MITECS : Situatedness/Embeddedness

Brian Cantwell Smith FIRST DRAFT

The "situated" movement -- situated language, situated cognition, situated behavior -- arose in the 1980s as a reaction against the then-dominant classical view of mind. The classical approach, inherited from logic and the metamathematical tradition, and embodied in various symbolic and representational models, viewed cognition as: individual, in the sense that the paradigmatic cognitive site was assumed to be a solitary person; rational, in the sense that deliberative, conceptual thought was taken to be the primary exemplar of cognitive activity; abstract, in the sense that issues of physical realization and material environment were treated as of secondary importance; detached, in the sense that thinking was treated as logically distinct from action; and general, in the sense that cognitive science was presumed to be in the business of looking for universal principles, true of all individuals and applicable in all possible circumstances. Situated approaches reject (one or more of) these assumptions, arguing instead that cognition should be treated as continuous with other forms of human activity, and understood as: social, in the sense of being located in human settings among communities of people; embodied, in that material aspects of the body are both pragmatically and theoretically significant; concrete, in the sense that physical constraints of realization and of circumstance are of the utmost importance; located, implying that context-dependence is a general feature of human endeavor; engaged, in that interaction with the surrounding environment is primary; and specific, in that what people do varies, dramatically, depending on particular facts about them and their local situations. In sum, the situated movement, rejecting the traditional model as an abstract and falsely-idealized myth, aims to "get the world back" by recognizing the local, material context as a resource in the very center of human epistemological activity, not a distraction or complexity at the periphery.

Within these broad outlines, situated approaches vary widely in detail, ranging from less theoretically "expensive" proposals, which incorporate some degree of context-dependence into classical frameworks, to more radical suggestions with substantial methodological and metaphysical consequences.

Among the best worked out, if simplest, are "situated language" proposals to handle indexicals, tense, and other linguistic constructs that allow people to use the same words to say different things in ways that depend systematically on the circumstances of use. As even young children realize, words like 'here', 'I', and 'now' -- and the present tense -- can be used on different occasions, by different individuals, to refer to distinct people, places, and times. Barwise and Perry label as meaning the stable "rule" or pattern that the child learns (e.g., that 'I' is used by a speaker to refer to him or herself), and as interpretation the distinct semantic values that different uses denote on different occasions. Thus, on their account, when two people shout "I'm right! You're wrong!" their utterances coincide in meaning, but differ in interpretation.

So regarded, meaning can be modelled as a function from context to interpretation. The model has wider generality. For example, the phrase '4:00' can similarly be assigned a (single, constant) meaning, mapping locations of utterance onto different times, according to contextually-determined day (and time zone). The approach can also be extended to anaphora, ambiguity, and other forms of circumstantially-determined interpretation. At a theoretical level, accomodating such examples requres shifting focus from sentence types to individual utterances, and from a view of inference as truth-preservation to a more generalized picture of inference as reference-preservation (to understand, e.g., why tomorrow we use 'yesterday' to refer to what today we refer to with 'today').

Many, however, feel that situated intuitions run deeper, pointing out that these purely linguistic examples remain compatible with a classical view of cognition as individualistic, deductive, and even relatively abstract. A further step is taken by Kirsh, who sees an agent's embedding situation not only as a semantical resource for determining reference, but as a material resource to simplify the thought process itself. You don't have to remember something, if you can place it in your visual field -- or measure two objects to see which one is longer, if you can simply place one on top of the other and see which one sticks out. In general, that is, people can often "let the world do the computing for them," and determine things by inspection, rather than shouldering the entire load deductively. And if the world happens not to provide exactly what they want, they can sometimes re-arrange it a little so that it does. A near-mythic example, cited by Lave (p. ???), describes someone who, when asked to make 3/4 of a recipe that called for 2/3 of a cup of cottage cheese, measured out 2/3 of a cup, smooshed it into a flattened circle, and cut away 1/4 of the resulting figure.

Not surprisingly, proposals to look at things this way have been accompanied by rearrangements in disciplinary allegiance, reducing emphasis on ties to logic and mathematics, and increasing connections with such non-traditional literatures as Heiddeggerian discussions of equipment and tool use, and Gibsonian theories of non-representational ("direct") perception and ecological realism. These developments have been accompanied by a revival of interest in robotics, especially of a non-representational variety. Norman's work on psychologically accessible design illustrates a similar approach: he argues for the direct manifestation of the function of artifacts in their structure--arguing, for example, for arranging the buttons on a stove panel in such a way as to mirror the arrangement of burners on the stove top, to minimize conceptual confusion.

