Tim van Gelder

Department of Philosophy, University of Melbourne Parkville VIC 3052. tgelder@unimelb.edu.au

Abstract: A survey of the multifarious roles philosophers play in cognitive science.

When the various disciplines participating in cognitive science are listed, philosophy almost always gets a guernsey. Yet, a couple of years ago at the conference of the Cognitive Science Society in Boulder (USA), there was no philosophy or philosopher with any prominence on the program. When queried on this point, the organizer (one of the "superstars" of the field) claimed it was partly an accident, but partly also due to an impression among members of the committee that philosophy is basically a waste of time. Philosophy, they thought, is mostly obscure bullshit that does little to help, and much to hinder, real progress in cognitive science.

As a philosopher, I found this a tad embarrassing. Yet I could hardly deny that philosophy is mostly obscure bullshit. So should we take the hint and quietly retreat to those traditional topics (metaphysics, epistemology, ethics etc.) where nobody much cares what we do? Or should we stick around, protesting (too much?) that philosophy really does have something to offer to cognitive science, and please can we still be part of the club?

This line of thought was obviously getting out of control. It would not be long before my conscience forced me go out and get a real job. Pulling out of this intellectual tailspin was not going to be easy. With elbow on knee and chin on fist I furiously pondered the situation. What, exactly, do philosophers contribute to cognitive science? And are we worth what they pay us?

This was truly my darkest hour. At that moment, however, years of graduate school began to pay off. All that wrestling with Kant and Hegel had turned my mind into a slick dialetical machine. I was able to answer any question and back it up with an appearance of support. Surely any philosopher worth his travel allowance could explain why philosophers deserve travel allowances. And I wasn’t just any philosopher. I was a particular philosopher, identical with nobody but myself, the sole referent of Russellian proper name. If anyone was up to the task, it was surely me.

As the clouds parted and the horizon leveled out, I gradually discerned the outlines of not just one but a whole series of answers, beckoning like golden arches against the evening sky. There is no one thing, but rather many things that philosophers contribute to cognitive science. They play not one but many roles. Philosophers are like balls of string in the kitchen drawer of cognitive science: hairy, round, and hopelessly tangled up, but every now and again they’re just what you need.

If this case is to be made with the clarity and precision for which philosophers are renowned, we must start by defining terms. Who or what is the philosopher? This is trickier than it looks. You can’t identify them by their subject matter, for philosophers will talk about almost anything. Nor can they be identified by their academic affiliation, for many cognitive scientists outside philosophy departments engage in philosophy, and many inside those departments haven’t done any philosophy in years. As it turns out, the best approach is to identify philosophers by their method. Psychologists standardly run and analyze experimental studies, neuroscientists study brains, linguists gather and systematize linguistic data, and computer scientists write programs. Similarly, what philosophers primarily do is argue, though they also rely heavily on conceptual clarification, and to a lesser extent on historical perspective. Of course, other cognitive scientists can and often do use these methods as well; the point is that philosophers specialize in them.

If this is right (which it is), philosophers are uniquely well-qualified for a diverse range of essential roles within cognitive science, including:

The Pioneer. It has been said, of philosophy generally, that it's job is to struggle with those problems that nobody else quite knows how to address. This insight applies to cognitive science in a fairly direct way. Thus, suppose you had the general intellectual skills of the philosopher, but none of the particular skills of the psychologist, the linguist, or the computer scientist. What issues would you be best able to address? Clearly, it would not be specifically psychological, linguistic or computational issues. It would be issues that, for whatever reason, are not amenable to treatment by means of those other specialized disciplines. Now, an important subset of these issues are ones which can, with enough hard work, be turned into questions that can be addressed by specialized scientific techniques. An important role for philosophers, then, is to be the pioneers: to tackle problems that nobody else knows how to handle yet, in the hope of transforming them into scientifically tractable questions.

