Interactivist Summer Institute

July 22 - 26, 2003

Copenhagen

Integration, Abstraction, and Concept Learning

Wayne Christensen

This talk will explore the role of integration and simulation in concept learning. I begin by criticizing the Laurence and Margolis (2002) model of concept acquisition, which I argue is too weak to explain cognitively interesting forms of concept learning. The model seeks to avoid Fodorian radical nativism by opting instead for a simple bottom up associative account. However it is unable to explain why concept learning is not prone to propagating error as misclassified objects get incorporated into the "concept". To address this issue theories of concept acquisition must explain how error correction is possible, without appealing to prior possession of the concept or its constituents on pain of falling back into radical nativism.

As an alternative I propose a model that draws on evidence from neuroscience, cognitive psychology and situated cognition. The model postulates that integration plays a central role in conceptualization because it involves similarity/difference comparison and promotes abstraction. Suggestive evidence that integration does indeed play a central role comes from neural evidence that concept retrieval is mediated by convergence zones that link perceptual, motor & affect systems (Tranel et al. 1997). One important form of active integration occurs in imaginative simulation, in which executive and parietal system evoke tacit knowledge in the process of exploring possible scenarios. Based on the evidence of Schwartz and Black (1996) I argue that simulative reasoning plays a central role in the articulation of declarative knowledge and thus acts as a mediator between declarative and implicit systems. Extending the account of induction given by Schwartz and Black I claim that automation and altered tacit knowledge also ensues from concept learning, leading to a cyclic process in which altered patterns of behavior drive further conceptualization. This cyclic process helps explain error correction in concept learning, and the fact that conceptualization is continually being shaped by interaction, implicit learning and imaginative processes explains why error correction does not depend only on prior concepts.