Approaches to cognitive modeling

Abstract

This issue of TiCS features two side-by-side opinion articles on two influential approaches to modeling cognition. In the past 30 years, connectionist and dynamical systems approaches have developed models of language, cognition, and development that focus on mechanism and implementation. Within this framework, cognitive functions are viewed as emergent phenomena, grounded in simpler, lower-level processes. Probabilistic models of cognition entered the cognitive science arena more recently, spurred by advances in the mathematics and computer science of probability. Probabilistic approaches adopt a top-down perspective, focusing as a first step on the characterization of the abstract principles that underlie cognitive functions.On pages 348–364, Griffiths et al. [1xProbabilistic models of cognition: exploring representations and inductive biases. Griffiths, T.L. et al. Trends Cogn. Sci. 2010; 14: 357–364Abstract | Full Text | Full Text PDF | PubMed | Scopus (156)See all References][1] and McClelland et al. [2xLetting structure emerge: connectionist and dynamical systems approaches to understanding cognition. McClelland, J.L. et al. Trends Cogn. Sci. 2010; 14: 348–356Abstract | Full Text | Full Text PDF | PubMed | Scopus (131)See all References][2] provide an overview of their respective approaches and their conceptual foundations, and give a flavor of the range of phenomena their modeling efforts have addressed. Each group comments on the companion article, clarifies issues of disagreement, and outlines how progress can be achieved. The exchange is accompanied by seven letters written by experts in the field, who comment on the two target articles from a variety of different viewpoints, hence providing a broader perspective on the core issues the articles address.I hope that this exchange and accompanying commentaries will be informative of where these two approaches to cognitive modeling currently stand, how they relate, and how they differ. The ultimate aim of the exchange is to provide a glimpse into how progress in cognitive modeling can be made, and, hopefully, this debate will stimulate further discussion and research.