On the Origin of Objects
by Brian Cantwell Smith, The MIT Press, 1996, 420 p.

reviewed by André Vellino
National Research Council, Ottawa, 30 Sept. 1997

On the Origin of Objects is a seminal book. It challenges us to rethink our entire world-view and revitalizes the perennial questions of philosophy: what is knowledge?, what is it knowledge of?, what is the nature of consciousness?, what is meaning?, what is personal identity?, how ought I to act?, what is politics?, what is beautiful? It is also a courageous book: to address such sweeping questions entails creative metaphysical contemplations of a kind that will no doubt ruffle the feathers of academic philosophers. Even though he does not try to answer all these questions nor even tackle most of them head-on, Smith offers the reader a fresh metaphysical foundation from which pluralistic solutions to those problems may be fashioned.

Smith's arguments are complex and demand the reader's careful attention. To understand the core of Smith's book the reader must follow an exacting journey through its labyrinth of concepts, distinctions and arguments: it is useless to go straight to the conclusion without first going through the phases of problem "analysis" and solution "construction." The book is also impossible to compress. Any attempt to summarize or paraphrase even the general thrust of the argument will inevitably do it some injustice or other.

The starting point for Smith's multi-threaded excursion into a new metaphysics is not a new one for either AI or cognitive science. The desire to provide a solid foundation for a cognitivist computational theory of mind (the theory that mind is computational in nature) has motivated many a tome in recent years [1-7]. Smith's take on the problem, however, quickly leads into murkier philosophical waters than his predecessors had dared to venture before. In trying to break out of the mould of contemporary cognitive science--mind/body duality, representation schemas, semantics of computation etc.--Smith attempts to re-orient the reader to a new way of thinking about the questions.

The foundations of any field are either philosophical or mathematical, and this one is certainly not mathematical. Notwithstanding his disclaimer that this book is not philosophical either, Smith might have considered another title for it, one that better reflects its foundational and philosophical bent; something like "A Prolegomena to any Future Computational Theory of Mind", perhaps. Indeed, this book is a sort of preamble to a much more detailed and finely articulated 4-volume encyclopedic analysis of computational theory which Smith promises us for the future. We are left anticipating that the results will be to the foundations of computing as Knuth's trilogy was for the art of computer programming, i.e., a tour de force.

The basic problem with which Smith is wrestling that if we want to build software systems that are able to represent and reason about the world in which they operate, we need to understand both what is represented and the representation, not to mention the process of reasoning and norms of rationality. Yet the conventional model of unambiguous "objects" that have "relations" to one another and which need only to be represented in some symbolic computer language is a completely non-explanatory and inadequate account of both the world and its representation in a computer. Some new and foundational perspective is required.

Like other philosophical system builders that propose new metaphysical and epistemological foundations, Smith believes this task is necessary because of the failures of the past. He is right, for example, in noting that there has been some fuzzy thinking about semantics. The problem of semantics isn't solved by the proposition that computing is symbol manipulation, he says, since symbol manipulation is intentional (it is easy to guess where Smith stands on the "Chinese room" argument). Nor is the problem of intentionality eliminated by equating computation with the behaviour of finite-state automata. Indeed, it turns out that, for Smith, intentionality isn't reducible to anything.

Where then, does Smith stand on philosophical issues? In terms of familiar categories, one could say that his epistemology is that of a realist (in the sense that "there is a world out there" which our theories attempt to capture), that his methodology is naturalist (in the sense that explanations of all phenomena must be grounded in nature), and, although he denies it, his metaphysics is largely that of a materialist (in the sense that the furniture of our experience-and the essence of computation-is a material/physical phenomenon). But for many reasons with which it is easy to sympathize, Smith distrusts these familiar categories and the sorts of prejudices they conjure up. He wants his analysis to make room for different ways of speaking and different formalisms (i.e. pluralism) and so he doesn't side with any particular camp. Instead he tries to position himself vis a vis well-known theories (of truth, of semantics, of representation, etc.).

