The New Science of the Mind
Page 1
From Extended Mind to Embodied Phenomenology
Mark Rowlands
Preface and Acknowledgments ix
1 Expanding the Mind 1
2 Non-Cartesian Cognitive Science 25
3 The Mind Embodied, Embedded, Enacted, and Extended 51
4 Objections to the Mind Amalgamated 85
5 The Mark of the Cognitive 107
6 The Problem of Ownership 135
7 Intentionality as Revealing Activity 163
8 The Mind Amalgamated 189
Notes 219
References 229
Index 239
I suppose that for a book entitled The New Science of the Mind there isn't, in fact, a lot of science in it; and some of the science that does find its way in isn't particularly new. The underlying reason is that the expression "new science" is as much aspirational as descriptive. There is, as yet, no new science, not understood as something comparable to mature cognitive science in its classical guise. Rather, "new science" picks out a cluster of related views that are emerging, or have emerged, from a variety of disciplines, including cognitive and developmental psychology, situated robotics and artificial intelligence, perceptual psychology, cognitive neuroscience, and philosophy. The role of philosophy is not so much to detail these developments as to provide a logical or conceptual foundation for them. That is the task of this book. It might have been called "Foundations of the New Science of the Mind," but that is a bit of a mouthful.
The book is written for philosophers, cognitive scientists, and any interested lay persons who want to understand what is being talked about when people throw around phrases like "situated cognition," "embodied cognition," "the extended mind," and their perhaps more arcane variants ("enactivism," "vehicle externalism," "locational externalism," "architecturalism," and the like). When writing for a disparate constituency, the danger is, of course, that at least some of this constituency, at least some of the time, might be tempted to think that I am in the business of teaching their grandmothers to suck eggs. I have tried to avoid this as much as possible, and in cases where I can clearly envisage it happening, I have inserted the offending passages into boxes.
Over the past decade or so, conversations with Fred Adams, Ken Aizawa, Andy Clark, Shaun Gallagher, Richard Menary, Robert Rupert, John Sutton, and Mike Wheeler have helped shape my thinking about matters embodied, embedded, enacted, and extended. Their influence has no doubt found its way into this book in a variety of ways. My thanks to all.
Tony Chemero and Mike Wheeler were kind enough to read an earlier draft of this book, and made a number of very helpful suggestions which significantly improved the final version. The mistakes that remain are entirely my fault.
At MIT Press, my thanks to Tom Stone for getting this project off the ground in the first place, and to Marc Lowenthal and Philip Laughlin for seeing it through to completion. Finally, my thanks to Judy Feldmann, for faultless copyediting (as usual).
1 The Expanding Mind?
There is a new way of thinking about the mind and things mental that has started to seep out of the ivory tower and set up residence in popular consciousness.' Actually, to call it a new way of thinking about the mind is not entirely accurate. It is an old way of thinking about the mind that has taken on new form. Previously the preserve of a few scattered, and distinctly renegade, philosophers and psychologists spread out over the centuries, this way of thinking about the mind has started to acquire what many would regard as more persuasive credentials. This is because it is now emerging, in a reasonably consistent and recognizable way, from the confluence of various disciplines in cognitive science, including situated robotics and artificial intelligence (Webb 1994; Brooks 1994; Beer 1995), perceptual psychology (O'Regan and Noe 2001; Noe 2004), dynamical approaches to developmental and cognitive psychology (Thelen and Smith 1994), and cognitive neuroscience (Damasio 1994).
Some people think-and, for what it's worth, I am one of them-that upon this new way of thinking about the mind will be built a new science of the mind. The new science in question will employ different methods for studying the mind, and will supply explanations of mental processes that are, at least in some ways, quite different from traditional accounts. But these transformations in methods and forms of explanation are just symptoms of something far deeper and more important. Fundamentally, the new science would be new because it is underwritten by a novel conception of what sort of thing the mind is. The subject matter of this book is not the mooted new science of the mind but the conception of the mind that underlies it-a conception that can, I think, be evaluated independently of whether anyone actually gets around to making a science of it.
Traditional attempts to study the mind are based on the idea that whatever else is true of mental processes-perceiving, remembering, thinking, reasoning, and so on-they exist in brains. Mental processes are either identical with brain processes or exclusively realized by brain processes (see box 1.1). The word "traditional," here, is slightly idiosyncratic. The scientific study of the mind is not much more than a hundred years old, and in that time it has undergone several significant transformations: introspectionism, gestalt psychology, behaviorism, and finally, from the early 1960s onwards, cognitive science. Cognitive science, in its traditional form, is based on the idea that mental processes-specifically cognitive processes, for these are the purview of cognitive science-are abstract "programs" realized in the "hardware" of the brain (an analogy with computers guided much of the early work in cognitive science). The principal tasks of cognitive science are, accordingly, to identify the programs (cognitive psychology) and work out how these programs are implemented in the brain (cognitive neuroscience). For reasons that will be fully explained shortly, I am going to refer to cognitive science, in its traditional form, as Cartesian cognitive science.
