Biologists have long sought concepts to describe the ways in which organisms are adapted to their environments. Social scientists have likewise sought concepts to describe the ways in which people become acculturated participants in the social worlds around them. Yet it has been difficult to approach these phenomena with the methods of computational modeling. We can see at least two reasons for this difficulty. The first is that the tradition of modeling in artificial intelligence developed around a concern with cognition, that is, mental processes understood to intervene between stimuli and responses in human beings. Although minority traditions such as ecological psychology reacted against this approach to studying human life, they have not been able to translate their concepts into computational mechanisms that match the expressive power of symbolic programming. The second reason is more subtle: if one conceives both organisms and their environments as spatially extended mechanisms that can be explained according to the same principles then the boundary between them (the surface of the body) is not particularly different from, or more interesting than, the rest of the total organism-environment system. The challenge for computational modeling, then, is to conceptualize agents' adaptations to their environments in ways that neither treat agents as isolated black boxes or dissolve them into one big machine.
For these purposes, we find it useful to distinguish between two aspects of an agent's involvement in its familiar environment: its embodiment and its embedding. ``Embodiment'' pertains to an agent's life as a body: the finiteness of its resources, its limited perspective on the world, the indexicality of its perceptions, its physical locality, its motility, and so on. ``Embedding'' pertains to the agent's structural relationship to its world: its habitual paths, its customary practices and how they fit in with the shapes and workings of things, its connections to other agents, its position in a set of roles or a hierarchy, and so forth. The concept of embedding, then, extends from more concrete kinds of locatedness in the world (places, things, actions) to more abstract kinds of location (within social systems, ecosystems, cultures, and so on). Embodiment and embedding are obviously interrelated, and they each have powerful consequences both for agents' direct dealings with other agents and for their solitary activities in the physical world. Our principal focus in this article is on embedding, and particularly on the ways in which agents maintain relationships to objects that are functionally significant for their tasks.
In this paper we develop some thoughts about embodiment and embedding as follows: