- Joachim Hoffmann: ABC: A Psychological Theory of Anticipative Behavioral Control
When the information processing approach arised in the sixties of the last century, cognition was defined as "... referring to all the processes by which the sensory input is transformed, reduced, elaborated, stored, recovered, and used" (Neisser, 1967, p.4). Accordingly, cognitive processes have been considered as being stimulus driven. I will argue that this approach is basically misleading as it ignores the determining role of intentions: cognitive processes do not serve to process given stimulation but to support the production of desired or otherwise anticipated stimulation.
The ABC theory (Anticipative Behavioral Control) exemplifies this tenet with respect to the acquisition of behavioral competence. It is assumed that evolution brought about elementary learning mechanisms by which efferent activation patterns (motor commands) became automatically associated to co-occurring reafferences (sensory effects) in such a way that reactivations of certain afferent patterns gain the power to address the efferences they formerly were the effects of (the ideo-motor principle). Furthermore, such ideo-motor associations become conditionalized if the contingency of the motor-sensory connections depends on current circumstances, that is, on the current state of the acting organism.
The talk will present theoretical considerations as well as experimental evidence in support of the ABC theory, in particular referring to animal and human associative learning and to the impact of behavioral effects on the selection, initiation, and execution of simple voluntary acts. Finally, speculations about how the motor output might be controlled by a cascade of increasingly specific sensory anticipations are discussed.
- Christian Balkenius: Anticipation in Purposive Behavior
While it is tempting to think of motor control as a mapping from sensors to effectors, this ignores both what the sensory-motor mappings tries to accomplish and what the sensors are trying to code. We do not see photons even though that is what our photoreceptors detect. Nor do we normally strive to contract muscles even though that is the function of the signals leaving our brain. Instead, we perceive a world around us and perform goal-directed actions on objects. Since the world is not stationary, such actions depend critically on an ability to anticipate the relevant states of the world as well as the consequences of our actions. To this end, sensory and motor information must be processed in a number of interacting time frames ranging from very short-term predictions that compensate for processing delays in the sensory system to an appreciation for the future consequences of actions. These ideas have been implemented in robots that show highly adaptive purposive behavior and and fast learning. The approach differs from both the classical and the reactive approach to robotics in that anticipatory models are short- term and local and result in shorter response times than is possible in a purely reactive system.
|