Cartesian cognitive science is, in many respects, a broad church. There are many important differences in the way cognitive science has developed over the years. For example, in early cognitive science, the emphasis was very much on the "programs" or cognitive "software"; early cognitive science understood itself as engaged in the task of providing abstract formal descriptions of cognitive processes. However, from the mid-1980s on, this emphasis gradually gave way to a renewed emphasis on "hardware" in the form of connectionist or neural network approaches: approaches to understanding cognition based on neurally realistic models of its underlying architecture.' Neural network models base their accounts of cognition on a hardware that is explicitly (if roughly) modeled on the brain (Rumelhart, McClelland, and the PDP Research Group 1986). It is not clear that these two approaches are incompatible. It may be that neural network models are merely accounts of how the more abstract formal descriptions of cognitive processes come to be implemented in the brain. However, nor it is clear that these two approaches are compatible: it may be that neural network models have properties that preclude their being described at a more abstract level by formal descriptions of the sort typically employed.'
We do not need to worry about the details of this dispute. What unites these differing faces of Cartesian cognitive science is an unquestionedindeed seemingly banal-assumption: whatever else is true of mental processes, whether they are abstract formal processes or patterns of activity in a neural network (or both)-they are processes that occur inside the head of the thinking organism. Cognitive processes-the category of mental processes with which cognitive science is concerned-occur inside cognizing organisms, and they do so because cognitive processes are, ultimately, brain processes (or more abstract functional roles realized exclusively by brain processes). It is this unquestioned assumption that makes Cartesian cognitive science Cartesian. And it is this assumption that I sha
ll try to undermine.
The new way of thinking about the mind is inspired by, and organized around, not the brain but some combination of the ideas that mental processes are (1) embodied, (2) embedded, (3) enacted, and (4) extended. Shaun Gallagher has referred to this, in conversation, as the 4e conception of the mind.4 The idea that mental processes are embodied is, very roughly, the idea that they are partly constituted by, partly made up of, wider (i.e., extraneural) bodily structures and processes. The idea that mental processes are embedded is, again roughly, the idea that mental processes have been designed to function only in tandem with a certain environment that lies outside the brain of the subject. In the absence of the right environmental scaffolding, mental processes cannot do what they are supposed to do, or can only do what they are supposed to so less than optimally. The idea that mental processes are enacted is the idea that they are made up not just of neural processes but also of things that the organism does more generally-that they are constituted in part by the ways in which an organism acts on the world and the ways in which world, as a result, acts back on that organism. The idea that mental processes are extended is the idea that they are not located exclusively inside an organism's head but extend out, in various ways, into the organism's environment. We shall examine each of these ideas in much more detail in the following chapters. These characterizations are very rough, and in several ways inadequate; but they will probably give us enough to work with for present purposes.
Each of these ideas-embodiment, embeddedness, enactedness, and extendedness-has been understood as denying, or at least questioning, the central assumption of Cartesian cognitive science: mental processes are identical with, or exclusively realized by, brain processes. It is not clear, however, whether it is correct to understand all of the ideas in this way. I shall argue, later in this chapter and in more depth in chapter 3, that not all strands of 4e are equally anti-Cartesian. Moreover, even if it were true that all of these ideas deny the central assumption of Cartesian cognitive science, this denial would take a quite different form in each case. And once we venture beyond the simple denial, we shall see that the ideas of embodiment, embeddedness, enactedness, and extendedness are far from equivalent. Indeed, not only are these ideas different, some of them may actually be incompatible with the others. Indeed, at least one of the strands of 4e can, and indeed has, been employed as a sort of Cartesian fifth column-one designed to acknowledge the force of anti-Cartesian arguments but also strictly limit their scope. In chapter 3, we shall begin the process of tidying these ideas up: working out exactly what each one claims, the ways in which it denies (if it does) the central assumption of Cartesian cognitive science, and thereby working out which ideas are and which are not mutually compatible.
Box 1.1
Identity and Exclusive Realization
For the purposes of this book, the difference between identity and exclusive realization is of no real importance-that is why I have hitherto used them in the same breath. But it is probably time to explain why the difference is of no real importance. First of all, identity itself can actually be understood in two ways. Broadly speaking, to say that mental processes are identical with brain processes is to say that they are one and the same thing as brain processes. It is not that there are two things there-mental processes and brain processes-that are correlated; there is only one thing there. However, there are two different ways of thinking about this. According to one, this is a claim about mental and neural processes understood as kinds-or as philosophers like to call it, types-of process (Smart 1959). So, the claim is that kinds or types of mental process are identical with kinds or types of brain process. This is known as a type identity theory. According to this, mental processes are one and the same thing as brain processes in much the same way that water is H20, and as lightning is an electrical discharge to Earth from a cloud of ionized water particles.
