AI Seminar
-------------------------------
Tuesday, October 4th, 2005
4:00 pm - 5:30 pm
175 ATL (Large Conference Room)

"Computational and Empirical Explorations of Fast, Complex Cognition: 
Interactive Skill, Language Processing, and Humor Perception"
Prof Richard Lewis
Psychology
University of Michigan

===============================

In this seminar I'll provide an overview and highlights of three ongoing 
research projects that all have the aim of developing detailed 
computational models of fast, complex cognition, and the fixed cognitive 
architecture that supports it.

(1) The first project is about understanding the nature and development 
of skilled human-computer interaction. I'll describe a new approach to 
cognitive modeling, Cognitive Constraint Modeling (CCM), that can be 
used to analytically identify the asymptotic bounds on human performance 
in specific task situations.  CCM derives a detailed schedule of 
cognitive processes via constraint satisfaction over explicitly declared 
performance objective functions (e.g., "go as fast as possible"), task 
constraints, and constraints on cognitive architecture.  This is joint 
work with NASA Ames and Manchester University.

(2) The second project is about understanding the nature of perhaps the 
most intricate and rapid of cognitive skills: real-time language 
comprehension. Our focus is on the fast compositional processes of 
extracting linguistic structure from text, and the working memory that 
supports this. We have developed a process theory based on two key 
principles of human memory---decay and similarity-based 
interference--embodied in an ACT-R computational model that 
incrementally parses sentences and generates reading times. This is the 
first model in psycholinguistics to make single-parameter quantitative 
predictions of reading time across multiple experiments and paradigms. 
This is joint work with Potsdam University.

(3) The third project is about understanding the nature of another 
rapid, complex cognitive skill: appreciating the humor in jokes and 
cartoons. I'll present some results from this new project that attempts 
to gain insight into humor perception using the modern tools of 
cognitive psychology--in particular, high speed eye-tracking and 
pupilometery--along with the 68,647 New Yorker cartoons published since 
1925. This is joint work with The New Yorker magazine.

Yes, there will be cartoons.