The revision of the central notion of representation for cognitive systems starts from considering that these are characterized by their operational closure, wherein to outline where the represented external environment (the world) ends and where the internal processes of the cognitive system, which represent it, begin is hardly to define. Such systems do not operate by representation but instead of representing an independent world they enact a world as a domain of distinctions that is inseparable from the structure embodied by the cognitive system:
Steps to a Middle Way
The Cartesian Anxiety
Representation Revisited
In the discussion of cognitivism we distinguished between two senses of representation, which we now need to recall. On the one hand, there is the relatively uncontroversial notion of representation as construal: cognition always consists in construing or representing the world a certain way. On the other hand, there is the much stronger notion that this feature of cognition is to be explained by the hypothesis that a system acts on the basis of internal representations. Since it might seem that these two ideas amount to the same thing, we need to refine our distinction somewhat.
We can begin by noting a relatively weak and uncontroversial sense of representation. This sense is purely semantic: It refers to anything that can be interpreted as being about something. This is the sense of representation as construal, since nothing is about something else without construing it as being some way. A map, for example, is about some geographical area; it represents certain features of the terrain and so construes that terrain as being a certain way. Similarly, words on a page represent sentences in a language, which may in tum represent or be about still other things. This sense of representation can be made even more precise. If, for example, our concern happens to be with languages in a more formal setting, we can say that the statements of a language represent their conditions of satisfaction. For example, the statement "snow is white"-taken literally is satisfied if snow is white; the statement "pick up your shoes"again, taken literally - is satisfied if the shoes are picked up by the person being addressed.
This sense of representation is weak because it need not carry any strong epistemological or ontological commitments. Thus it is perfectly acceptable to speak of a map representing the terrain without worrying about such things as how maps get their meaning. It is also perfectly acceptable to think of a statement as representing some set of conditions without making further assumptions about whether language as a whole works this way or whether there really are facts in the world separate from language that can then be re-presented by the sentences of the language. Or we can even talk about experiential representations, such as the image I have of my brother, without making any further assumptions about how this image arose in the first place. In other words, this weak sense of representation is pragmatic; we use it all the time without worry.
The obviousness of such an idea, however, is quickly transformed into a much stronger sense of representation that does carry quite heavy ontological and epistemological commitments. This strong sense arises when we generalize on the basis of the weaker idea to construct a full-fledged theory of how perception, language, or cognition in general must work. The ontological and epistemological commitments are basically twofold: We assume that the world is pregiven, that its features can be specified prior to any cognitive activity. Then to explain the relation between this cognitive activity and a pregiven world, we hypothesize the existence of mental representations inside the cognitive system (whether these be images, symbols, or subsymbolic patterns of activity distributed across a network does not matter for the moment). We then have a full-fledged theory that says (1) the world is pregiven; (2) our cognition is of this world-even if only to a partial extent, and (3) the way in which we cognize this pregiven world is to represent its features and then act on the basis of these representations.
We must, then, return to our earlier metaphor, the idea of a cognitive agent that is parachuted into a pregiven world. This agent will survive only to the extent that it is endowed with a map and learns to act on the basis of this map. In the cognitivist version of this story, the map is an innately specified system of representations-sometimes called a "language of thought" -whereas learning to employ this map is the task of ontogeny.
Many cognitive scientists will object that we have presented a caricature. Are we not presupposing a static conception of representation, one that overlooks the rich detail of the inner structure of a cognitive system and unjustifiably construes a representation as merely a mirror? Is it not well known, for example, that visual perception is considered to be a result of mapping the physical patterns of energy that stimulate the retina into representations of the visual scene, which are then used to make inferences and eventually to produce a perceptual judgment? Perception is seen as an active process of hypothesis formation, not as the simple mirroring of a pregiven environment.
This objection, though somewhat fair, misses the point. Our point is not to caricature a sophisticated research program but simply to render explicit some tacit epistemological assumptions in as clear a fashion as possible. Thus although everyone agrees that representation is a complex process, it is nonetheless conceived to be one of recovering or reconstructing extrinsic, independent environmental features. Thus in vision research, for example, one speaks of "recovering shape from shading" or "color from brightness." Here the latter features are considered to be extrinsic properties of the environment that provide the information needed to recover "higher-order" properties of the visual scene, such as shape and color. The basic idea of a world with pregiven features remains.
The complaint that we have presented a caricature would, however, be justified were we not to acknowledge the subtlety and sophistication of cognitive realism in relation to the classical opposition between realism and idealism in philosophy. In the hands of cognitive realism, the notion of representation does undergo something of a mutation. The power of this mutation is that it seems to offer a way out of the classical opposition between realism and idealism.
This opposition is based in the traditional notion of representation as a "veil of ideas" that stands between us and the world. On the one hand, the realist naturally thinks that there is a distinction between our ideas or concepts and that which they represent, namely, the world. The ultimate court of appeal for judging the validity of our representations is this independent world. Of course, each of our representations must cohere with many others, but the point of such internal features is to increase the probability that globally our representations will have some measure of correspondence or degree of fit with an outer and independent world.
The idealist, on the other hand, quickly points out that we have no access to such an independent world except through our representations. We cannot stand outside of ourselves to behold the degree of fit that our representations might have with the world. In fact, we simply have no idea of what the outside world is except that it is the presumed object of our representations. Taking this point to the extreme, the idealist argues that the very idea of a world independent of representations is itself only another of our representations – a second-order or metarepresentation. Our sense of an outer ground thus slips away, and we are left grasping for our internal representations, as if these could provide a sure and stable reference point.
