IEEE Technology and Society Magazine - March 2020 - 56

child through algorithms applied to sensor input from
what is described as a smart space. This supports the
application of computational models to assess the
behavior of the child and infer his psychological disposition - in other word, to exercise total autonomy over
the child through feedback loops.
ABA is described as enabling an approach to changing socially useful behaviors that employs scientifically
established principles of learning to bring about these
changes [9]. ABA works by forcing the child to act in
superficially socially acceptable ways. Such therapy
may require up to 40 hours contact per week, so there
are obvious efficiency advantages in deploying a robot
in such a radical behaviorism that treats humans, rats,
and machines as having equal value.
However, ABA is based on a flawed view of autism.
For example, operant conditioning may remove stimming, but fails to understand why the child enters into
behavior such as arm flapping [17]. Transparency
requires not only that we are honest about explaining
the nature and limits of the robot, but also the limits of
the programming that has been done, and the philosophy behind the programming. It also requires that we
are honest about the nature of and the purpose of the
work we are having the robot do in a social setting.
In the case of the DREAM project, not only the limits
of NAO robots should be defined, but also the limits
and controversies concerning ABA. And yet no mention
is made of issues with ABA. ABA seeks to enforce social
conformity by making the autistic child act normally on
the outside. It makes limited attempts to understand, to
find meeting points, to create an environment where
the child feels safe, accepted, and competent. This
deployment of NAO robots to implement ABA may be
seen as oppressive and subordinating. ABA fits the child
into social expectations. The child is subjected to suffocating intense robot therapy that does not address the
underlying emotional needs of the autistic child.

Identity
Robots have a limited set of behaviors. The limited
set of stimuli will require a fixed set of responses. A
robot such as NAO used in ABA training seeks conformity from the child. The NAO is programmed with
standard stimuli and responses. The aim is to get the
child to behave in a standard socially acceptable way
through hours of repetitive social conditioning. In
doing so it suppresses individuality. Distinctive or
unique skills that the child may have are subordinated to social standardization. In other words, it affects
the child's identity.
We all have different skills and these are part of our
identity. Our specific way of thinking is part of our identity. Autistic children have a different way of thinking.

While they may be poor at social relations, they may be
rich in other skills. Autism should be seen as a difference in informational processing rather than a behavioral disorder [13]. While the resulting social problems
are not minor, ways of dealing with them can be developed that involve a meeting between the autistic and
the neurotypical. Children can be taught to compensate in order to interact with society and flourish, but
that should not be at a cost of fracturing their identity
as a person.
The use of the NAO robot by the DREAM project
involves not only programming the robot with standard
behavior, but demanding standard behavior. But there is
a further problem in that the ABA training enforced by
the robot is only rule based. It is static learning. It does
not enable the child to understand and adapt. It effectively papers over the child's identity, creating a veneer
of conformity and homogeneity, making the autistic
child look the same as everyone else when they're not.
It is like dressing the child in a behavioral school uniform. This may fit certain situations but inhibits the
child's character and personality. Of course, the robot
must do this because it must reduce the behavioral variety of the child to match the limits of the robot. The
robot cannot adjust its behavior, based on rules, to
every situation. It has a limited behavioral repertoire.
The ethical consideration of social robots must
examine the connection between robot technology
and human identity. The presence of the robot may
alter the child's identity and the question is: what is the
ethical effect of that? For example (and this connects
identity with autonomy in ACTIVE ethics), an autistic
child will have certain things he can do for himself that
involve control and responsibility. Indeed controlling
and managing the sensory environment is very important for the child.
In a care home, you should never do anything for a
client that they can do for themselves, even if they struggle to do it. The struggle and achievement are part of
the client's identity. Responsibility should take precedence over pity. Thus the human must always assert
himself over the robot and not become a passive recipient or even victim of the robot. If identity is not conserved and promoted as a result of robot interaction,
then an ethical issue has arisen that needs addressing.

Value
In ACTIVE ethics, value concerns the value we put on
the human. If we value the individual highly we will protect their data, we will design the robot with a humancentered focus. We will also encourage the human to
take as much control and responsibility as possible. We
will be less concerned about automation than communication and our design effort will focus on the interface

IEEE TECHNOLOGY AND SOCIETY MAGAZINE

∕

MARCH 2020

IEEE Technology and Society Magazine - March 2020

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