ABSTRACT
Soar is a computational architecture for the realization of intelligent, behaving agents. It is also a theory of the human cognitive system. As is characteristic of both such entities, it exhibits self-modification. Thus, over time, Soar's behavior changes as a function of its knowledge state and the interaction of the tasks it undertakes, the processing that arises, and the learning that is a consequence. Because of these characteristics, the Soar project is populated by a group of scientists who are unusual in that they continually ask themselves questions such as “Where did this particular competence come from?” or “How did this piece of knowledge come to be constructed?” Those same features mean that Soar can act as a vehicle for the direct study of mechanisms of human cognitive development, an enterprise that involves the investigation of essentially identical questions.
