[E-voting] Minimum requirement to juge voting "technology"

[Colm MacCarthaigh]

On Fri, Sep 10, 2004 at 05:38:11PM +0100, Timothy Murphy wrote:
> On Friday 10 September 2004 17:03, Colm MacCarthaigh wrote:
> 
> > But do you have a proposal, and indication, an idea, anything in
> > fact which would aid someone to discern between what is a reasonable
> > request and what isnt?
> 
> I've already answered that question - I would leave it to the person
> in charge of the election (maybe the Returning Officer), with anyone
> having the right to appeal to a court if they felt aggrieved.

And how would they know? And how would I know whether I should even
feel aggrieved?

In a traditional count, the failure mode is discrete and relatively
probabalistic. With human counting, it easy to understand that a margin
of say 5 votes is more likely to have resulted in an incorrect ordering
of candidates than a margin of 100 votes. Since it's more likely 3 votes
could be miscounted by humans than 51.

With electronic counting, this is not so obvious. There is no
probabalistic assessment of likelyhood of error versus various magins.
While an extremely great or small margin, or a series of power-of-two
margins might raise suspicion there is no obvious reason to suggest that
a margin of 5 could be more or less error or tamper prone than a margin
of 100.

So, how do we ever establish criteria? Is feeling aggrieved enough?
Most losing candidates feel aggrieved, and presented with no evidence
that there could be a miscount, shouldn't they *always* ask for
a paper-based count?

> > In an election where candidate A gets 1600 votes, and candidate B
> > gets 1400, how I guess whether tampering or error is likely or not?
> 
> Political activists usually have a pretty good idea of how candidates
> are going to do, eg from exit polls and from the response to
> canvassing.  If a candidate from one party did noticeably worse in one
> constituency that the party did nationwide, it would be reasonable to
> enquire why this was so, and if there was no apparent reason then the
> paper votes should be counted.

In my experience, speaking as a longtime political activist and
canvasser, Political activists usually overestimate how well they will
do. But dispensing with this, pollsters, politicians and activists
are frequently surprised by election results.

If the software has been tampered, or if error is present, it may very
well be there on a national basis, and it's entirely possible the
effects are only visible in marginal constituencies. But this goes back
to my original point, there is no probabalistic breakdown of margin-size
Vs error-likelyhood.

> Nb ICTE has won a battle but it hasn't won the war.  If you put
> forward a sceme that seems as cumbersome as the present system then
> VVAT will be rejected by the political parties.

I don't think a system need be as cumbersome as the current, for one
thing printed ballots instantly dispense with at least an hour or two of
spoils reckoning and would eliminate most of the human error in the
system.

> IMHO, the argument should be, "You only have to make one slight
> adjustment to the present system, to print out the ballots.  Then you
> will have all the advantages of e-voting and you will justify the
> expense of all these magnificent machines."

The machine expenditure is lost, hopefully they will never be used.
Mainly because they are terribly inaccessible (small screen size, small
writing, confusing user interface) and are totally useless to the blind.

-- 
Colm MacCárthaigh                        Public Key: colm+pgp at stdlib.net