[E-voting] Question on optical scan

Adrian Colley aecolley at spamcop.net
Thu Jan 20 00:14:20 GMT 2005

On Wed, Jan 19, 2005 at 02:14:06PM +0000, Margaret McGaley wrote:
> Another, more expensive, option would be to have some device which helped
> voters cast their vote, but didn't record the vote electronically. This system
> would have more user interface issues, I feel, and would seem like a waste
> (you can imagine a response that said "why doesn't the printer just save the
> vote"), so I'd be more comfortable advocating the stickers.

I can't help being reminded of the devices which helped many of
Florida's voters mark their ballots by punching holes in them, with
varying degrees of success due to wear and tear on the punch devices.  I
find Brian's warnings (of chad-like problems with stickers) convincing.
It's hard to think of a ballot-marking tool (other than a simple pencil)
that might be immune to similar problems.  Maybe the National Disability
Authority has some ideas on how best to enable voters with impaired
vision or dexterity to vote in secret.

On Wed, Jan 19, 2005 at 03:42:44PM +0000, Brian O'Byrne wrote:
> A variation on this theme that would overcome most of the concerns I raised:
> Have a sheet of transfers instead of stickers. Have a box beside each candidate.
> Place a transfer above that box and scratch to transfer it onto the ballot.
> Being simple ink instead of a sticker it cannot be removed and cannot stick
> things together. It can be dry and indelible instantly.

My only worry is that it might be easy for a voter to half-stick the
transfer in one of a number of unexpected ways (e.g. half-scratched,
peel-and-lick as if with a postage stamp, scratch-oops-peel-restick),
following which the transfer could unstick and become a "sticky chad".

On Wed, Jan 19, 2005 at 03:50:14PM -0000, Aengus Lawlor wrote:
> Even with a pencil and paper system, we're not talking about
> "handwriting recognition" - we're talking about recognizing characters
> from just 10 digits. With clear examples posted showing how to write the
> numbers, there's a fairly small amount of room for confusion, and what
> room does exist will be lessened by the fact that the numbers aren't
> random - you can't have a 7 if you don't already have a 6.

On Wed, Jan 19, 2005 at 04:57:13PM +0000, Ryan Meade wrote:
> It would be interesting to confront a modern OCR system with a
> real-world sample of Irish ballots - I suspect it would perform quite
> well.  Ballots are usually filled out quite clearly, and any ballots
> I've seen which might present an OCR system with a problem would be
> considered 'doubtful' under our current system anyway.

I imagine the simplest effective thing in the end will be pen/pencil,
paper, an informational poster showing how numerals should be shaped for
maximum clarity, and very picky OCR which will reject every doubtful
ballot.  Rejected ballots are returned to the voter, who gets another
go unless the presiding officer loses patience.  This would eliminate
spoiled votes and ensure that the electronic and paper records will not
get out of sync by reason of operator error alone.

I notice that Rebecca Mercuri is now advocating optical scan too.


GPG 0x43D3AD19 17D2 CA6E A18E 1177 A361  C14C 29DB BA4B 43D3 AD19

More information about the E-voting mailing list