[E-voting] experience with VVAT

Michael McMahon michael at hexmedia.com
Mon Mar 27 12:11:54 IST 2006


Colm MacCarthaigh wrote:

>On Mon, Mar 27, 2006 at 11:03:28AM +0100, Michael McMahon wrote:
>  
>
>>I think if we had to design a system *today* from scratch for the
>>Irish election system, this is the one I'd go for.  Even at this
>>stage, given the cost of storing the DREs (€1 million per year) it
>>could make sense to just scrap them, and buy scanners (obviously a
>>much smaller number of them) instead.
>>    
>>
>
>I'm the complete opposite. I'd buy ballot printers, and nothing else.
>Gets rid of confusion about possible spoiled ballots and helps the
>voter make sure their ballot is cast correctly.
>
>  
>
That seems like a very expensive solution for the problem of unintended 
spoiling
of votes. It would be a lot cheaper to just run a few ads on the telly, 
explaining
how not to spoil your vote, which I think should be done anyway.

Though I don't see a problem with it in principle. It's just an issue of 
cost.
Ballot printers for people who are visually impaired might not be a bad 
idea,
but hardly necessary for the majority. Where we really disagree is on 
the need
for human counts.

>>Scanning hand-written numbers is obviously more complicated than the
>>simple marks in this system, but OCR software is easily capable of it.
>>In fact, I would see it as an advantage, that gives the audit more
>>purpose (in a practical sense) because of the inevitability that some
>>people's hand-written scrawl would be unrecognisable and it is
>>important that the machines correctly identify these ballots.
>>    
>>
>
>Having been on several spoils ajudication committees, I can tell you it's 
>impossible. Besides, how could we know when to audit the output of the
>scanners in the ordinary course of events? They would be subject to
>potential fraud themselves.
>
>  
>
Right, but the scanners would not be making any such adjudication. They 
would be calibrated
to reject anything that does not look like a very clear set of 
preferences. The rejected ballots
would have to be adjudicated manually as before. The purpose of the 
audit is then to check
that the scanners are actually doing what they are supposed to do..

And like I've said many times before, the audit has to be well defined, 
mandated
in law, and standard procedure for every election. There could be  a QA 
process where
ballots are checked randomly as they are scanned in each batch, but the 
important step
is the audit itself, which would be done after the official electronic 
vote file is published.
Then, samples of ballots would be taken at random
and checked individually against this file. Any errors would require the 
relevant batch to
be manually validated (ie. each ballot in the batch would be checked to 
be the
same as its electronic equivalent).

I'd see no need for human counts, except in the unlikely event that 
someone proves the
official system made a mistake.

Michael



More information about the E-voting mailing list