The Fourth Way: A Comment on Halpin's "Philosophical Engineering"

Abstract

It is common these days to distinguish between three kinds of cognitive science or artificial intelligence: classical, connectionist, and (something like) embodied-embedded.Of course, all such attempts at neat-and-tidy categorization are undoubtedly guilty of over-simplification in one way or another.For example, researchers sometimes build models that combine aspects of more than one approach (e.g. when conventional connectionist networks are used as control systems for embodied agents).That noted, however, one method for separating out our three kinds of cognitive science, so as to understand more clearly their basic theoretical commitments, would be to identify, in a very general way, the sorts of machine that each takes to capture the fundamental character of intelligence.If we adopt this strategy, classicism will be defined by the manipulation of symbols using structuresensitive processes, connectionism by unfolding patterns of activity in neurally inspired networks of simple processing units, and embodiedembedded thinking by complete autonomous robots engaged in perceptually guided motor activity.One of the many fascinating claims in Harry Halpin's strikingly original article "Philosophical Engineering: Towards a Philosophy of the Web" (Halpin 2008) is that the Web constitutes a fourth conceptual anchor for the notion of mind as machine.Halpin's view, in short, is that the Web provides a general model of a computational machine that compels us to rethink the notion of representation, while simultaneously radicalizing our conception of cognition through a vindication of the idea that minds may be realized partly by factors located beyond the skin.In this comment on Halpin's article, I shall engage briefly with just some of the issues that confront us once we take this fourth way.With apologies for the immediate whiff of self-centredness, I shall begin by considering an argument from Halpin's paper that responds explicitly to some of my own previous work.I have been known to claim (e.g.Wheeler

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 8 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.