Page 18 - Towards Trustworthy Elections New Directions in Electronic Voting by Ed Gerck (auth.), David Chaum, Markus Jakobsson, Ronald L. Rivest, Peter Y. A. Ryan, Josh Benaloh, Miroslaw Kutylowski, Ben Adida ( (z-lib.org (1)
P. 18
10
E. Gerck
5
Intuition
14
who are ideally honest and error-free ran an
What if perfect human clerks
election, would the result be trustworthy?
No, not necessarily. Trustworthiness of the election outcome would still depend
on whether a number of requirements are met, such as that no one can vote more
than once or vote on behalf of another. This example refutes the oft repeated
idea that the trustworthiness of an election is entirely dependent on the people
who count the votes. Even if the people who count the votes are ideally honest
and use flawless devices with flawless software, the trustworthiness of the election
still remains elusive.
Election outcome trustworthiness depends on voters, but not only on their
honest behavior. It also depends on voters understanding the instructions and
being able to let their intent be represented accurately and reliably in the votes
they cast, unlike with the notorious Florida “butterfly ballot”.
What if the election is not run by such ideally honest human clerks or comput-
ers? We expect that additional requirements would have to be imposed in terms
of election outcome trustworthiness. We may require that ballots must be han-
dled in the presence of at least two clerks. But, would two clerks allow “enough”
risk reduction to allay fraud concerns? What do we mean by “enough”?
Many will recognize that not only there is no good model to design these
important requirements to, and one must make up the rules by trial and error,
but there is no inner metric to comprehensively define “enough” in terms of risk
for each step and for the voting process as whole.
What we are missing is a better model. The deterministic model described by
“cast ballots, then count them” fails to provide us with guidance for improving
the trustworthiness of the election outcome.
Instead, we start with the fresh observation that avotingsystem is an in-
formation transfer system. The tally results contain information that was sent
from each voter, where, according to Information Theory [9, 11], information is
essentially uncertain in nature —although voting is deterministic.
Uncertainty which arises by virtue of freedom of choice on the part of the voter
is desirable uncertainty. Uncertainty which arises because of interference on that
freedom of choice (e.g., caused by faults, attacks and threats by adversaries) is
undesirable uncertainty, which we call errors. We are, thus, motivated to include
the possibility of errors by means of a probabilistic description of undesirable
uncertainty. Information Theory provides a good correspondence here as well,
where the undesirable uncertainty is called interference (or noise).
In such a model, voting system requirements are not arbitrary. Requirements
are created and dictated by the goal of minimizing interference. The intuition is
that an information-theoretical voting model should be able to optimally combat
interference and, thus, improve the trustworthiness of the election outcome.
14
Or perfect computers. Originally, and as late as the 1920s, computers were defined
as human clerks who performed computations. Computers were expected to perform
obediently with paper and pencil, for as long as it was necessary, but without insight
or ingenuity.