Historically, the role of philosophers as pioneers in cognitive science is indisputable. Virtually all the major topics of cognitive science—the nature of intelligence, of knowledge, perception, action, imagination, concepts, mental representation, the relationship of thought processes to the brain, etc. etc.—were first opened up and addressed by philosophers. For example, the existence and nature of "the language of thought" —a key theoretical postulate of mainstream cognitive science —was extensively debated in medieval times . If there are now non-philosophical methods of tackling these issues, these methods generally developed out of the prior philosophical treatments and testify to the success of philosophers (whoever they may have been) in dealing with them.

Further, the most basic ideas around which contemporary cognitive science is constructed are all philosophical achievements. For example, the materialist metaphysical stance that is unthinkingly accepted by the vast majority of cognitive scientists—the idea that human thought processes are the operation of a certain kind of physical system—was first proposed, debated and refined by philosophers. The foundational hypothesis of classical cognitive science—that thought is, more precisely, a form of symbolic computation—was a philosophical invention. And even the notion of computation was first carved out by philosophers. Cognitive science owes its existence, in part, to the pioneering efforts of philosophers such as Leibniz, Frege, Russell & Whitehead, and Turing, who are collectively responsible for the fact that we now understand quite well how the orderly manipulation of appropriately structured physical objects can, in a certain way, bridge the apparent gap between the physical world of causes and effects on one hand and semantic and even mental properties such as meaning, truth, and intelligence on the other .

Perhaps philosophers were pioneers in the origin of cognitive science, but don't play any such role now. Is there reason to believe that philosophers play an ongoing pioneering role? Yes; there are plenty of examples of issues within recent cognitive science that were first tackled by philosophers, and where those initial labors were subsequently turned to practical advantage by other cognitive scientists. Philosopher Clark Glymour has pointed out that many developments in machine learning in artificial intelligence have been a matter of adapting and implementing theories of scientific knowledge developed by philosophers (e.g., the Dendral and Meta-Dendral expert systems, which applied Vienna Circle philosopher of science Carl Hempel's theory of explanation and his instance-based approach to hypothesis confirmation). Another quite different example is the more recent upsurge of interest in, and progress on, issues of modularity in cognitive architecture (i.e., the extent to which the human cognitive system is made up of quite distinct modules which independently handle distinct cognitive tasks). This current attention to modularity is due in significant measure to Jerry Fodor's pioneering monograph The Modularity of Mind . In that book he re-opened an important topic that had lain relatively dormant for many years and, in a fiery crucible, heated equal parts of argument, conceptual clarification, and historical perspective to produce a strong new modularity thesis that has become the primary point of orientation for all subsequent work in the area (see, e.g., , ).

The Building Inspector. Descartes once likened human knowledge to a house built on foundations of questionable integrity . Adapting this famous image to cognitive science, we can think of each specific scientific mode of enquiry as proceeding within the terms of a set of basic theoretical and methodological assumptions. These assumptions constitute its foundations, and sometimes need to be examined. The examination proceeds in a number of conceptually, if not temporally distinct stages. First is simply articulating the basic assumptions themselves. The second is inspecting them for various kinds of defects (vagueness or incoherence, inconsistency with each other, outright falsity). The third and most difficult stage is to reconstruct the foundations, if necessary. Now, since these foundations are what makes everyday kinds of scientific inquiries possible, scrutinizing foundations cannot be part of those everyday inquiries; it is a very different kind of intellectual activity. The tools of the philosopher—argument, conceptual analysis, and historical perspective—turn out to be very useful here, though they must be supplemented with the kind of subtle insight that only comes from deep familiarity with the area itself. Consequently, anyone engaged in inspecting foundations must be, if not a philosopher already, at least temporarily setting aside normal scientific methods and pursuits in favor of philosophical reflection; and one of the most common roles that philosophers actually do play is that of inspecting foundations. (Symptomatic of this is the fact that so many articles or books in the philosophy of cognitive science use the term "foundation" in their title: e.g., , .) Extending Descartes' image, we can think of philosophers as the building inspectors of cognitive science.

