IEEE Technology and Society Magazine - June 2020 - 58

reinforce a surveillance apparatus used by states or corporations for control and profit in a way that damages
both the social and natural worlds.6
Community energy systems (CES), which could be
considered as an exemplar of socio-techno-ecological
systems, illustrate the first challenge. A community
energy system takes a primarily commons-oriented
approach to energy generation, distribution, consumption, and storage, rather than a market-oriented
approach (cf., [27]). The "socio" component of the system comprises the human stakeholders; the "ecological" component is concerned with ensuring that the
actions of those stakeholders in accomplishing their systemic goals do not adversely affect their physical surroundings or other living organisms. In principle,
SmartMeters should be the sensors that are inserted
into the ecological system that are "reading" the state of
the environment, and should also be the software
agents that support the human stakeholders in managing the balance between getting what they want out of
the system without doing any environmental damage in
the process. Artificial Intelligence, in the form of algorithmic deliberation and reasoning should be the technological "glue" that links the socio- and the eco-logical.
However, four layers of interaction between human
stakeholders and artificial intelligence have been identified in the management of energy as a common-pool
resource (as, for example, in a CES). These are:
delegated - humans out of the loop: some operations are controlled by artificial intelligence, which
works without (or with limited) user awareness; for
example, in CES, micro-grid load-balancing and voltage control can be implemented by regulating refrigerator compressor cycles;
■■ programmable - humans on the loop: daily operational requirements are specified by members of the
CES, and artificial intelligence resolves the constraints (and disputes); for example, with programmable appliances such as dishwashers, the CES
members specifies constraints (completion time,
cost), and delegates to the appliances the task of
negotiating a collectively agreeable schedule;
■■ interactive - humans in the loop: artificial intelligence detects problematic situations and indicates
active intervention is required by humans; for example, a SmartMeter can initiate collective action for
overload prevention, although any such intervention
has to be sensitive to social situations (hence supportive not determiner); and
■■

6

Similarly, there is some discussion whether digital solutions to controlling the coronavirus crisis are potentially leading to excessive tracking,
monitoring, and control [26]; this could be called "disaster surveillance"
(cf., [27]).

58

■■

attentive - humans are the loop: each stakeholder
may occupy multiple different roles besides producer
and consumer in a CES, for example, as an investor or
a "citizen," they will be called upon to consider investment decisions, engage in interactions with third parties, help resolve internal disputes, and so on.

Therefore, the deeper challenge for algorithmic
reflexive governance in socio-techno-ecological systems is to manage the four binaries across all four layers. However, herein lies the roots of the second
challenge. Of course, there are the practical aspects of
implementing such a system; but there is a theoretical
argument about the potential for success. Going back
to Ashby's Design for a Brain [4], it is argued that for
stability (or in our case, sustainability) of a system, its
regulator has to be more complex (have more states)
than the system being regulated. Therefore, if by conjoining human and computational intelligence, we create a system that is more complex, and also more
complex than the ecological system, it should be possible to regulate it; if, on the other hand, the joining of
human and computational intelligence results in a diminution of human intelligence or competence [29], [30],
then we are perhaps no better off, and possibly worse,
if our humanity is diminished as well as depleting and
despoiling the natural world.
However, there are at least two open questions here.
Firstly, there is the claim that, with certain assumption
any regulator must also model what it regulates [31]. It
could be argued though that "harmonious" regulation of
a socio-techno-ecological system does not require the
capacity to model the entire system fully, as opposed to
"optimal" control (with "optimality" defined with respect
to the regulator and not to other components, or the
system as a whole), which does. Secondly, when the regulatory system is itself embedded in the system being
regulated, is it the meta-deliberation that we have
argued for here, that offers the necessary complexity?
But for this, data and modeling are surely required,
unless we want to pursue "the perverse desire - to
remain ignorant" [32]. The exploration of these questions, and more, is an enterprise for future work.

Resolving Tensions in Reflexive Governance
In summary, we have proposed an algorithmic basis for
implementing deliberative processes which can help to
resolve the four tensions in reflexive governance. We
have argued that algorithmic reflexive governance can
monitor the impact of institutions on ecosystems, and
vice versa, and is a critical foundation for reshaping core
values and reconfiguring current practices, especially
when incongruence between those values and practices
is exposed, especially when those values are concerned

IEEE TECHNOLOGY AND SOCIETY MAGAZINE

∕

JUNE 2020
IEEE Technology and Society Magazine - June 2020

Table of Contents for the Digital Edition of IEEE Technology and Society Magazine - June 2020
