Note: Date change from October 26 to November 16!
Professor Rick Lewis (Psychology, UM) will deliver our colloquium today. For more on Rick, see the news article here. The title and abstract of his talk are given below.
Towards Deriving Competence and Performance
The purpose of this talk is to illustrate an idea for developing theory in cognitive science: our capacities for language, thought and action (and their expression as behavior) can be understood as solutions to optimal control problems that are bounded by information processing constraints. That is, rather than invoking cognitive bounds as limits on the approach to rational thought and action, we use bounds as part of the definition of rationality. The idea has roots that go back at least to work on signal detection theory and bounded rationality in the 1950s. We consider the application of the idea to two problems in (psycho)linguistics: deriving the control of eye-movements in reading, and deriving properties of grammar. Both pieces of work are in early stages, but in each case it is possible to see how assumptions about processing bounds are a source of explanatory and predictive power. The examples also illustrate how the approach spans levels of abstraction, from milliseconds in reaction time to abstract properties of grammar, such as the presence of compositionality and local ambiguity. I will end on a speculative note by relating the approach to Chomsky's "Strong Mimimalist Thesis", advancing a (probably controversial) reason for why it is a promising approach to explaining language: it abstracts over possible trajectories of biological and linguistic evolution in a powerful way, by focusing on just those constraints on such trajectories about which we have the most hope of making informed conjectures---the structure of mental architecture.