On the non-phenomenological use of the word ‘behaviour’ in adaptive behaviour and robotics research

Since Simon[1] and Brooks[2], robotics researchers have understood that the behaviour of a robot system is a function of the structure and dynamics of its physical construction, the environment in which it finds itself, and its internal control system. Simultaneous with the growth in popularity of this view there has been an increasing use of the word ‘behaviour’ to refer to non-phenomenological objects. It is now common to refer to a subroutine in a robot controller as a ‘behaviour’, the word designating a set of instructions rather than the outcome of those instructions as they are executed. We believe that behaviour-as-subroutine is an unhelpful word, because it implies that the subroutine is the entire cause of the robot’s visible, external actions. It confuses the subroutine with the observable behaviour of the system. This is in conflict with the most important ideas behind adaptive systems, behaviour-based robotics and ethology, in which behaviour is considered the result of coupled interaction between active controller components, physical construction and environment. This paper presents some examples and a simple experiment to show that the two meanings of ‘behaviour’ are distinct and incompatible. We suggest that the non-phenomenological use of the word ‘behaviour’ is an unhelpful mistake and assert the usefulness of the classical meaning.