For example, cognitive psychologists typically assume (a) that the performance of individual human subjects on cognitive tasks such as memorizing is the direct result of the way relevant information is mentally represented and manipulated; (b) that subjects are all basically alike in their mental organization vis a vis such tasks; and (c) that, consequently, one can infer the general structure of that mental organization on the basis of data on performance gained by averaging over trials on many individual subjects. Assumptions like these form part of the foundations of contemporary cognitive psychology; specific hypotheses and experiments within the discipline are constructed upon them, and psychologists standardly pay them little attention as they go about their business. However, it is not obvious that such assumptions are correct or unproblematic, and they deserve to be carefully scrutinized. To some extent the proof of the pudding will be in the eating; the value of the assumptions will be borne out by the empirical success or failure of actual work in experimental cognitive psychology. But for a number of reasons we cannot rely solely on this empirical measure. On one hand, those who accept a certain set of foundational assumptions can always explain away an apparent lack of empirical success by claiming that more research time and effort is still needed. On the other hand, a mode of scientific enquiry can appear quite fruitful in its own terms, while hindsight reveals that this fruitfulness was largely an illusion sustained by unquestioned acceptance of the very assumptions at issue. And even if some mode of scientific enquiry does grind to a halt, indicating severe problems with its foundational assumptions, this tells us only that something was wrong with those assumptions, and not what was wrong. Philosophical reflection is required in order to understand and eliminate the problem.

The Zen Monk. The fact that philosophers can bring some general intellectual skills to issues nobody else can handle is no guarantee that the philosophers will make any progress. What often happens instead is that they get caught up in arcane and irresolvable debates that, as time goes on, become increasingly removed from everyday work in cognitive science. The philosopher becomes like the Zen monk—a figure supported by the community to ponder those imponderable issues that everyone thinks should be thought about by someone, but for which nobody else has time or patience. In theory, the philosopher-monk eventually reaches a state of enlightenment, but unfortunately that enlightenment is necessarily incommunicable to those who have not undertaken the requisite prolonged course of meditation and asceticism.

The Cartographer. The philosopher Wilfred Sellars once characterized philosophy as aiming "to understand how things, in the broadest possible sense of the term, hang together, in the broadest possible sense of the term" . Applied to the philosophy of cognitive science, this means that one role of philosophers is understanding and describing how all the various elements of cognitive science fit together (or conflict, as the case may be). In doing this they produce large-scale conceptual maps of the discipline; we can thus think of philosophers as the cartographers of cognitive science.

Again, it follows straightforwardly from the characterization of the philosopher given above that cartography will be the philosopher’s job. Of course, someone who is normally, say, a psychologist, can also be a cartographer, but in doing so she will not be using any of the distinctive skills of the psychologist, such as experimental design, statistical analysis, etc.; these skills are obviously irrelevant to the task. She will have to fall back on general skills such as argument and conceptual clarification, i.e., on the methods of the philosopher.

This paper is, in its own modest way, an exercise in cartography. But for better examples, one need only look at many of the classics of the philosophy of cognitive science literature. In the last ten to fifteen years numerous book have been written on Artificial Intelligence. Most of these have been technical introductions to AI programming, written by and for computer scientists. The philosopher John Haugeland's book Artificial Intelligence: The Very Idea , by contrast, aims at an understanding of AI in terms of its wider historical and conceptual context. He sets out by sketching the history of certain key ideas and problems, from Copernicus and Galileo through Descartes, Hobbes and Hume. He then spends two whole chapters just articulating what a computer is, in terms of concepts such as formal systems, digitalness, medium independence, algorithms, automation, meaning and interpretation. An overview of different kinds of computer architecture leads into a conceptual survey of various AI research programs: machine translation, heuristic search, microworlds, frames, and an assortment of important potential difficulties: the frame problem, pragmatic sense, imagery, feelings, ego involvement. The point here is the kind of understanding of AI the reader is given. It is not the sort of inside, hands-on knowledge of the technical practitioner; rather, it is an understanding of what AI is in terms of how it fits into a larger framework of ideas and philosophical ambitions.

