IEEE Technology and Society Magazine - Winter 2014 - 27

job entails and what is expected of them. Their tasks
and duties may be explicitly stated in training manuals, procedures, protocols, codes of ethics, and regulations, but engineers may also learn about these as they
become acquainted over time with the informal norms
and rules within organizations. The engineer's acceptance of a job is taken as a sign that he or she accepts
the assignment of tasks and duties that go with the job.
When an engineer fails to live up to specified responsibilities he or she can be reprimanded by colleagues
or superiors through informal conversations or through
more formal processes, e.g., annual evaluations. In
high profile cases such as occurred with the Challenger disaster and the Y2K problem, the public may
demand an explanation, an accounting from the profession. These activities all contribute to shared norms of
responsibility; they constitute responsibility practices.
Responsibility practices involve formal and legal
frameworks as well as informal social conventions
and are usually a mixture of both. A soldier is responsible for following orders given by a superior officer;
this responsibility is reinforced by military culture
and daily practice but it is also formally expressed in
military codes, codes that have been refined in military court decisions. In other contexts, responsibility
practices may be contested. For example, informally
if a search engine bot delivers pornographic material
in response to a search word entry, some would blame
the searcher, others might blame the search engine
designers or the website owners.
The notion of responsibility practices places the
focus on the shared understanding of what is expected,
what is owed, and what are likely consequences of failure within a sociotechnical network. Using this notion
involves examining what constitutes responsibility in a
particular network and why practices have been constituted as such: what are blame- or praiseworthy actions;
when is one required to account for one's behavior;
what would cause feelings of guilt and shame, regret,
and a sense of moral reckoning; and so on. Using the
notion of responsibility practices draws attention to the
formal and informal mechanisms and strategies, such
as legal procedures, organizational rules, and social
norms that promulgate and enforce responsibility.
Once established, responsibility practices are continuously challenged and renegotiated. New technologies, for instance, tend to change the way people do
things, how they relate to each other and, thus, how
responsibilities are distributed and acted on. For
example, the introduction of UAVs has led to a range
of new procedures and protocols in military operations [22]. Pilots had to learn new skills, commanders
had to develop new strategies, and decision-making
processes have undergone significant changes. This
has meant in turn that new practices have had to be
developed and negotiated for assigning and ascribing
IEEE TECHNOLOGY AND SOCIETY MAGAZINE

WINTER 2014

responsibility, not only within military organizations
but outside as well. Increasingly prominent discussions in the mainstream media about the ethics of
drone warfare show that the U.S. is still in the process
of developing new norms, policies, and laws that govern the use of these unmanned systems.
In order to address responsibility concerns involving artificial agents, a careful analysis of responsibility practices and how they shape and are shaped by the
design of a particular technology is required. When
responsibility practices are considered in the early
stages of development (as indicated with Recommendation 2), then policy and practices that assign and
specify who is responsible for what can be put in place
and enhanced as the technology reaches completion.
It is important to take into consideration how the relationships between the relevant actors take shape, how
ideas about obligations, duties and accountability
evolve, and what mechanisms and strategies are put
in place to enforce these responsibilities on multiple
levels. It also means that we have to create the appropriate conditions for people to be responsible: institutional mechanisms, law, and policy, etc.

Issues of Responsibility in
Artificial Agent Technology
The overarching principle and four recommendations
listed above work together to constitute a framework
for thinking about artificial agent technology (and
other emerging technologies) that enables consideration of issues of responsibility. To effectively address
the responsibility issues in artificial agent technology, it is important to address the issues early on.
The issues can only be fully seen (early on or later
on) when the technology is viewed as a sociotechnical
system, when claims about autonomy are used judiciously and kept in context, and when responsibility
is understood to be a matter of social practices rather
than something in an individual.
The recommendations are broad both in the sense
that they apply to many different actors and also in
the sense that they imply different sorts of behavior
for different actors. For example, they encourage
researchers, engineers, and developers to look beyond
technical specifications and imagine what kind of
interactions between a technology and its users can
occur; they remind journalists and policy makers to
be critical of abstract future visions and look more
closely at the possibilities and limitations of the technology in particular contexts. The various actors thus
need to reflect on the technical, social, and ethical
aspects of artificial agent technologies.
In facilitating consideration of issues of responsibility, these recommendations are targeted to improve
technological development. Artificial agents of the
future promise to perform some tasks better than
|

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