If we build it, will they verify?
Assuming we can build the magic web site, what would it take to recount those ballots as described, independently, off the web site? Let's posit 1 second response time per image. For a state the size of Minnesota, that's 2.5 million seconds, or 30 days. To independently check the national vote... that's 4 years.
Of course, we could divide up the verification work among multiple people. All of whom you'd have to mutually trust. So how many people would need to converge on the outcome of the web site audits in order to trust the result? How many people would have to differ, and by how much, to actually go look at the paper?
Whose verification software will they use?
Would all of those people who wish to verify the tally use precisely the same software to interpret the web ballot images? If so, who provides that software, and why should we trust their honesty, or the program's accuracy in interpreting ambiguous marks on paper? After, a well known web security technique (CAPTCHA) you may be familiar with relies on the difficulty of accurate software-based interpretation of arbitrary marks in a image file - and the ease of humans, at doing precisely the same!
Even considering CAPTCHA, many of my colleagues persist in their blind technocentric faith that people are unsuitable processors for the interpretation of ballot marks, but "dumb" optical scanners are; and that people are so challenged to count way up, to a few hundred or so, in public, every couple of years that they will continue to need machines to help them.
But even if the ballot interpretation software is magically precise, since ballot format varies by state, county, and precinct ... who will update the ballot verification software to handle these variable formats - and why should we trust them? On the other hand, if everyone is free to provide their own ballot interpretation software, what would prevent someone from tailoring a program to interpret ambiguous ballot marks in favor of their party and candidate, and disputing the count made by partisans of the other candidate?
Election verification considered as a game
I suspect Rivest and Smith approach elections as an especially intriguing kind of intellectual puzzle, and they expect the participants in the electoral process to resolve such electoral disputes with the same kind of reasoned peer-review process we find in academia.
On my part, I predict an unending series of recounts and disputed outcomes, on top of a yet more new and expensive computer infrastructure of Byzantine and unnecessary complexity. As is now the case, in this proposed new scheme, the only elections we will know the outcome of with even moderate certainty will be the ones we recount by hand.
How much simpler, to just count it by hand in the first place?
So even though Rivest and Smith's scheme would require a major investment in new IT infrastructure, be extraordinarily difficult to secure, could endorse fraudulent elections, and, in practice would be unverifiable by the public via the web site images... at least it's also far more complex and vastly more expensive than the alternative. That alternative is simply counting the ballots in public, by hand, with multiple witnesses - the only voting protocol, by the way, considered robust enough for use in war zones.
And just one more thing: would it be too much to ask people who propose ivory-castle web-based systems to do a reality-check with the professionals?
1 | 2
