A computational model of avoidance behavior

Learned avoidance behavior is critical to animal survival but has proven difficult for animal learning theorists to model. The authors propose a computational model of the highest layer of an hierarchical control system responsible for learning and behavior. The proposed layer consists of a network of associative control processes (ACPs) that employ the drive-reinforcement learning mechanism. A network of ACPs has been shown to successfully predict both classically and instrumentally conditioned behavior. The output of the authors' model is a goal-directed, whole-animal behavioral command that is used in lower layers of the control system to guide motor responses. In computer simulations, the authors' model developed, through trial and error learning, the active responses necessary to avoid shock in a one-way shuttlebox experiment. The model extends Mowrer's (1960) revised two-factor theory of learning.