ABSTRACT
A recent issue of the journal Cognitive Science (1993, vol. 17, no. 1) centered around a debate between two “camps” within the field, the “situated action” (or SA) camp and the “traditional,” symbol processing camp. Though the debate in that journal suggests that, at some levels, symbol processing and SA are incommensurable, this paper disputes that view. If the message of the SA community is taken to be that traditional approaches neglect the importance of the environment, then not only is the message an important one, but the typical symbol processing system is guilty as charged. However, this does not mean that, in principle, symbol processing systems must have this limitation. The two approaches can work hand-in-hand to produce more general and more accurate computational models. A framework of building models of the environment and having models of cognitive agents work with those models is proposed, from which a smooth integration of SA and symbol processing is not only possible, but desirable. The framework proposed here is instantiated with a production system called S-CAPS, and the efficacy of building models of both the problem-solver and the problem environment is demonstrated.
