-
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.
|