballot paper and track traces of your DNA these days 😛

When the taxpayers is footing the bill for the gathering of that information you’ll have a valid comparison JR.

The article clearly states that they’re withdrawing them, the EC has still made no public statement on exactly what they’re doing. Perhaps they’d like to do so.

When interviewed this morning they were talking about spitting results back to manual intervention where the OCR software was uncertain.

This still leaves the many cases where OCR is absolutely certain and completely wrong.

Jonathon Reynolds10:35 am 13 Aug 08

@iohnboy:

Remember this is a fall-back after they arsed up electronic voting, which has now been withdrawn despite no public admission of what went wrong.

You are totally incorrect – electronic voting will be used again at the upcoming election. Give Elections ACT a call yourself on (02) 6205 0033 to confirm.

The electronic voting system remains susceptible to manipulation of votes either during the casting of an actual vote or after a vote has been cast. This is because the system continues to lack a voter verifiable paper audit trail (something that we could go back to to scrutinise the electronic vote against). You have to have blind faith that what you intended to lodge as your vote is actually what gets recorded electronically.

I have more confidence in a system that is scanning and recognising data from ballot papers as there is still the ability to scrutinise the original hard copy, manually marked ballot paper.

Oh and in regard to your conspiracy theory… I’d make sure you are wearing thick rubber gloves when you mark your ballot paper… you don’t want those nasty political parties working out who you are and how you voted. It is all too easy to use fingerprints from the ballot paper and track traces of your DNA these days 😛

VYBerlinaV8_the_one_they_all_copy10:32 am 13 Aug 08

Security through obscurity is bad as just because something is very unlikely to happen doesn’t mean it will never happen.

In some instances security through obscurity actually works against you. When evaluating new cryptographic algorithms the security community around the world is often invited to review and attack the algorithm. In such instances, the strength of the security is in it’s genuine effectiveness, thus removing reliance on maintaining secrecy.

Different country, different voting system….& I don’t think the AEC really had much to do with it.

Leave that to the AEC, they do a stand-up job.

Because blind faith in electoral commission’s decisions without public investigation into the counting machine has never caused problems or controversy in the past.

Common though people….handwriting recognition? Identify 20% of voters? How exactly? Where are the handwriting samples to compare with for that 20% of voters? It’s not up to me (or any of you) to minimise potential exposure. Leave that to the AEC, they do a stand-up job.

Security through obscurity is bad as just because something is very unlikely to happen doesn’t mean it will never happen.
It should still be a part of somebody’s risk matrices, and potential exposure to it should be minimised.

Always glad to give them a giggle.

> Have you met many party hacks?

Plenty. I married one even. I guarantee that they’d be laughing heartily at your paranoia.

Have you met many party hacks?

I guarantee there are sad enough people who will be using this data if they get the chance.

Now if you’re cool with that then that’s great, but such a major change to the mechanism of vote counting, for mine, should not have been sprung on us so late in the day.

Security by obscurity. Who gives a flying F*&% how people who write with loopy twos vote? Just because you could in theory find such obscure things out, it doesn’t mean anybody would actually be interested enough to bother.

I’m still not sure I want anyone with the dataset on a portable HD being about to figure out the voting trends of people who write with loopy twos.

Graphology has some benefits, but it is too broadly inaccurate to be of any real use to the Government. So you can all settle down and take the Alfoil of your heads now.

oh, good.

hand writing??

after doing a count one election, hand writing is a bit of a stretch. perhaps they should give out crayons – fill in the blanks with your numbers, people – they might just keep in the squares….

seriously, though, if they are going to use the ICR technology to pick up votes, the system better be tested a few hundred times. The form scans are ok, but if they have someone who has botched the form, is too embarrassed to ask for another, and has written the info beside the boxes, it won’t get picked up by a traditional form process scan.

why don’t they just put in a series of coloured handles – push em down in sequence of your vote, punch card gets punched with the sequence, very old computer records your vote.

Daily Digest

Want the best Canberra news delivered daily? Every day we package the most popular Riotact stories and send them straight to your inbox. Sign-up now for trusted local news that will never be behind a paywall.

By submitting your email address you are agreeing to Region Group's terms and conditions and privacy policy.