What then do these philosophical theories have to do with everyday objects like tables and chairs, or even with the problem of how to represent them for a computer? What Smith wants to do in his inquiry into "objects" is

"..[to find out what] is needed for a complete picture of intentionality, semantics and ontology ...[for which one needs] to understand how a conception of objects can arise on a substrate of infinitely extensive fields of particularity" [p. 191].

There are all sorts of things wrong with this clumsy formulation, as Smith himself readily admits, and he eventually offers us the notion of "registration" to shed some light on the matter. But to understand how our conception of objects arises at all we first need to understand the distinction between "particulars" and "individuals." Examples of particulars include the constitutive elements of what we ordinarily call (physical) objects (the physical cause of this patch of green/blue experience) but also events (such as Diana's accident) and processes (the writing of this review). Contrast this with universals (e.g. the generic concept "canoe") and abstract categories. The distinction between particulars and individuals-and this is an important one-is that individuality is what makes an object discrete and separable from its background. An individual is constituted of particulars, but what makes it an individual is "... what allows us to say of one object that it is one; or two that they are two" [p. 119].

One might well think that the science of physics would do quite well as a foundational theory of particularity, individuality and objecthood. After all, physics is the star candidate for an explanatory theory of all there is, and Smith has to explain why it is not enough. The argument, which reads like a romp through 80 years of philosophy of science--ranging from the ontological presuppositions of classical and quantum mechanics to Popper's "three worlds" theory--concludes that physics has something to say about particulars but is silent about individuals. One of the lessons drawn from this foray into physics is the criterion of ultimate concreteness:

No naturalistically palatable theory of intentionality-of mind, computation, semantics, ontology, objectivity-can presume the identity or existence of any individual object whatsoever.[p. 184, italics in the original]

It turns out that what distinguishes an individual from the "rest of the world" is something abstract and dependent on the subject that is doing the abstraction. Hence the problem of what "objects" are and how we "register" them remains.

What, then, is "registration"? In brief, we "register" the world (we take it to be a certain way, to be composed of certain things) not because we conceive it to be this way or even perceive it to be that way (as the 18th century British Empiricists might have it) but because its being what it is requires an intentional relation to a subject. For Smith, the intentional nature of objects and their registration in our minds will never be reduced, as an eliminative materialist would have it, to some causal story about neurons firing in some particular way as a result of stimuli of a certain kind. Registration, it seems, is a primitive and irreducible intentional relation between mind and object.

The modalities of what precisely counts as registration, by whom and how it takes place, obliges Smith to sketch a theory of causality. Here again, nothing is settled but a metaphysical picture is painted that is plausible: the world is neither entirely deterministic--a world of intertwined cogs and gears--nor is everything physically and causally disconnected from everything else. Smith goes for the middle ground, one that captures both the intuition that what I am doing now is not terribly relevant or causally influential to whether you are or are not currently drinking a glass of water, and the intuition that my use of garlic as an ingredient for the pasta sauce I cooked last night had a significant influence on my guests' enjoyment of the meal.

That bears on registration in the following way: an object is registered by a subject not only by the causal link between them (the "world" impinging on the senses, for example), but also by how it does not causally relate to the subject. In particular, an object's continued existence or movement in space bears an intentional relation to the subject that has as much to do with its causal relations to the subject as it does to the causal disconnection between the subject and the object.

In addition to a theory of causality, Smith needs a theory of reference (and, therefore, Truth). For it must be possible for me to register objects not only as present, here and now, in my perceptual field, but also as enduring, "stable" objects that preserve their identity somehow. This process of stabilization requires the ability to deal with indexicals (such as this and now) which, in turn, demands a theory of reference.

To explain this process of object registration, Smith spends chapter 7 (one of two on registration) speaking of "s-regions" and "o-regions" instead of the more vexed "subjects" and "objects" and of the important (and, I take it, intentional) "letting go" of one by the other, as an essential requirement for registration. I think Smith means that the process of individuation that takes place in selecting a "thing" from its environment and registering it as a thing (or object) requires an (intentional) act of separation or discrimination.

