IEEE Technology and Society Magazine - September 2017 - 76

formal computational theories of relations and value. On
the other hand, social organization has been studied in
some depth by systems researchers, where progress
already made can contribute to the formalization of this
aspect of socially-sensitive systems design.
Several key challenges arise from our analysis of
insufficiencies in existing state-of-the-art approaches.

Challenge 1: Social Awareness and
Intentional Social Action
In computing and biological systems alike, additional
"devices" can help increase the scope of the system's
sensors (helping us to perceive more) and effectors
(helping us to do more). Many such contributions can be
unintentional on the part of the device or agent. Hence
a device like a rear-view mirror, a telescope, or a microphone can add to the range of our sensors. There is
adaptive behavior on our part in using these devices but
there is no intelligence in the basic form of these devices. A bird's sudden trill, stranger's glance upwards, or a
cat's ear movements may all convey information to us
that we can adaptively use, but again without any intentional communication of such to us. On the effector
side, bulldozers and bullhorns, amplifiers, spears, and
guns are just a small number of the inventions we have
developed to enhance the scope of our actions.

One of the key issues to address is
how to represent and reason with
social values, mathematically and
computationally, without reducing
them to base quantities.

In previous work [4], [5], we have emphasized the
importance of reflective processes and self-awareness to
the adaptive behavior of complex systems. And we have
noted that the information generated by a self-aware system can in turn be communicated to other agents, greatly facilitating cooperation. What has not been sufficiently
explored however is the role that others play in a system's ability to correctly formulate how its behaviors and
capabilities fit (or not) into a given operational situation.
Others can literally see things that the system itself cannot; they have a different scope, different viewpoints,
and different experiences - all of which can be critical
to correctly understanding how a given system is contributing or not to some goals within a specific situation.

As discussed in our introduction, uncertainty can be
reduced by using a collective of agents to enhance the
scope of sensing and actions, adding a diversity of experiences, perspectives, and expertise. The challenge is to
design these collectives, and their diversity more specifically, such that the system (in terms of its parts) supports
itself in a mutually beneficial way, that adds knowledge,
capability, or reasoning in support of the system's goals.
When we consider the design of systems in this way,
there are several main questions: Where does it help to
have multiple eyes and hands and how does one coordinate them? Where does it help to have multiple agents
watching out for the success of both themselves and others? Where does it help to approach complex problems
with multiple viewpoints and experiences - and how
does one communicate and integrate those viewpoints?

Challenge 2: Capturing Social Value
Here, we identify a major issue with existing, purely utility-based approaches for capturing the notion of social
(human) value in socio-technical systems. Fundamentally, such systems should be capable of incorporating and
managing the notions of value we, as humans, associate
with social aspects of the world. In many existing
approaches to designing socio- technical systems, for
example in much of the multi-agent systems literature,
an approach is taken whereby direct (often numerical, or
at least ordinal) comparability of alternatives may be
assumed, based on a universal commodification of such
values. We argue that this approach is, in general, insufficient to capture both a diverse range of and nuance
associated with human social values such as obligation,
empathy, peace, and justice. Yet these are the things that
often truly matter the most to humans and the human
societies that form part of socio-technical systems.
One of the key issues to address is how to represent
and reason with social values, mathematically and computationally, without reducing them to base quantities.
For example, it has been seen how, when assigning a
financial value to a moral, one dilutes the moral value of
its force. The nursery that introduced fines for late pickups found that collection time just got later; and when
they rescinded the fines, pick-ups got later still. They had
replaced moral incentives (i.e., good parenting and consideration to proxies) with a coarse financial one, and
when the financial incentive was removed as well, the
moral one did not immediately re-assert itself [26]. Much
the same commodification of human values has been
seen in e-commerce and social media, with concepts like
loyalty, trust, and friend all being diluted of their social
quotient and social potential by representation as numeric terms. There are, however, logical frameworks in
which it is possible to represent and reason with values
formally without falling into this trap [35].

IEEE TECHNOLOGY AND SOCIETY MAGAZINE

∕

SEPTEMBER 2017

Table of Contents for the Digital Edition of IEEE Technology and Society Magazine - September 2017