At first sight, contemporary cognitive science seems to offer a way out of this traditional philosophical impasse. Largely because of cognitive science, philosophical discussion has shifted from concern with a priori representations (representations that might provide some noncontingent foundation for our knowledge of the world) to concern with a posteriori representations (representations whose contents are ultimately derived from causal interactions with the environment). This naturalized conception of representation does not invite the skeptical questions that motivate traditional epistemology. In fact, to shift one's concern to organism-environment relations in this way is largely to abandon the task of traditional a priori epistemology in favor of the naturalized projects of psychology and cognitive science. By taking up such a naturalized stance, cognitive science avoids the antinomies that lurk in transcendental or metaphysical realism, without embracing the solipsism or subjectivism that constantly threatens idealism. 'The cognitive scientist is thus able to remain a staunch realist about the empirical world while making the details of mind and cognition the subject of his investigations.
Cognitive science thus seems to provide a way of talking about representation without being burdened by the traditional philosophical image of the mind as a mirror of nature. But this appearance is misleading. It is true, as Richard Rorty remarks, that there is no way to raise the traditional skeptical questions of epistemology in cognitive science. Global skepticism about the possibility of cognition or knowledge is simply not to the point in the practice of science. But it does not follow, as Rorty seems to think, that the current naturalized conception of representation has nothing to do with the traditional image of the mind as a mirror of nature. On the contrary, a crucial feature of this image remains alive in contemporary cognitive science - the idea of a world or environment with extrinsic, pregiven features that are recovered through a process of representation. In some ways cognitivism is the strongest statement yet of the representational view of the mind inaugurated by Descartes and Locke. Indeed, Jerry Fodor, one of cognitivism's leading and most eloquent exponents, goes so far as to say that the only respect in which cognitivism is a major advance over eighteenth- and nineteenth-century representationism is in its use of the computer as a model of mind.
As we have seen, however, cognitivism is only one variety of cognitive realism. In both the emergence and society of mind approaches (and in the schools of basic elements analysis for the experiential pole of our investigation), the notion of representation becomes more and more problematical. We did not explicitly question this notion in our discussion of the varieties of cognitive realism, but if we look back on our journey, we can see that we have slowly drifted away from the idea of mind as an input-output device that processes information. The role of the environment has quietly moved from being the preeminent reference point to receding more and more into the background, while the idea of mind as an emergent and autonomous network of relationships has gained a central place. It is time, then, to raise the question, What is it about such networks, if anything, that is representational?
To make this question somewhat more accessible, consider once again Minsky's discussion toward the end of Society of Mind. There he writes, "Whenever we speak about a mind, we're speaking of the processes .that carry our brains from state to state ... concerns about minds are really concerns with relationships between states - and this has virtually nothing to do with the natures of the states themselves." How, then, are we to understand these relationships? What is it about them that makes them mindlike?
The answer that is usually given to this question is, of course, that these relationships must be seen as embodying or supporting representations of the environment. Notice, however, that if we claim that the function of these processes is to represent an independent environment, then we are committed to construing these processes as belonging to the class of systems that are driven from the outside, that are defined in terms of external mechanisms of control (a hetero-nomous system). Thus we will consider information to be a prespecified quantity, one that exists independently in the world and can act as the input to a cognitive system. This input provides the initial premises upon which the system computes a behavior-the output. But how are we to specify inputs and outputs for highly cooperative, self-organizing systems such as brains? There is, of course, a back-and-forth flow of energy, but where does information end and behavior begin? Minsky puts his finger on the problem, and his remarks are worth quoting at length:
Why are processes so hard to classify? In earlier times, we could usually judge machines and processes by how they transformed raw materials into finished products. But it makes no sense to speak of brains as though they manufacture thoughts the way factories make cars. The difference is that brains use processes that change themselves-and this means we cannot separate such processes from the products they produce. In particular, brains make memories, which change the ways we'll subsequently think. The principal activities of brains are making changes in themselves. Because the whole idea of self-modifying processes is new to our experience, we cannot yet trust our commonsense judgement about such matters.
What is remarkable about this passage is the absence of any notion of representation. Minsky does not say that the principal activity of brains is to represent the external world; he says that it is to make continuous self-modifications. What has happened to the notion of representation?
In fact, an important and pervasive shift is beginning to take place in cognitive science under the very influence of its own research. This shift requires that we move away from the idea of the world as independent and extrinsic to the idea of a world as inseparable from the structure of these processes of self-modification. This change in stance does not express a mere philosophical preference; it reflects the necessity of understanding cognitive systems not on the basis of their input and output relationships but by their operational closure.
A system that has operational closure is one in which the results of its processes are those processes themselves. The notion of operational closure is thus a way of specifying classes of processes that, in their very operation, tum back upon themselves to form autonomous networks. Such networks do not fall into the class of systems defined by external mechanisms of control (heteronomy) but rather into the class of systems defined by internal mechanisms of self-organization (autonomy). The key point is that such systems do not operate by representation. Instead of representing an independent world, they enact a world as a domain of distinctions that is inseparable from the structure embodied by the cognitive system.
We wish to evoke the point that when we begin to take such a conception of mind seriously, we must call into question the idea that the world is pregiven and that cognition is representation. In cognitive science, this means that we must call into question the idea that information exists ready-made in the world and that it is extracted by a cognitive system, as the cognitivist notion of an informavore vividly implies.
But before we go any further, we need to ask ourselves why the idea of a world with pregiven features or ready-made information seems so unquestionable. Why are we unable to imagine giving up this idea without falling into some sort of subjectivism, idealism, or cognitive nihilism? What is the source of this apparent dilemma? We must examine directly the feeling that arises when we sense that we can no longer trust the world as a fixed and stable reference point.