Reconstructing the Cognitive World: The Next Step

by Michael Wheeler

In Reconstructing the Cognitive World, Michael Wheeler argues that we should turn away from the generically Cartesian philosophical foundations of much contemporary cognitive science research and proposes instead a Heideggerian approach. Wheeler begins with an interpretation of Descartes. He defines Cartesian psychology as a conceptual framework of explanatory principles and shows how each of these principles is part of the deep assumptions of orthodox cognitive science (both classical and connectionist). Wheeler then turns to Heidegger's radically non-Cartesian account of everyday cognition, which, he argues, can be used to articulate the philosophical foundations of a genuinely non-Cartesian cognitive science. Finding that Heidegger's critique of Cartesian thinking falls short, even when supported by Hubert Dreyfus's influential critique of orthodox artificial intelligence, Wheeler suggests a new Heideggerian approach. He points to recent research in "embodied-embedded" cognitive science and proposes a Heideggerian framework to identify, amplify, and clarify the underlying philosophical foundations of this new work. He focuses much of his investigation on recent work in artificial intelligence-oriented robotics, discussing, among other topics, the nature and status of representational explanation, and whether (and to what extent) cognition is computation rather than a noncomputational phenomenon best described in the language of dynamical systems theory.

Wheeler's argument draws on analytic philosophy, continental philosophy, and empirical work to "reconstruct" the philosophical foundations of cognitive science in a time of a fundamental shift away from a generically Cartesian approach. His analysis demonstrates that Heideggerian continental philosophy and naturalistic cognitive science need not be mutually exclusive and shows further that a Heideggerian framework can act as the "conceptual glue" for new work in cognitive science.

Genres: Philosophy

357 pages, Paperback

First published April 14, 2005

Reviews

[Joshua Stein · Rating 3 out of 5]

Wheeler's Reconstructing the Cognitive World is an interesting approach to contemporary cognitive science. He positions himself as "against Cartesianism" during the early sections, which is unfortunate. Most contemporary writers on the subject take Descartes to be horribly wrong on a number of the important issues, and so it's sort of a silly label to put on contemporary neuroscience. Of course, Wheeler really means that there are certain problematic assumptions about the object/subject division that are carried over from Descartes, but it's a weird way of setting the issue up.

The exposition of Heidegger and contemporary AI is probably the best articulating of this popular Dynamic Systems account I've come across; it's just not as clear about the underlying philosophical baggage or as careful in accounting for the recent philosophical history as, say, Rockwell's Neither Brain Nor Ghost All things considered, Wheeler does a good job at illustrating a very difficult philosophical position, whatever misgivings I have about the choices in the storytelling.

I recommend the book for those who are interested in the embodied cognition account and don't want to slog through the massive text of Lakoff and Johnson's Philosophy in the Flesh . While, philosophically, I prefer Rockwell, I think that most people coming to the cognitive science will appreciate Wheeler's style and writing more than Rockwell, especially those who have some familiarity with Heidegger and Dreyfus to piggyback on in reading the book. Worth a read to those who are already interested in this field.

[James Davies · Rating 3 out of 5]

This is a unique attempt to reframe contemporary cognitive science and challenge some of the field's core assumptions.
I thought Wheeler's initial critique was rather good – a nuanced reading of Descartes shows that it's more tricky to be non-Cartesian than one might think. One can elevate the body and its deep connection to the mind, but Wheeler argues that the fact one is even thinking of the body as being connected to the mind shows that these are still being thought of as two distinct, separate entities, albeit deeply intertwined. Descartes thought that, too.

Wheeler offers good discussions of cognitive science's niggling problem of how to treat the question of context. How do cognitive agents understand context, and effortlessly switch between different contexts? This is, of course, the Frame Problem. Wheeler argues that thinking in terms of representations with discrete, identifiable meanings will cause the Frame Problem to constantly emerge. Representations are not distinct from their context – if we possess them at all, they are thoroughly integrated within their contexts.

I'm concerned about Wheeler's ultimate positive thesis, though. Firstly, it comes across as rather messy. There are strands of half-finished ideas, and there is a consistent paucity of examples, and a corresponding paucity of engagement with cognitive-scientific literature. All we get are a few cherry-picked examples here and there. In particular, two examples – a cricket's auditory system, and a triangle-detecting robot – are supposed to "prove" Wheeler's theses about cognition. On no occasion does Wheeler ask himself whether these primitive systems are really apt analogies for explaining all action-based cognition. Note that these two examples are highly task-specific: they have a single job only (for the cricket, to move towards a mate based on sound; for the robot, to move towards the triangle and away from the rectangle). I felt a little offended that Wheeler thought it was OK to pass off such crude examples as worthy of his task. For it must be remembered that he is trying to mount a challenge against the vast majority of work being performed in cognitive science.
Surely, such an ambitious goal deserves a strong challenge, but throughout it felt rather limp, and that Wheeler was averting his gaze as he skipped around genuine, concerning questions for his whole project. Just what is a representation, then? I felt that he didn't really try to grapple with this question in any serious way. He argues for the possibility of a pure, ready-to-hand state involving no representations. Although he thinks we rarely achieve such a state in practice, nonetheless he carves this out as a distinct, in-principle possibility. For me, even a smidge of anti-representationalism is still anti-representationalism, which is fine, but it must be recognised that this is a radical position, and radical positions demand a radical defence. Unfortunately, no proper defence really eventuates.

But Wheeler is a smart character, nonetheless, and there are plenty of great insights in here. I appreciate his attention to questioning how cog sci is framed. It does seem plausible that our (Cartesian) beliefs do shape the way we do science, and I'm glad that this is being challenged. Ultimately, the positive thesis is a scattered, mixed-bag approach, about whose experimental promise I feel skeptical. I'd say it's worth a read, nonetheless, and is illuminating on some ongoing issues in cognitive science and AI.