IEEE Technology and Society Magazine - June 2018 - 85

literacy level of the general public in computer science,
however, there is no analogue whatsoever for public
understanding of algorithms. If rules and regulations
were to be expressed in code, only a tiny fraction of today's
society would be able to read them.
There is compulsory need, for both citizens and governors, be able to code the rules that govern the algocracy in
a format that is understandable to policy makers and to
the public. We also need to make sure that we will be able
to understand whether the code is actually serving the
purposes it has been built for, or if it is instead bugged or
hacked. In this regard, we envision three key requirements: better programming languages, inspectability of
code, and users' computing literacy.
Concerning programming languages, we emphasize
that future algorithmic governance for Smart Cities
needs to be inherently distributed and mobile. Accordingly, programming for Smart Cities will require much
better programming languages than currently used for
distributed programming where, for instance, there is little support for the verification of the current behavior of
programs. This is confirmed by the tremendous amount
of middleware that exists to cover the programs' shortcomings. Yet these middleware programs do not in tegrate well with the host language of the code [19].
Powerful languages that allow complex distributed code
to be written in a "clean" way (such as AmbientTalk [20])
have not yet made it to the mainstream. What makes a
"good" language for this job? An important yardstick for
measuring the quality of a language can be found in
Brooks' paper on complexity in software engineering
[21]. Today's programming languages put far too much
emphasis on the accidental complexity of a distributed
system. Languages that allow a distributed programmer
to only focus on the essential complexity are still being
researched at this time.
For inspectability, algorithms are often referred to as
"black boxes" in that it is not apparent to the casual
observer exactly how the algorithm works. Tomorrow's
intelligent environments and algocracies should be
based on "grey box" systems that citizens can read and
tweak along various "levels" of participation. Just like
there is a distinction between a constitution and normal
laws, distinctions have to be made between various levels of code so that some code can be easily tweaked by
direct democratic processes (à la WikiPedia), whereas
other code is proverbially carved in stone. In other
words, the code that runs the algocracy needs to be
exposed in a "grey box" fashion, where different shades
of grey will probably be needed. For instance, in the
area of pervasive computing and the Internet of Things,
approaches to user-level programming for configuration
of smart environments, based on simple and understandable "if this then that" rules (see e.g., www.ifttt.com),
JUNE 2018

∕

and hiding more mundane programming details, go in
that direction of a "grey box" approach.
In a broader perspective, societal engagement would
also include building "institutions and tools that put the
society in-the-loop of algorithmic systems, and allows us
to program, debug, and monitor the algorithmic social
contract between humans and governance algorithms"
[22]. The need for transparency, accountability, and
explainability for the increasingly prevalent AI "blackboxes" has been noted in [23], where a layered model
involving technical, ethical, legal, and social aspects
needs to be taken into account.
In parallel with the development of an understandable
"gray box" approach to programming, we need to solve
one of the main factors hampering efforts for a healthy
algocracy. This is the need for citizens to have at least a
basic literacy in computing, and - if they are not able to
program - they should be capable of judging the actions
and the quality of the programs that govern them (at least
when exposed in their "gray box" form). Unfortunately,
computer science as a basic scientific field is absent in the
high school systems of most countries. A notable exception is the U.K. where "Computing" is part of the high
school curriculum since 2015. Even for people who will
never program in their entire lives, a good basic understanding of what is programming, and what is an algorithm, is necessary for being a citizen in the algocracy!

Humans in the Loop and the Wisdom of Many
An algorithm is weightless and only worth the weight
people put in it, so that some degree of safety from the
potential dangers of algorithmic governance can come
from the "wisdom of the crowd." There is a need for
users to be able to provide feedback to the system in a
forum, or through a mechanism for collectively commenting on the algorithm's performance, so that problems can be identified and signaled. The same pervasive
sensing technologies that feed the algorithms with data
can be exploited by users to monitor the environment
and signal (and share information about) problems
caused by existing Smart City algorithms. For instance,
the fact that home heating systems are biased so that
they do not meet user needs and only act towards some
municipality goals, can be discovered by a multitude of
users (or by their algorithmic guardians) by accessing
sensor and actuator data. Eventually the users can then
make the facts emerge to global awareness. Indeed, it
should be a general goal for governments (at all levels,
from national to municipal), whenever they start relying
on algorithms to control cities and make decisions on
our behalf, to involve human citizens in the loop at the
highest levels of the participation ladder [3], i.e., as partners, in delegating authority, and as co-managers of the
system or algorithm.

IEEE Technology and Society Magazine

http://www.ifttt.com

Table of Contents for the Digital Edition of IEEE Technology and Society Magazine - June 2018