To some extent, this shift in theoretical focus away from such classic examples as theorem-proving and chess, towards more concrete activities such as walking down a hallway and cooking dinner, has undermined the view that ratiocination is the central or highest form of human activity. It is starting to be more common to find dance used as a general metaphor for human activity, in place of deduction or problem solving. The historical irony is strong. Descartes, a forebear of cognitive science, believed that movement, sensation, and appropriate reaction to the physical world lay within in the province of "mere brutes." For him, it was high-level conceptualization that was challenging, and paradigmatic of what it is to be human. But times have changed. From the simple prowess of making one's way across a Tokyo subway station platform at rush hour, to the phenomenological character of conscious experience, it has turned out to be the perceptual, the sensational, and the qualitative that have stubbornly resisted computational modeling, more than the purely conceptual or deductive.

The change in focus from thinking to navigating was accompanied by a temporal shift, from advance planning to on-the-fly improvisation. In the robotic arena, Brooks, Agre and Chapman, and Rosenschein have all argued that embodied agents can inventively exploit specific facts about their physical circumstances to avoid explicit representation and reasoning. Perhaps the strongest theoretical blow, however, was struck by Suchman, who claimed that most human activity, rather than implementing pre-conceptualized plans, consists instead of an incessant, creative, improvisational appropriation of the vast array of resources that environments regularly make available. Not only do people rarely "figure it all out in advance," she argues, but also, and perhaps more significantly, the stories they tell should be understood as after-the-fact reconstructions whose role is to retrospectively render activity intelligible (and perhaps accountable), rather than as veridical reports of how it originally came to be.

Suchman's work, coming out of anthropology, reflected another shift in disciplinary affiliation, this time to include a group of socially-oriented disciplines, where situated approaches have also been on the rise, such as sociology, science studies, feminist epistemology, and philosophy of science. To date, some of the stronger methodological consequences of a deep situated approach have been more evident in these adjacent disciplines, though they are becoming more salient in cognitive science as well. One example is a shift in emphasis away from in vitro studies towards in vivo accounts: "thick narrative descriptions" of real people operating in real-life situations, instead of statistical results based on laboratory experiments. More serious issues have to do with the nature of knowledge itself -- whether what we know may not be circumstantial in ways that go deeper than the obvious fact that, in any given situation, we are more likely to focus on what is nearby than on what is remote. These arguments tie directly into considerations of pluralism, of the contestation of clashing world views, of inadequacies intrinsic to mathematical modelling, and into socio-political critiques of science, where it is argued that the content of even as supposedly stable a form of knowledge as scientific theories may only be explicable with at least partial reference to social, political, and personal agendas of the scientists. Haraway's "Situated Knowledges," a landmark article from a feminist critic of science that is representative of these concerns, has begun to gain substantial currency within cognitive science.

Methodological and epistemological considerations such as these lead in turn to what is perhaps the strongest version of a situated approach, where the notion of situatedness is carried all the way through to metaphysics. Return to the first example given above, where it was claimed that indexical expressions can be viewed as having a single, stable, constant "meaning," even if they vary, on different occasions, in their "interpretation". Some argue that this approach to contextual dependence does not go far enough: that meaning, too, can be (often is) bent and shaped by contingent exigencies of the situation. Winograd gives an example in which the term 'water' in the question "Is there any water in the refrigerator?" can vary in meaning, depending on whether the person asking is thirsty, is worried about humidity and condensation, or is testing a child's understanding of the constitution of eggplants. Smith considers a case where two friends, at the end of a late-night conversation, bend and shape the meanings of the words they use -- rather in the way that blues guitar players bend notes to suit the exigencies of a song. What makes such examples metaphysically significant, as Smith, Cussins, and others argue, is that it is not just meaning that varies, in such cases, but ontology as well -- not just that your friends use the word 'chair' (or 'theory') differently from you, but that what it is to be a chair (or a theory) may be different for them, as well. Not only do what we do, what we say, and how we get at the world depend on facts about our circumstances, on such a view, but how the world is -- the world we talk about and live in and get at -- may depend on circumstance, as well. Taken to its logical extreme, that is, situatedness can lead to a view of a much more plastic world that, in contrast to the realism implicit in the classical picture, is at least partially constructive. This is situatedness with a vengeance.

