Freeing machines from Cartesian chains

Abstract

The impact of technology on thinking about behaviour has shifted from mechanistic descriptions towards the computational stance of cognitive science and classical Artificial Intelligence. All these approaches share an output-oriented black-box rationalism, which is also the foundation of neo-Darwinistic accounts of behaviour. To gauge the limitations of this type of explanations and of ethological methods in particular, I analysed the behaviour of simple robots as if they were living creatures. This revealed interesting patterns but did not take the lid of the black box. The self-organized cooperative behaviour of the robots could only be understood if feedback from environmental changes was considered. Furthermore, the robots were not designed by "engineering from scratch" or a "problem-solving approach", but instead by an almost task-free attitude without preconceptions like "imperfect design" and "behavioural errors". This questions the use of a priori stated "costs" and "benefits", and thus is at odds with the starting points of normative and rationalistic theorizing.