IEEE Technology and Society Magazine - Winter 2014 - 25

more likely to overlook the responsibilities of human
actors other than operators and users - those human
actors that create the conditions for the operation of
the technology. This changes when the technical components that constitute artificial agents are understood
in terms of their role in a broader network of human
and non-human components. Failure to recognize this
may lead to a failure to see that part of the development of a technology is designing the way humans
will operate in or interact with the system and interact
with artifactual components. In the case of artificial
agents, this means, among others, those who decide
when and where to deploy the agents, those who control the artifact in operation, and those who monitor
and maintain the agents.
Recommendation 2:
Responsibility issues should be addressed when artificial
agent technologies are in the early stages of development.
Responsibility issues often seem to be addressed
(if at all) only when a technology is introduced in realworld settings. Google's on-the-street experiments
with their autonomous car quickly generated a debate
about the applicability of existing legal frameworks
and liability laws in California and Nevada [23]. The
technology was already developed to a significant
extent, and decisions about how the car should behave
had been made along the way. Before these experiments, discussions about responsibility involving
autonomous cars were relatively few and did little to
influence the design of the car.
This is not uncommon for new technologies. In the
early stages of development the focus tends to be on
technical feasibility. UAVs, for instance, have been
in development for over fifty years. Yet, the discussions about responsibility have only recently gained
momentum especially because of the use of drones
in military operations. To be sure, issues of responsibility with regard to autonomous robots have been
raised early on. The handful of roboticists who work
on designing robots to be ethical [2], [3] are an illustration of this. Yet, it was only after these technologies were introduced on a larger scale that the debate
really took off.
The problem of addressing ethical issues after a
technology is introduced and put into use is that the
design decisions that have already been made may
preclude or constrain the possibilities for particular
kinds of responsibility assignments. It is also possible that responsibilities for various aspects of a new
system may be assumed or given cursory attention
during one phase or another, e.g., testing, rather than
being assigned explicitly and carefully with an eye to
the users or the context of use. This makes the issues
harder to address and may do damage to acceptance of
the new technology.
IEEE TECHNOLOGY AND SOCIETY MAGAZINE

WINTER 2014

An important trend in recent scholarship on ethics and technology is to identify and analyze the
ethical issues raised by new technologies while these
technologies are still in the early stages of development [6], [28], [32]. Some refer to this endeavor as
anticipatory ethics [8], [15] or anticipatory technology ethics [6], [7]. Value sensitive design (VSD) and
responsible innovation [29], [30] are examples of
such approaches. Such anticipatory ethics approaches
generally build on a sociotechnical perspective. They
analyze the sociotechnical processes that underlie the
development of a new technology in order to intentionally address ethical issues before the technology
has stabilized.
An example of how concerns about responsibility can be part of early development stages is when
designers build in monitoring systems to record information on the operation of a vehicle with an eye to
determining the chain of events in cases of accidents.
Here the monitoring system may be used not just to
trace back what happened but also to decide who is
accountable for an accident or failure. Another example is user interface design that can be designed to
facilitate or constrain responsible operation of a UAV
or other agent (see, for example, [11]).
In emphasizing the value of addressing responsibility
issues early on, Recommendation 2 helps to ensure that
responsibility issues do not fall through the cracks, and
decreases the possibility that artificial agents that defy
human control will be implemented. Intentionally bringing ethical concepts and concerns into the early stages
of technological development has the potential more
broadly to lead to better technologies. We are more likely
to get technologies that are aligned with moral notions
and less likely to get technologies that violate moral sensibilities and principles. We are more likely to eschew
lines of research or designs that ultimately would have
been rejected by consumers, users, or the public.
Recommendation 3:
Claims about the autonomy of artificial agents should
be judiciously and explicitly specified.
In the discourse on artificial agents, a great deal of
attention is given to the autonomy of artificial agents.
A number of scholars have devoted attention to futuristic visions, similar to those found in popular media,
in which artificial agents will have or even exceed the
capacities of human beings (see, for example, [26]).
Such visions suggest that future agents will be increasingly autonomous, operating entirely independent of
direct human control. This prospect underlies many of
the concerns about whether or not humans can still be
held responsible. Moreover, the idea of increasingly
autonomous artificial agents leads to speculation and
discussions about the moral rights and moral standing
of these agents [13], [34].
|

Table of Contents for the Digital Edition of IEEE Technology and Society Magazine - Winter 2014