Book Reviews
411
Representations, Targets, and Attitudes. ROBERT CUMMINS. Cambridge, MA (1996). 153 pp. ISBN: O-262-93252-X. $25.00.
MIT Press (A Bradford
Book),
Representational Ideas: From Plato to Patricia Churchland. RICHARD A. WATSON. Dordrecht, Netherlands: Kluwer Academic o-7923-3453-1. $99.00
Publishers
(Synthese
Library,
Volume 250) (1997).
172 pp. ISBN:
One of the natural components of the cognitive viewpoint in information studies is the literature of epistemology and the philosophy of mind. This literature seeks to understand enigmas such as the mind/ body problem, the nature of thought and knowledge, and mental representation. The implications of this body of thought go beyond philosophy, to fields such as artificial intelligence, cognitive psychology/ science, and linguistics. Philosophers of mind engage in what may be called the final, boundless frontier on Earth: the mind seeking to understand itself. In particular, as Cummins (1996) sums it up, the topic in the philosophy of mind for some time now has been to explain the relevant notion of mental representation . . (p. 1) Philosophy of mind impacts information science most in its discussion of representation. Intuitively, we can say that there should be some relationship between a persons mental content with respect to an information problem, and how he or she evaluates documents retrieved from an information retrieval system (or, for that matter, how he or she interprets and indexes documents). For example, a user may engage in the act of comparing information retrieval (IR) system output (lists of documents) with some internal (i.e. mental) notion of the problem at hand when s/he makes relevance judgments. Two recent books provide opposing arguments for theories of representation. Of particular interest to IR research are the issues surrounding the representation and causality. Neither Watson nor Cummins are information retrieval researchers and therefore, any correlations between what they write and IR, have been drawn by the reviewer. Watson is an authority on Descartes, but he does not support a rehashed version of the non-extended mind engaged in interpretation of vibrations of the pineal gland. He is well aware of advances in neurophysiology, especially those of Patricia Churchland. Watsons criticism of materialist notions of mental representation is based mainly on the failure of materialists to specify what constitutes an idea. Specifically, one of the primary characteristics of an idea is ones awareness of it. If, as materialists suggest, ideas are no more than patterns of neural events, then what function of the brain is aware of such events? Certainly, these neural patterns may bear resemblance to the objects they represent, but no neurophysiologist has yet explained how this awareness may take place within the brain. Watson recommends that ontological (metaphysical) considerations be present in discussions of the nature of ideas. In addition, Watson maintains a causal stance on representation. That is, representations are caused by events and objects in the world. It is here that Watson reveals his sympathies with Hume. Cartesians did not believe it necessary that representations either resemble, or be caused by, events or objects they represent. Cummins, on the other hand, is uncomfortable with the idea that representations themselves play an active role in cognitive processes. This is not to say they play no part. They do. However Cummins maintains that a representations content may be used by applications (i.e. attitudes, beliefs, desires, etc.) to interact cognitively with targets (objects in the world). It is difficult to tell, but Cummins seems to be advocating some notion of innate schemata (representations). That is, there is not a causal relationship between targets and mental representations. If there was a causal relationship between targets and representations, according to Cummins, there would be no explanation of how mental representations could be misapplied to targets. He maintains that attitudes may apply representational content in error, but representational content itself is independent of its use or functional role (p. 86) Cummins is a tough read. He warns of such in the first chapter, stating that this volume is not for the uninitiated. Part of the reason for the density of the book is that he spends the first six chapters arguing against opposing representational theories with the assumption of the readers intimate domain knowledge. This assumption is both a strength and a weakness. Certainly, Cummins is breaking some ground. He argues persuasively for the role of applications and attitudes in representational theory. On the other hand, his writing is so directed at his peers that it becomes, at times, coy. But this is a minor issue. The major issue is his dismissal of causal theories of representation. Take, for example, our situation in IR. A user approaches an IR system with some sort of information problem in mind-a representation of the problem-constructed to some degree of coherence. Our user is able (depending on the completeness of her construction) to form some guery and then observes the results. Cummins attacks causal theories which might interpret our Situation as one in which the user, as a result of observing the results of her search, might change his or her representation of her information problem. The results may contain information that debunks one of the users assumptions about the problem (of course, much of this discourse depends on our assumptions of users intentions: whether they intend to use information to
412
Book
Reviews
confirm what is already known; or whether they actually seek to extend knowledge; but this is a different
subject). In this case, objects and events have had a direct causal effect on a representation. Cummins would counter saying that if objects have a causal link to their representation, then there should be no error. But there is error, and therefore, causal theories are false. Cummins does not speculate on error rate: he is a theorist. Another theorist, Jerry Fodor, counters error arguments with a common sense approach: A meandering river erodes, its outside bank. False or vacuous; so a philosopher might argue. Take it straight-as a strictly universal generalization-and it is surely false. Think of the case where the weather changes and the river freezes; or the world comes to an end; or somebody builds a dam: or somebody builds a concrete wall on the outside bank; or the rain stops and the river dries up . or whatever. You can, of course, defend the generalization in the usual way-by appending a ceteris paribus clause: All else being equal, a meandering river erodes its outside bank. But perhaps this last means nothing more than: A meandering river erodes its outside bank-unless it doesn’t. That, of course, is predictively adequate for sure. Nothing that happens will disconfirm it; nothing that happens could. (Fodor, 199.5,p. 244) In other words, what do we gain in predictive power from an analysis of error? If we can show empirically that, in general, a users ideas about his information problem change as a result of interaction with an IR system, then is the consideration of error in the application of a representation enough of a problem to disregard a causal notion? Watson might say no, but he might be more interested in the ontology of the user. He may ask, What mechanisms in the mind/brain are at work that allow the brain to recognize its own state changes. Watson maintains that no reductionist theory has yet explained the mental functions. If they [materialists, reductionists] claim that having an idea-a state, an organization, an event in the brain-is knowing, they appeal to magic (Watson, 1995, p. 139, brackets added by reviewer) . Therefore, Watson wants explanations of specific physical or mental events that allow a brain to be aware of its own activities. In addition to presenting a comprehensive overview of western thought on the problem of representation, if not specifically on causality, Watson has brought the unthinkable into cognitive studies. For all the achievements by cognitive scientists, they have elected to ignore metaphysical and affective factors in human life. One reason for this oversight is that such factors don’t lend themselves to study in digital environments. Another is that sophisticated, late twentieth century intellectuals find such notions beneath them. It may be time for a reappearance. What value are such efforts to information science? On one hand, we might say they are of little value. We are, by and large, a practical field. We wish to build systems that will seamlessly deliver information to users with precision and recall measures of 1.0. Why should we care about things such as languages of the mind, or whether mental representation is isomorphic or homomorphic? Does knowledge of such issues really translate into better information retrieval systems? However, over the past 20 years or so, information science has experienced a move toward the study of users and how they interact with information systems. This move is best expressed in the literature pertaining to the cognitive viewpoint (e.g., Belkin, 1990). While it is true that some of the issues over which philosophers of mind wrangle are too esoteric to be of much theoretical or practical use for information science, much of the literature should be seriously considered for inclusion in theory building in a field that is wanting in such.
REFERENCES Belkin, N. J. (1990). The cognitive viewpoint in information science. Journal of InformationScience, 16, I I- 15. Fodor, J. (1995). Persistence of the attitudes. In W. Lyons (Ed.), Modem philosophyof mind,pp. 242-271. London:
Everyman. Louisiana State University School of Library and Information Science
DAVID B. ROBINS