Review - Reconstructing the Cognitive World
The Next Step
by Michael Wheeler
MIT Press , 2005
Review by Pak Hang Wong
Aug 22nd 2006 (Volume 10, Issue 34)

Until recently, cognitive science (and the philosophy of it) has been very much dominated by the 'Representationalist approach', the basic idea of which, as Wheeler puts it, is that ''there are internal representation... in the cognizer's mind, entities or structures (the representations) that stand in for (typically) external states of affairs'' (p.6). For its acceptability amongst the researchers of the discipline, Wheeler labels this approach to cognitive science as the 'orthodox cognitive science'. However, this predominant approach to cognitive science is presently challenged by a new approach in the discipline, which is known as 'Embodied-Embedded cognitive science' [henceforth, the 'EEC' approach], which takes the brain, body and environment as the integral parts of a cognitive system. And in Reconstructing the Cognitive World, Wheeler attempts to spell out the inadequacy of the orthodox cognitive science, and offer an alternative by drawing insights from the EEC approach. All this may sound relatively familiar, and Wheeler's book may appear as if it were just another book on the arguments against orthodox cognitive science. But, it isn’t so. Wheeler's interesting twist lies in is his defense of the EEC approach, which he argues is Heideggerian in nature. Thus, what Wheeler is trying to do in his book is to articulate and explicate the EEC approach, and to show how Heideggerian philosophy can help to illuminate the underlying principles.

Wheeler's strategy is simple and straightforward. First, by re-investigating the works of Descartes, Wheeler explicates the underlying principles of Descartes' philosophy of mind. And from them, Wheelers draws out eight principles that are at work in Cartesian Psychology (p38/53). After surveying some research from (or, closely related to the family of) orthodox cognitive science, he argues that although orthodox cognitive science has done away the dubious tenets of Descartes' philosophy of mind, e.g. dualism, it remains committed to the eight principles he has identified. Wheeler's main conclusion is the following: as orthodox cognitive science is committed to the subject-object dichotomy, and sees cognitive system as a general purpose reasoning machine which gets its inputs from perception and bodily states, and computes the inputs to generate relevant outcomes (this is often demonstrated by the computer analogyox cognitive science), it will simultaneously commit to the principle of explanatory disembeddedness and disembodiement, and, therefore, the orthodox cognitive science is, by its nature, context-independent.

Then, Wheeler proceeds to introduce the working framework for the EEC approach, i.e. the dynamical systems, which he defines as ''any system in which there is state-dependent change, where systemic change is state dependent just in case the future behavior of the system depends causally on the current state of the system'' (p.91). What's more, Wheeler argues that the computational system, which is generally regarded as the working framework for the representational approach, is actually a subset of dynamical systems. This will be important, as Wheeler's aim is not to overthrow the representationalist approach entirely, but to point out that orthodox cognitive science, by focusing on this approach, has misplaced their explanatory priority of our cognitive abilities.

In the remaining chapters, Wheeler begins his task of defending EEC, and to tell his story of how Heideggerian philosophy helps to illuminate the underlying principles of this approach. Wheeler begins by embracing the possibility of collaboration between Heideggerian philosophy and cognitive science. In the course, he unfolds the Heideggerian philosophy for human science, which consists of two inter-related spectrums of explanation: (i) the constitutive explanation, which aims to identify and clarify the constitutive character of human agency; and (ii) the empirical explanation, which is about the human agent's relations to the character. Although the two spectrums of explanation cannot be reduced to one another, a paradigm shift in either of them is going to affect the other. After that, Wheeler explains the core of Heideggerian philosophy that is relevant to cognitive science, i.e. the modes of encounter. For Wheeler (and Heidegger), there are three modes of encounter between agents and various entities in the world: (i) readiness-to-hand, (ii) un-readiness-to-hand, and (iii) presence-at-hand. Readiness-to-hand describes the encounters of which an entity is equipment for our use, and such use of the equipment is hitch-free; the term 'smooth coping' is used to describe circumstances in which our use of equipment for our ends is hitch-free. In 'smooth coping', Wheeler claims that there is no room for the subject-object dichotomy, as the agent is unaware of any object with determinate properties, and there is only the experience of the on-going task. When will an agent become aware of the determinate properties of the entities then? According to Wheeler, it is only when one takes a theoretical attitude towards the entities by investigating them out of the context [as in scientific investigations] that one will become aware of the properties. But (i) and (iii) are the two extreme modes of encounter in Heideggerian philosophy, the mode an agent is normally in is something in-between, i.e. 'un-readiness-to-hand', which happens when 'smooth coping' is disturbed, and the agent has to become a practical problem solver.  Although not as pure as 'readiness-to-hand', 'un-readiness-to-hand' remains inclined towards 'readiness-to-hand', but not 'presence-at-hand', as an agent does not attempt to investigate the entities out of a specific context, and it remains as 'an equipment' (or, a search for an equipment) for the agent's purpose. It is this 'un-readiness-to-hand', Wheeler argues, that constitutes our everyday cognition. And it is clear that, if the Heideggerian is right, then cognitive science is by its very nature context-dependent. Still, the question remains: what will the Heideggerian’s context be? Wheeler provides us an answer by spelling out Heidegger’s world. In a nutshell, 'world' is the totality of happenings in which an activity took place; and, a better label for this notion of world is 'background' of/for an activity (p.147). As the Heideggerian world is intended to describe the agent's background against the specific purpose she is going to achieve by her activity (or by her use of equipment), the question concerning normativity naturally follows. It is here, Wheeler suggests, that the importance of culture and biological selection in constituting Wheeler's Heideggerian world. Here, Wheeler's exposition of Heideggerian philosophy and its relevance to cognitive science is, indeed, very interesting. However, he does not seem to have offered any positive argument for his Heideggerian claim that cognitive science should prioritize everyday cognition over other cognitive abilities, except by citing Heidegger. One may, of course, happily admit that it is an important project to figure out how everyday cognition works, but, nonetheless, refuse to join Wheeler in thinking that it has its priority over the representation-involving cognition.