There is another-currently more popular-way of understanding the identity theory. According to this, the identity between mental and physical holds between individual instances-or, as philosophers like to call them, tokens-of each kind (Davidson 1970). It is the individual episode of pain that I, a particular person, feel at a particular time (e.g., 4:19 PM on March 21, 2008) that is identical with a particular firing of a brain process (that takes place in me at this time). This is an identity between individual episodes, instances, or tokens, rather than an identity of general kinds or types. This view is known as the token identity theory.
The token identity theory has proved more popular for reasons that connect up with the notion of realization mentioned above. The idea of realization is drawn from the computer metaphor that dominated early cognitive science (Putnam 1960). One and the same program can be run on different sorts of computer-and these computers can, within limits, be built in different ways. Therefore, we can not identify the program with any particular configuration of hardware. But, nonetheless, the program cannot be run without some hardware or other. Therefore, the idea is that though the program cannot be regarded as the same thing as configurations in the underlying hardware-because it can be run on different hardware-the program is, in any particular case, realized by given type of hardware. To say that A realizes B, therefore, is to say, very roughly, that A makes B possible by providing a physical basis for it.
The token identity theory became popular because it is compatible with the idea that mental kinds are functional (see box 2.1) kinds that are, in any given instance, realized by some or other physical hardware-but not necessarily the same hardware in all cases. The idea that mental kinds are functional kinds is the claim that they are best understood in terms of what they do. To see what this means, consider another example of something defined by what it does. A carburetor is a physical object located somewhere in the innards of a car's engine (or older cars anyway-fuel injection systems have replaced them in more recent models). What is a carburetor? Roughly, it is (or was) something that takes in fuel from the fuel inlet manifold, takes in air from the air inlet manifold, mixes the two in an appropriate ratio, and sends the resulting mixture on to the combustion chamber. It is fulfilling this role that makes something a carburetor, and anything that fulfils this role in a car thereby counts as a carburetor. Most carburetors tend to look pretty similar. But this is at best a contingent fact, because it doesn't matter what a carburetor looks like as long as it fills this role. The details of its physical structure and implementation are of secondary importance compared to the role it fills, for it is filling this role that makes something a carburetor, and not the details of its physical structure or implementation. Of course, not everything physical thing is capable of playing the role of a carburetor. A lump of Jell-O inserted into your car engine would have a hard time mixing fuel and air-or doing anything except melting, for that matter. A lump of Jell-O is simply not the right sort of thing for fulfilling the functional role of a carburetor. So, the details of the how the functional role is physically implemented are not irrelevant. But as long as you have a suitable physical structure-one that is capable of fulfilling the role of a carburetor, then it doesn't matter what it is as long as it, in fact, fulfills this role. If it does, then it is a carburetor.
At the level of tokens, each individual carburetor is a physical object. There is a token identity between a carburetor and a physical thing. However, since carburetors can, in principle, be realized by different kinds of object, there is no type identity between carburetors and specific types of physical object. That is, it is not plausible to say that a carburetor as a kind is the same thing as this kind of physical object, because other kinds of physical objects are capable of playing the role of carburetors-and thus of being carburetors. The kind or type carburetor is realized by physical kinds or types, but it is not the same thing as them.
By far the most popular version of materialism over the past few decades has combined an identity of mental and physical tokens with a functionalist account of mental kinds or types. So, the general idea is that me
ntal processtokens are identical with brain process-tokens, but mental process-types are realized by brain process-types. Since, it has almost universally been assumed, there is nothing outside the brain that realizes mental types, this realization is exclusive to the brain. Together these claims-token identity combined with exclusive neural realization-form the default view of the nature of mental phenomena that 4e attacks.
To avoid these difficulties, and to avoid anticipating the results of arguments yet to be given, I shall, for now, simply talk of non-Cartesian cognitive science (or, depending on context, of the non-Cartesian conception of the mind that underlies this science), and understand this as made up of at least some, but not necessarily all, of the strands that make up 4e: embodiment, embeddedness, enactedness, extendedness. The diversity and potential incompatibility of the ideas of embodiment, embeddedness, enactedness, and extendedness do present us with a problem. If a new, non-Cartesian, cognitive science were to be built on these ideas, and if some of these ideas didn't really cohere with the others, then the prospects for our new science would be at best uncertain and at worst bleak. What we need, it seems, is a way of identifying with precision the content of each of these ideas, and, on this basis, putting them together in the best way we can and discarding the ones that cannot be thus assimilated. This, in effect, is where philosophy comes in. It would not be entirely true to say that philosophy has driven the development of the ideas of embodiment, embeddedness, enactedness, and extendedness. It is true that the history of philosophy has contained figures that have developed views of the mind that are entirely amenable to, and may even be (philosophical) versions of these ideas. We shall encounter several of these figures later in the book. However, most of the recent impetus for 4e comes from developments in fields of cognitive science such as situated robotics, developmental cognitive psychology, and theories of visual perception. The primary role of philosophy is not to provide new empirical evidence for nonCartesian cognitive science, but to place this science on a solid conceptual footing.