The Dilettante. Drawing up a good map requires knowing the various regions that it covers pretty well, and when the map is supposed to cover some significant chunk of a domain as vast and heterogeneous as cognitive science, this is a pretty tall order. By practical necessity, philosopher-cartographers almost never know the particular topics that they are trying to fit into the map as well as the specialists on that topic, and as a result the specialists often regard the cartographer as superficial or misinformed. (In the worst case, all relevant specialists will take this attitude; the philosopher becomes something of a dilettante, purporting to know about many things but having no really deep familiarity with any of them.) However, the appropriate response to this inevitable problem is not to stop drawing up maps altogether, but to continue striving to draw up better maps, which means cooperating with all the specialists in order to develop the best way of describing their particular region and how it fits with other parts of cognitive science.

The Archivist. Closely related to the role of cartographer is that of archivist. It is the philosopher who, more than anyone else in cognitive science, is expected to be the repository of accumulated wisdom: to have understood, seen the significance of, and remembered ideas and research programs of the past both good and bad. That philosophers in particular would play this role is just what you'd expect if, as suggested above, a distinctively philosophical way of approaching an issue in cognitive science is that of historical perspective. After all, much of that accumulated wisdom is philosophical work itself, and that which isn't, the philosopher should be aware qua cartographer anyway.

The Cheerleader. As the official repositories of accumulated wisdom, philosophers have acquired a certain measure of both authority and responsibility in determining what counts as a Good Idea. Consequently, when other cognitive scientists believe that their particular line of research has turned out to be Important, they like to have philosophers publicly bestow their official Seal of Approval, and even better, to explain to the world just how Significant that line of research really is. Of course, any philosopher who does see the Significance of that research must be quite perceptive. A pleasantly symbiotic relationship thus builds up between philosophers and cognitive scientists, with each pronouncing on the goodness of the other's work.

This can be put by saying that if the researchers are the players out there on the field of science, philosophers are sometimes the cheerleaders. This is not the place to discuss specific examples; suffice to say that while most major movements in cognitive science have had their share of philosophical cheerleaders, some have done particularly well; these include artificial intelligence, connectionism and computational neuroscience. On the other hand, it may be that some schools of thought, such as ecological psychology, have ended up less prominent in cognitive science than they might otherwise have been due to unfortunate deficiencies in this key area.

The Gadfly. Had Oscar Wilde been in the philosophy of cognitive science, he would surely have thought that the only thing worse than being refuted is being ignored. Thus philosophers often advance positions that are so strongly and provocatively stated that other cognitive scientists feel compelled to respond. This is the philosopher as gadfly. An excellent example is the notorious paper on connectionism by Fodor & Pylyshyn . The authors there stated bluntly, and argued deftly, that a whole new research program was either hopeless or a mere variant on their own "classical" world-view. Cognition, they claimed, is systematic, and connectionist models either fail to explain this, or they must be implementing some computational cognitive architecture of the kind Fodor & Pylyshyn had been recommending for years. Many connectionist psychologists and computer scientists found these charges so galling that they couldn't resist taking up the challenge of refuting the argument. In the process, Fodor & Pylyshyn's paper became one of the most discussed in the philosophy of cognitive science, and a whole new research niche was created.

In fact, the Fodor & Pylyshyn paper can be seen as exemplifying almost all the roles discussed so far. It was pioneering, in bringing to the forefront of attention a previously neglected phenomenon, that of systematicity. It was inspecting the foundational assumptions of whole research paradigms in cognitive science. It was cartographic, in that it began with a high-level sketch of the lay of the land in issues of cognitive architecture. It was archival, in that it likened connectionism to the seventeenth-century empiricist David Hume and twentieth century behaviorists such as Hebb, Osgood and Hull, and claimed to be merely re-presenting thirty-year old arguments against such views. It was highly enthusiastic in its support for one school of thought over another. To its credit, however, the authors were not playing the role of Zen Monk: the arguments it contained were clear, powerful, and highly relevant to other branches of cognitive science.