In any event, the registration of an object is no trivial matter. Apart from the problem of context ("A teacup's exemplification of the property of being a teacup does not inhere in its meager six ounces; it reaches back into British society" [p. 269]), there is the question of how we register higher-order cognitive objects like concepts and the process of non-cognitive registration--that which we register simply as a consequence of being alive and derives from our "encounter" with things. This means, that there must be an equilibrium (a "middle distance") between the degree to which an object is causally connected with the subject and the degree to which it is separate. Ditto with the extent to which the subject's concepts participate in the object's being what it is and the extent to which we are viscerally "engaged" with the object.

The upshot of Smith's analysis of registration is an affirmation of what he calls "a philosophy of presence". He writes:

[Its metaphysical viewpoint is] a commitment to One world, a world with no other, a world in which both subjects and objects-we and the things we interact with and think about and eat and build and till and are made of and give away as presents-are accorded appropriate place. ... that world is depicted as one of cosmic and ultimately ineffable particularity.... Neither formally rigid nor nihilistically sloppy, the flux [of particulars] sustains complex processes of registration: a form of interaction, subsuming both representation and ontology, in which "s-regions" or subjects stabilize patches of the flux, in part through processes of intervention, adjusting and building them and beating them into shape, and also through patterns of disconnection and long-distance coordination, necessary in order to take the patch to be an object, or more generally, to be something in and of the world. [p. 347]

But this book is not merely a metaphysical treatise on the philosophical foundations of computer science. Smith aims to touch the reader on many other planes as well. He warns us early on that:

The story to be told represents an attempt to unify and therefore, with luck to help heal, the schism between the academic-cum-intellectual-cum-technological on the one hand, and the curious, the erotic, the spiritual, the playful, the humane, the moral, the artistic, the political and the sheerly obstreperous on the other. [p. 94]

Although we get intimations about how this might be as the narrative unfolds, it is only in the conclusion that Smith explicitly sketches how a "philosophy of presence" helps to shed light on a wide range of issues: the representation problem in AI (how does a neural net "register" a face?); the realist/constructivist (instrumentalist) debate in philosophy of science (how does a physicist register a neutrino?); the Platonist/intuitionist/constructivist debate in mathematics (how does a mathematician register a number?) etc.

One wonders whether the aesthetics of the book itself is intended to achieve the goal of conveying the many-layered implications of his theory at another "register" in the reader's consciousness: the layout is original and creative, its structural composition is exceptionally ergonomic and the style of prose is not always complex and philosophical. Smith often tries to speak-easy and give us the gritty feeling for his intuitions, a welcome break from the often strong academic-cum-intellectual-cum-technological flavour of the remainder of the text. Every other page has an instructive drawing of some sort that illustrates a point and there are many sidebars that take the place of stretched-out parenthetical remarks. Near the end of the book, just as the reader is led into Smith's metaphysics and conclusion, there is an especially beautiful picture of a canvas by Adam Lowe. It is as though Smith thought that a contemplation of the art-piece by itself could convey the ineffable intuition for his philosophy of presence.

There is a lot to be grateful for in this book. It stimulates not just the intellect but also the spirit and the imagination. It offers a way to understand the stuff of objects and their inextricable ties to mentality in a way that evenly balances reason and logic with intuition and mystery. Perhaps its greatest value is to have rekindled the possibility of conversations about the ineffable, about the inter-relatedness among all things and to have shown how completely relevant these metaphysical discussions are to the science of computing.

Further Reading

[1] Jerry Fodor, The Modularity of Mind, MIT Press 1983.

[2] Zenon Pylyshyn, Computation and Cognition, MIT Press 1984.

[3] John Haugeland, Artificial Intelligence: The Very Idea, MIT Press 1985.

[4] Patricia Churchland, Neurophilosophy, MIT Press 1989.

[5] Daniel Dennett, Consciousness Explained, Boston: Little, Brown 1991.

[6] John Searl, The Rediscovery of the Mind, MIT Press, 1992.

[7] David Chalmers, The Concious Mind, Oxford, 1996.