Having explained the modes of encounters and the world, then the introduction of Heideggerian philosophy which is relevant to cognitive science Wheeler moves on to draw comparison between the Heideggerian framework (or EEC approach) and the Cartesian framework, and see which one fares better in the discipline in question. Wheeler starts by examining Heidegger's critique and Dreyfus's critiques of the Cartesian view. After showing the failures of Heidegger's critique, and some of the Dreyfus's view, Wheeler illustrates what, he thinks, is the most serious problem for cognitive science, i.e. the frame problem, which Wheeler has borrowed from Fodor's definition, that ''the problem of putting a 'frame' around the set of beliefs that may need to be revised in the light of specified newly available information'' (p.178). The problem for orthodox cognitive science, as he argues, is that it relies on the context-independent representation, but an agent faces dynamic and changing contexts in everyday cognition; context-independent representation is, by its nature, impoverish to handle everyday cognition.

So, if the context-independent representation cannot do any good to cognitive science, what else can? Wheeler introduces 'action-oriented representation', which is ''determined by the fact that they (the representations) figure in a process in which what is extracted as contextually relevant is fixed by the needs and the previous 'experiences' (...) of a (mildy) intelligent agent, acting in its particular environment, and given its distinctive navigational 'project''' (p.197), as an alternative to orthodox cognitive science. But then, drawing evidences from research in cognate disciplines, e.g. evolutionary robotics, he identifies a phenomenon which is known as (non-trivial) causal spread. Briefly, causal spread describes those phenomena which ''turn[s] out to depend, in unexpected ways, on causal factors external to the system'' (p.200). In cognitive science, this means that intelligent behavior is dependent on bodily and environmental factors. This paves the way for Wheeler to set up his Heideggerian framework of cognitive science, which puts great emphasis on the world, i.e. the background [of an activity] in which the agent is involved. He, then, provides us with a detail analysis of how Heideggerian philosophy can shed light on the EEC approach.

But, it is not enough merely to recover the Heideggerian principles within the EEC approach. For Wheeler, it is more important, as the title aptly suggests, to reconstruct cognitive science by proving that a Heideggerian framework is the proper one for the discipline. How can Wheeler prove this? He attempts to do so by showing that a Heideggerian framework can handle the frame problem. As the frame problem appears to be a deadlock for any framework that is Cartesian in nature, if Wheeler succeeds in showing that a Heideggerian framework can handle the problem then it will be superior to the Cartesian, and a paradigm shift may possibly occur. Yet, in handling the frame problem, Wheeler's strategy is not to provide a reply to the problem, but to dissolve it before it arises. He claims that once the researcher does away Cartesian biases and shifts to a Heideggerian framework, the frame problem no longer arises. Wheeler offers a 3-step plan for this reconstuction: (S1) refuse to take the seductive Cartesian route from the manifest context sensitivity of intelligence to the need for detailed inner representations; (S2) provide an enabling explanation of thrownness; and (S3) secure adaptive flexibility on a scale sufficient to explain open-ended adaptation to new contexts, but do so without a return to Cartesian representation-based general-purpose reasoning. These three steps, as Wheeler explains, are available once we drop the Cartesian framework for everyday cognition, as the world [and thus the context] is necessarily built-in to an explanation for intelligent behavior, and, therefore, it is no longer necessary to find a way to bring in the context-dependency of everyday cognition [it is also no longer necessary to accommodate the role of bodily and environmental factors via inner, context-independent representation]. In this way, then, the frame problem no longer emerges. However, as I understand the frame problem, i.e., as a problem for empirical investigation, or as a working mode of Artificial Intelligence, it is a problem of specifying how an agent correctly collects and represents the necessary information, not for a particular task, but for general behaviors in everyday life. Of course, Wheeler's action-oriented representation fits perfectly to explain an agent's intelligent behavior in a specific task, and is relieved from the need to capture the context in which an agent's intelligent behavior is taking place, as the action-oriented representation itself specified and limited the range of information.. But, Heidegger's world, the context, is arbitrary and ambiguous, and, as Wheeler sees it, one cognitive task is essentially connected to another cognitive task and so on. The world, or the context, then will become enormously big for a working model of cognitive science. Thus, those who favor Wheeler's action-oriented representation must admit that either it only works in laboratory-controlled circumstances, but not in everyday cognition, or a better definition of the world, or context, has to be given before we can evaluate its success.

By taking cognitive science and Heideggerian philosophy side by side, I think Wheeler's book is an interesting and thought-provoking one, relevant to cognitive scientists, philosophers of mind and cognitive science and philosophers who work in the Continental tradition. Although I do not agree with his view on the frame problem, I think Wheeler has raised many important questions for those who work in orthodox cognitive science. Furthermore, I believe that Wheeler's book will be an important one, as it opens up the conceptual spaces for a new paradigm for cognitive science.

© 2006 Pak Hang Wong

Pak Hang Wong was formerly an Assistant College Lecturer at The Community College at Lingnan University. His research interests are in Epistemology, Philosophy of Language, and Philosophy of Mind.