At the broadest level, it is safe to say that the basic tenet of the situated movement has been largely accepted. That language, cognition, and activity are fundamentally context-dependent is by now almost de rigueur . What remains open to debate is just how far this situated intuition should be taken. Whereas everyone agrees that language is indexical, a smaller number are prepared to de-thrown ratiocination as the hallmark of the human, and those who are prepared to take context-dependence through to metaphysics remain in a distinct minority. How the issues are resolved may depend on the extent to which the various communities advocating a situated approach -- linguistics, cognitive science, AI, sociology, epistemology, feminism, philosophy, science studies, etc. -- can collaborate in effecting this transformation in our self-conception.

Related entries

ROBOTICS
PLANNING
INTELLIGENT AGENT ARCHITECTURE
ANIMAL NAVIGATION
ECOLOGICAL PSYCHOLOGY
FRAME PROBLEM

References

Agre, P., and D. Chapman. (1987). Pengi: An implementation of a theory of activity. In The Proceedings of the Sixth National Conference on Artificial Intelligence . American Association for Artificial Intelligence. Seattle: Morgan Kaufmann, pp. 268-272. Barwise, J. and J. Perry. (1983). Situations and Attitudes . Cambridge, MA: MIT Press. Brooks, R. A. (1997). Intelligence Without Representation. Artificial Intelligence , 47: 139-159 (1991). Revised version in J. Haugeland (Ed.), Mind Design II: Philosophy, Psychology, Artificial Intelligence -- 2nd ed., rev. and enlarged . Cambridge, MA. MIT Press, pp. 395-420. Cussins, A. (1992). Content, Embodiment and Objectivity: The Theory of Cognitive Trails. Mind 101, pp 651-688. Gawron, M. and S. Peters. (1991). Anaphora and Quantification in Situation Semantics (CSLI Lecture Notes, No. 19) . Stanford, CA: Center for the Study of Language and Information (Chicago Univ. Press, dist.). Haraway, D. (1991) Situated Knowledges: The Science Question in Feminism and the Privilege of Partial Perspective. In D. Haraway, Simians, Cyborgs, and Women: The Reinvention of Nature , New York: Routledge. Haugeland, J. (1997) What is Mind Design?. In J. Haugeland (Ed.), Mind Design II: Philosophy, Psychology, Artificial Intelligence -- 2nd ed., rev. and enlarged . Cambridge, MA. MIT Press, pp. 1-28. Kirsh, D. (1995). The intelligent use of space. Artificial Intelligence 72: 1-52. Lave, J., M. Murtaugh and O. de la Rocha. (1984). The dialectic of arithmetic in grocery shopping. In B. Rogoff and J. Lave (Eds.), Everyday Cognition: Its Development in Social Context. Cambridge, MA: Harvard University Press, pp. 67-94. Rosenschein, S. and L. Kaelbling. (1995). A Situated View of Representation and Control". In Agre, P. A. and S. Rosenschein (Eds.), Special Issue on Computational Research on Interaction and Agency, Artificial Intelligence , January/February 1995. Smith, B. C. (1996) On the Origin of Objects . Cambridge, MA: MIT Press. Suchman, L. A. (1987). Plans and Situated Action. New York: Cambridge University Press. Winograd, T. (1985). Moving the Semantic Fulcrum. Linguistics and Philosophy 8(1): 91-104.  

Further Reading

Agre, P. A. (1997). Computation and Human Experience . New York: Cambridge University Press. Agre, P. A. and S. Rosenschein (Eds.). (1995). Special Issue on Computational Research on Interaction and Agency, Artificial Intelligence , January/February 1995. Clancey, W. J. (1997). Situated Cognition: On Human Knowledge and Computer Representations . New York: Cambridge University Press. Clark, A. (1997). Being There: Putting Brain, Body, and World Together Again . Cambridge, MA: MIT Press. Dreyfus, B. (1979). What Computers Can't Do . Hutchins, E. (1995). Cognition in the Wild . Cambridge, MA: MIT Press. Latour, B. (1986). Visualization and cognition: thinking with eyes and hands. Knowledge and Society: Studies in the Sociology of Culture Past and Present 6: 1-40. Lave, J. (1988). Cognition in Practice: Mind, Mathematics, and Culture in Everyday Life. New York: Cambridge University Press. Lave, J. and E. Wenger. (1991). Situated Learning: Legitimate Peripheral Participation . New York: Cambridge University Press. Norman, D. A. (1987). The Psychology of Everyday Things. New York: Basic Books. Vera, J. and H. Simon. (1993). Situated action: a symbolic interpretation. With replies. Cognitive Science 17(1): 7-48. Winograd, T. and F. Flores. (1986). Understanding Computers and Cognition: A New Foundation For Design . Norwood, NJ: Ablex.