![[Mitchell book cover]](http://www.cs.cmu.edu/~tom/mlbook.gif)
Abstract: Understanding and analyzing how human beings respond, adapt, and react is a major scientific endeavor. Human beings are inherently complex and how we behave and interact is not easily modeled or quantified. At the same time, real-world decisions are made based on supposed expectations of human behavior. So much of classic computational modeling has focused on exploring physical and biological phenomena which are based on scientific theory. Whereas, in human modeling, socio-cultural factors are a critical component and not readily expressed in mathematical terms. As such, the realm of socio-cultural modeling creates a need to understand how or when to leverage the classical approaches coupled with the need to incorporate such socio-cultural research concepts. Success of the work can only be accomplished through true interdisciplinary and cross-disciplinary research spanning multiple fields including computer science, engineering, mathematics, and the social sciences. In this talk, we discuss the way forward for effective and efficient computational socio-cultural modeling. I will present new work in socio-cultural representation, the capability to infuse such factors into social networks analysis, and new techniques for efficient real-time analysis of human network constructs.
CogSci Colloquium by Jeff Elman, 4/20, 4-5pm, PY 101.
Jeff Elman, University of California, San Diego
Title: The role of event knowledge in sentence processing: Arguments
against a mental lexicon
Abstract: For many years, rules were where the action lay in language research. Words were seen as arbitrary, unsystematic, and relatively uninteresting. Over the past decade, however, there has been increasing interest in the lexicon as the locus of users' language knowledge. There is now a considerable body of linguistic and psycholinguistic research that has led many researchers to conclude that the mental lexicon contains richly detailed information about both general and specific aspects of language. Words are in again, it seems. But this very richness of lexical information poses representational challenges for traditional views of the lexicon. In this talk I will present a body of psycholinguistic data, involving both behavioral and event-related potential experiments, that suggest that event knowledge plays an immediate and critical role in the expectancies that comprehenders generate as they process sentences. I argue that this knowledge is on the one hand precisely the sort of stuff that on standard grounds one would want to incorporate in the lexicon, but on the other hand cannot reasonably be placed there. I suggest that in fact, lexical knowledge (which I take to be real) may not properly be encoded in a mental lexicon, but through a very different computational mechanism.