In the preface to this collection of essays, Scheutz offers a straightforward definition of computationalism as “… the view that mental states are computational states…” (ix). He then nods to the history of computationalism in cognitive science before mentioning the philosophical criticisms of functionalism (computationalism's conceptual progenitor) and some of the alternative accounts of cognition posed by connectionists and dynamicists. At this point, the reader might be tempted to catalog this book as just another volume in the philosophy of mind, but that would not do justice to the wide range of ideas presented. Rather than argue for this or that view of mind qua computer, Scheutz has assembled six essays (plus one introductory essay and an epilogue) that examine the notion of computation itself—from its historical roots in centuries-old ideas about mechanism, to a detailed examination of the intuitions that underlie the formal foundations of computing, to discussions of how symbols acquire their meaning and whether a computer must manifest a genuine sense of responsibility in order to possess genuine intentionality. While Scheutz is clearly interested in outlining the “next generation” of computationalism as it relates to questions about mind, this material will also prove illuminating to the philosopher of science who is interested in understanding how ideas about computation have been co-opted, sometimes uncritically, by a variety of traditions.

Consider the tangle of computational intuitions that connects cognitive modeling, cognitive psychology and artificial intelligence. For example, a proponent of the symbolic approach to cognitive modeling celebrates the fact that the production systems underlying most cognitive architectures are, “… Turing complete and thus can compute anything that is computable… thus, they should be flexible enough to model the wide array of computations performed by the human cognitive system” (Byrne Reference Byrne, Jacko and Sears2003). Conversely, the “naturalistically” inclined cognitive psychologist rejects the rational-choice models of decision making in favor of descriptive models that, on their face, seem to resist computational representation (cf. Klein Reference Klein1998). Finally, in the context of artificial intelligence, the reticent will admit to ambivalence about the nature of their work (cf. Kolodner Reference Kolodner1993), while others will fail to see the difference between arriving at the right solution and arriving at a solution in the right way. To them, cognitive modeling simply seems like a waste of time.

In chapter 3, Copeland begins to untangle some of these intuitions by introducing a distinction between “narrow mechanism” and “wide mechanism”—roughly, a distinction between the belief that the mind is a information-processing machine of the sort that can be simulated by a Turing machine as opposed to the belief that the mind is a machine but of a sort that cannot be simulated by a Turing machine. Copeland fleshes out this distinction historically, first reminding the reader that the Turing machine was an idealization of the uninspired cleric, the human computer, who calculates by rote using pencil and paper. In this context, Copeland argues that, “Turing's (and Church's) discovery was that there are limits to what human computer can achieve… ” (66; emphasis in the original) and he maintains that such limits need not apply to mechanism generally. Copeland then points to Turing's oracle-machine as an example of a machine that cannot be simulated by a Turing machine. Although one might quibble about the ultimate significance of an oracle machine (it's one thing to consider a notional device to underpin a hierarchy of undecidable problems, but another to talk about an oracle machine as an intuition pump to inform our ideas about mechanism), Copeland's discussion should nevertheless give pause to the computationally minded cognitive modeler. Indeed, the presumption that the Turing computability has anything to do with the “wide array of computations performed by the human cognitive system,” is up for grabs. If Copeland is right, it is neither entailed by a mechanistic view of mind, nor is it supported by a careful reading of Turing's own work.

Sloman continues in a similar vein in chapter 4 where he argues for “The Irrelevance of Turing Machines to AI”. Offering another blend of historical evidence and conceptual discussion, Sloman claims that our current understanding and use of computers is divorced from the formal theory of computation and that neither artificial intelligence nor our understanding of cognition depends on that theory. The alternative view that emerges is one of the computer as an engineering artifact with useful features that, “… were not defined in relation to recursive functions, logic, rule formalisms, Turing machines, and so on but rather had to do with using machines to produce and control sophisticated internal and external behavior with a speed and flexibility that was previously impossible for man-made machines” (107). Computers are useful for modeling cognition not because cognition is computational, but because real-world computers, like humans, must support “complex causal interactions” while interacting with other physical systems (89). Conversely, Sloman claims that understanding how to make the computer useful “gave us new ideas for thinking about minds” (109). Scheutz echoes the sentiment in his Epilogue where he discusses computation as “experimental philosophy” (178 ff.). Independent of the formal results that underpin and delimit abstract computation, the working computer provides a ready environment in which to model, simulate, and understand cognitive processes. This view of computing should give pause to the naturalistic psychologist who dismisses a computational representation of decision making for fear that it commits us to a view of decision making as essentially computational. It should also vindicate the cognitive modeler who endeavors to fix an appropriate level of abstraction at which mind and machine can be seen as analogous.

Computationalism goes a long way toward diagnosing and clearing up the tangle of intuitions that exists between cognitive science, artificial intelligence, and cognitive modeling (Agre's discussion of “The Practical Logic of Computing Work” provides an interesting perspective here). Although this is a significant step in its own right, Computationalism also delves into the theory of computing directly. Smith's discussion of “The Foundations of Computing” is a sober examination of what the theory of computing actually delivers. Suffice it to say, Smith's conclusions are both surprising and provocative. Indeed, Smith's work exemplifies the strength of Computationalism, namely, a willingness to look critically at both the theory and practice of a tradition that is otherwise seen as monolithic and certain. As computers continue to exert an ever greater influence on an ever wider variety of subjects, it is refreshing to take in a view of computation that attempts to understand rather than just promote.