IEEE Technology and Society Magazine - June 2020 - 55

institution if the collective experience was considered
incongruent with the operational policies.
The specification and implementation of an algorithm for this form of interactional justice in a multiagent system used ideas from distributive justice [13]
for the personal self-assessment; and then used opinion
formation over a social network [2] to aggregate subjective (diverse) self-assessments of fairness into a collective (consensual) assessment, as an indicator of the
institutional quality [6]. Each agent had certain values,
specifically "fairness", and a set of metrics that it used
to evaluate whether or not the outcome of institutional
deliberation was "fair" (the specific application was
resource allocation, so the metrics included legitimate
claims, utility, and satisfaction). From this and interaction with its neighbors, each agent formed an initial
opinion on fairness. These opinions then percolated
through the network, and agents updated their opinions
accordingly. Their final opinions could be used to motivate institutional change, if the collective opinion was
one of specifically unfairness, or more generally, an
incongruence between institutional rules and experience. Linking back to relevant expertise aggregation,
one option would be a majority vote on a policy to
address the problem, but another could be some form
of "civil" disobedience as a mechanism of bringing
about positive change [14] (and see below).

The Zone of Dignity
In a system of systems, the relationship between systems (i.e., the institutional architecture) can (in extremis)
be either hierarchical (single centralized decisionmaker) or polycentric (multiple autonomous centers of
decision-making). In a hierarchical system, loosely
speaking, information flows "up" and policies flow
"down." In an ideal world, perfect information drives
appropriate policy formation, i.e., evidence-based policy-making. In a sub-ideal world, information is abstracted up the hierarchy, and policies are reinterpreted down
the hierarchy.1 To counter this, some forms of corporate governance have introduced the concept of guardrails. The idea of guardrails is that in a hierarchy, the
"upper" (management) layer specifies that within specific boundaries, the "lower" (implementation) layer

can make any decision, or take any action, they want, in
order to complete a given task. Moreover, the guardrails
are not necessarily fixed: if the "lower" layer needs to go
outside the guardrails, it seeks permission from the
"upper" layer.
The idea of a guardrail can be seen within Ostrom's
Institutional Design Principle 7, which states that, in a
system of systems, the minimal recognition of the right
to self-organize should be observed [15]. In fact, many of
the failures to sustain a common-pool resource, reported in [15], could be traced to a failure of this principle.
The idea of a guardrail is also present in Ober's Demopolis [7], in a discussion of the continuum of distributive
justice from full equality to complete liberty. The
"poles" of this continuum are marked by excessive
paternalistic intervention to enforce perfectly equal outcomes at the egalitarian end, to a total absence of
state intervention to ensure that all citizens have
access to adequate resources at the libertarian end.
However, dignitarian considerations (such as non-infantilization by the removal of choice on the one hand,
deprivation of adequate resource to contribute fully as
a citizen on the other) set limits on how far the libertarian and egalitarian tendencies can push liberty and
equality. The avoidance of indignity for a democratic
regime2 defines the zone of dignity (ZoD), which determines the acceptable range of policy options for distributive justice, i.e., not too coercive, not too heartless.3
The ZoD therefore represents a kind of preference
meta-consensus (or guardrails) on the acceptable range
of (institutional) policy options.
To specify and implement a deliberative process for
algorithmic reflexive governance in a system of system,
we propose to adopt the idea of guardrails in an adapted
and extended ZoD. The extended ZoD has three components: a coordinate plane, metrics to locate an object in
that coordinate plane, and meta-rules to control the trajectory of an object (i.e., a set of rules) as it moves within
the plane. The proposed coordinate plane extends the
ZoD from a linear one to a two-dimensional plane, by
also including an axis of "resource extraction" or taxation, that the "upper" institution imposes on the "lower"
(cf., [16]).4 As shown in Figure 3, this characterizes four
2

1

One example: in the U.K., a policy of restorative justice, a process that
tries to resolve disputes to the mutual satisfaction of all concerned parties
through negotiation, was introduced by a regional U.K. police force [19].
The policy was interpreted differently by different groups: senior management understood the theoretical concepts, key values, and nuanced
differences between "instant restorative disposals" and "restorative conferencing"; middle management were focused on performance; and uniformed officers "on the beat" applied the policy as frontline practitioners.
The latter group found themselves conflicted between pressures to resolve
disputes without involvement of the criminal justice system, but also having to increase the number of "detections" (the number of cases resolved
with a ticket, charge, caution, etc.).

JUNE 2020

∕

Ober [7] also states that civic dignity is undermined when the citizens are
deceived into voting for policy options that are not in their best interests;
for example, a certain referendum in a North-Western country of Europe.

3

This raises the question of why there is not a maximal recognition of the
right to self-organize, as this implies that in a system of systems, unlimited
rights to self-organize do not affect the sustainability of a local commonpool resource. Maybe this is true, but it does not necessarily augur well
for the political regimes "at the edge," i.e., the leaf node of a hierarchy or
regions on the periphery of a state, e.g., some 1970s U.K. academic departments, organized crime, warlordism, feudalism, etc. See for example [20].

4

The notion of "resource extraction" is intentionally general and covers the
spectrum between "stationary bandits" [17] and meso-level water temples
coordinating rice planting [18].

IEEE TECHNOLOGY AND SOCIETY MAGAZINE

55
IEEE Technology and Society Magazine - June 2020

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