Connectionism and the Philosophy of Psychology

Abstract

Human cognition is soft. It is too flexible, too rich, and too open-ended to be captured by hard (precise, exceptionless) rules of the sort that can constitute a computer program. In Connectionism and the Philosophy of Psychology, Horgan and Tienson articulate and defend a new view of cognition. In place of the classical paradigm that take the mind to be a computer (or a group of linked computers), they propose that the mind is best understood as a dynamical system realized in a neural network. Although Horgan and Tienson assert that cognition cannot be understood in classical terms of the algorithm-governed manipulation of symbols, they don't abandon syntax. Instead, they insist that human cognition is symbolic, and that cognitive processes are sensitive to the structure of symbols in the brain: the very richness of cognition requires a system of mental representations within which there are syntactically complex symbols and structure-sensitive processing. However, syntactic constituents need not be parts of complex representations, and structure sensitive processes need not conform to algorithms. Cognition requires a language of thought, but a language of thought implicated in processes that are not governed by hard rules. Instead, symbols are generated and transformed in response to interacting cognitive forces, which are determined by multiple, simultaneous, (robustly) soft constraints. Thus, cognitive processes conform to soft (ceteris paribus) laws, rather than to hard laws. Cognitive forces are subserved by, but not identical with, physical forces in a network; the organization and the interaction of cognitive forces are best understood in terms of the mathematical theory of dynamical systems. The concluding chapter elaborates the authors' proposed dynamical cognition framework. Bradford Books imprint

Neural Machinery and Realization

“…Now we see that those investigations will not benefit from looking back to Marr. 1 See Heil (1992), Poland (1994), Horgan and Tienson (1996), Kim (1998), Shoemaker (2001, Wilson (2001), Gillett (2002), Polger (forthcoming), Shapiro (forthcoming), and Endicott (forthcoming).…”

Marr's Levels Revisited: Understanding How Brains Break

“…A Wilson–Cowan type model is a system of nonlinear differential equations that represent the activity of multiple populations of interacting neurons (Wilson & Cowan, ). (Horgan & Tienson, , , give us one way of understanding dynamical systems approaches from a Marrian perspective.) The study asks whether a particular system of nonlinear differential equations comprises the algorithm that allows the basal ganglia to implement a brain's computational goal of selecting the most salient action.…”

“…In 1982, David Marr advocated for three independent levels of understanding for any information‐processing device, like the human brain: the level of computational theory, the level of representation and algorithm, and the level of hardware implementation (p. 25). Cognitive scientists and philosophers have discussed the viability of these distinctions since then (see, e.g., Butler, ; Cummins, ; Dennett, ; Egan, , , ; Gilman, , ; Harnish, ; Horgan & Tienson, , ; Kitcher, ; McClamrock, ; Newell, , ; Poggio, ; Polger, ; Pylyshyn, ; Shagrir, ; Sterelny, ; Stevens, ; Verdego & Quesada, ). Now, more than 30 years out, we can ask: What can we learn from contemporary research in computational neuroscience that might shed light on the feasibility of Marr's original proposal?…”

Untitled

“…Any representational scheme that is of interest to cognitive science must, at least to a considerable extent, be compositional (Chalmers 1993, p.306;Hinton 1990, pp.2-3;Pollack 1990, p.78;Smolensky, 1991, p.288;Van Gelder, 1990, pp.355-356). Van Gelder (1990;1991a) and, more recently, Horgan and Tienson (1996) have pointed out that the part/whole relation between complex representations and their constituents is not the only formal relation available for the encoding of causally effective constituent structure. Horgan and Tienson state, correctly in our view, that 'the question is not whether constituents can play a causal role.…”

“…It has been claimed to be unsolvable for classical cognitive science (a.o. Dreyfus and Dreyfus, 1987;Horgan and Tienson, 1996) but easily solved by its main competitor connectionism (Churchland, 1989;Meyering, 1993). The frame problem arises when one attempts to model the human ability to keep track of relevant changes in the environment.…”