Indeed, one of the main advantages of the
proposed system over similar models, such as [9, 20] is
its ability to generalize. Rules need not all be given a
priori. Based on a set of training examples, the system
inductively generalizes and synthesizes rules that can
then be used in reasoning. Thus, from a psychological
standpoint, the model retains the self-adaptive nature
inherent in human learning, while allowing the
encoding of a priori knowledge. Experience can then
be combined with theory to yield a system capable of
confirming, refuting, and modifying its "innate"
knowledge based on empirical observations