Michael Decker
Vortrag auf dem internationalen Workshop "Social Aspects of Cooperating Objects Technologies". Technische Universität Berlin, 02.11.2006
Abstract
Robots are one of the rare technical systems that have already been comprehensively described and discussed in terms of their possible construction and effects before actually being built. However, considering the development in the last fifteen years, namely that much more microprocessors are used as so-called microcontrollers for other purposes than for computers (e.g. in cars, aeroplanes, houses, machine controls, satellites, mobile phones, washing machines, gaming machines, cameras, etc.), one can speak of an automation of the human environment. This development involves increased efficiency, "miniaturisation" and wide availability in other areas: motors, drives, batteries, materials, sensors, etc. General robotics has meanwhile reached a stage where dealing with robots does not necessarily lead to an odyssey through diverse disciplines such as mechanical engineering, electrical engineering, control engineering, software engineering, and algorithmics with countless involuntary intermediate stops. It is rather possible to assume a (minimum) availability and stability of hard- and software components, and the systems can be systematically constructed in an engineering manner.
These developments in robotics research allow the use of robots in application areas that have not yet been technicised. Here, robots will act in contexts in which, so far, only humans have acted. In these action contexts, humans will be replaced by robots. Reference to concrete contexts is of special importance, because presently, the general replacement of humans by robots is not intended. The issue of replaceability of humans in specific contexts of action was formulated very generally and formed the focus of an interdisciplinary technology assessment.
First, there is the issue of technical replaceability. A robot will only be used if it is technically capable of performing actions that are required to fulfil a specific task. A robot will always be technically evaluated in a means-end relation, where it competes with possibly other existing means that can be used for a specific end. The technical criteria for the use of robots are derived from this means-end relation.
Such means-end considerations quickly reach a point where it is necessary to assess the benefits of robots (economic replaceability). This implies not only strict cost-benefit considerations in economic terms where aspects of a broader definition of benefits are often not considered. This becomes obvious when evaluating so-called "service robots", because service is always associated with aspects such as friendliness, helpfulness, attentiveness, politeness. What arguments can be put forward for the right of the demand-side to these additional service aspects?
The position of the demand-side - the customers - has to be discussed also from another perspective. For example, it is necessary to examine whether the introduction of robots as actors leads to changes from the legal point of view (legal replaceability). Liability issues are of course of great importance, too. Who is liable for damage caused by a robot? On the other hand, there are questions of consumer protection. Because in future scenarios of robotics also laypersons will encounter robots. Do robots require special equipment for these "unexpected" encounters? And, in turn, is it necessary to prepare people for the possibility of such encounters? Are there any additional aspects to be considered for "learning" robots?
Finally, the question must be addressed as to whether there are action contexts in which the integration of robots should be excluded by modem societies. This might apply for example to the care of old and ill people, the education of children, or similar action contexts. The question of the areas in which autonomous robots should replace humans is strongly determined by the issue of the ethical replaceability. This ethical reflection in the mentioned means-end relation is particularly aimed at the level of ends, i.e. questioning whether the ends of using autonomous robots are ethically justifiable.
This contribution refers to the interdisciplinary TA-project "Robotics. Options for Human Replaceability" (Robotik. Optionen der Ersetzbarkeit des Menschen) in which several recommendations to act have been developed referring to robotic applications in general. Here, I would like to focus on liability aspects of leaning robots. One recommendation proposes a technical solution to enable the robot user to take over responsibility for the robot action: The robot indicates what it has identified as "worth to learn" and the robot user needs to explicitly accept this proposal, if the robot should learn that task. However, such a transparent learning algorithm causes problems which root in one of the basic challenges of artificial intelligence: Context awareness.