COGNITIVE SCIENCE 17, 7-48 (1993) Situated Action: A Symbolic Interpretation ALONSO H. VERA AND HERBERT A. SIMON Carnegie Mellon University The congeries of theoretical views collectively referred to as "situated action" (SA) claim that humans and their interactions with the world cannot be under- stood using symbol-system models and methodology, but only by observing them within real-world contexts or building nonsymbolic models of them. SA claims also that rapid, real-time interaction with a dynamically changing environment is not amenable to symbolic interpretation of the sort espoused by the cognitive science of recent decades. Planning and representation, central to symbolic theories, are claimed to be irrelevant in everyday human activity. We will contest these claims, as well as their proponents' characterizations of the symbol-system viewpoint. We will show that a number of existing symbolic systems perform well in temporally demanding tasks embedded in complex en- vironments, whereas the systems usually regarded as exemplifying SA are thoroughly symbolic (and representational), and, to the extent that they are limited in these respects, have doubtful prospects for extension to complex tasks. As our title suggests, we propose that the goals set forth by the proponents of SA can be attained only within the framework of symbolic systems. The main body of empirical evidence supporting our view resides in the numerous symbol systems constructed in the past 35 years that have successfully simulated broad areas of human cognition. During the past few years a point of view has emerged in artificial intelli- gence, often under the label of "situated action" (henceforth, SA), that denies that intelligent systems are correctly characterized as physical symbol systems, and especially denies that symbolic processing lies at the heart of - - This research was supported by the Defense Advanced Research Projects Agency, Depart- ment of Defense, ARPA Order 3597, monitored by the Air Force Avionics Laboratory under contract F33615-81-K-1539, and by the Office of Naval Research. Cognitive Science Program. under Contract No. N00014-89-J-1975N158. We are grateful to John Anderson, Susan Chipman, BOM~ John, Jeff Shrager, Jim Greeno, Joyce Moore, Phil Agre, and Lucy Suchman for their comments on earlier drafts of this article. Conversations with Allen Newell contributed much to developing and sharpening our ideas. As always, all of those who helped must be absolved from responsibility for the final product, which is ours alone. Correspondence and requests for reprints should be sent to Alonso Vera, Department of Psychology, Carnegie Mellon University, Pittsburgh, PA 15213.8 VERA AND SIMON intelligence. In fact, SA does not denote a single, sharply delineated posi- tion, but a whole congeries of closely related views that share a deep skep- ticism about the dominant role of symbol systems in the intelligence human beings exhibit, especially in their everyday behavior and in their response to complex or real-time situations. In this article, we wish to examine whether SA is actually antithetical to symbolic manipulation.To anticipate our conclusions, we find that there is no such antithesis: SA systems are symbolic systems, and some past and present symbolic systems are SA systems. The symbolic systems appropriate to tasks calling for situated action do, however, have specialcharacteristics that are interesting in their own right. Because there is no official credo to which all those usually associated with SA subscribe, and the points that different authors emphasize are sometimes quite different, wewill focus our remarks on the central theme: What is the role of symbol systems in intelligence? Later on, we will sort out and comment upon some of the substrands in the SA literature, but it must be understood that not all of the subprinciples would be accepted by all of those whom we identify with the SA label. If particular SA feet do not fit a particular shoe that we mention, they should not be squeezed into it; no single shoe will fit them all. Of course, the argument also works in reverse: The symbol-system point of view is no more monolithic than the SA view, and some of us who subscribe to the former view do not always recognize the caricatures of our position that appear in SA critiques of it. However, as suggested before, there does. appear to be a central theme that separates the two positions, and we will focus our attention on that. In the first section of this article we will state briefly what we mean by a physical symbol system, so as to provide a precise template with which we can compare SA systems. In the second section, we will review the accounts of SA that have been given by various of its proponents, taking account of the differences among them and the aspects they single out as salient. In the third section, we will examine some "classical" symbolic systems that were designed for responding, usually in real time, to real or synthetic but com- plex external environments. We will see how these systems resemble or dif- fer from those designed by the proponents of SA. Then we will compare the symbol systems with some of the more prominent SA systems that have been described and actually implemented. In the fourth section, we will set forth the theoretical conclusions we have drawn from our examination of the two kinds of systems, followed by a brief summary. 1. PHYSICAL SYMBOL SYSTEMS A physical symbol system is built from a set of elements, called symbols, which may be formed into symbol structures by means of a set of relations.SITUATED ACTION: A SYMBOLIC INTERPRETATION 9 A symbol system has a memory capable of storing and retaining symbols and symbol structures, and has a set of information processes that form symbol structures as a function of sensory stimuli, which produce symbol structures that cause motor actions and modify symbol structures in memory in a variety of ways. A physical symbol system interacts with its external environment in two ways: (1) It receives sensory stimuli from the environment that it converts into symbol structures in memory; and (2) it acts upon the environment in ways determined by symbol structures (motor symbols) that it produces. Its behavior can be influenced both by its current environment through its sen- sory inputs, and by previous environments through the