Dr. Mike Dawson's Psychoquium Talk Abstract
November 1995
Classical cognitive scientists have been primarily inspired by the Turing machine, which provides a complete and powerful description of what it means to be doing "information processing". But Alan Perlis writes "Beware the Turing tar-pit, where everything is possible, but nothing is easy." Connectionists have heeded this warning, and have tried to endorse a qualitatively different notion of "information processing". This move, too, has been fraught with problems. In particular, they have become mired in the Connectionist goo of their networks, which prevents network interpretation, and thus diminishes the impact of Connectionist theories.
My students and I have rashly entered this fray by working on methods for interpreting connectionist networks. In this talk, I will describe how cognitive neuropsychology can be informed by such techniques, by talking about the relationship between another sticky problem -- the "locality assumption" -- and lesions applied to an interpreted network?
Do we come out with sticky fingers? Let the audience be the judge!
Dr. Dawson is a professor in the Deprtment of Psychology at the University of Alberta.
Back to 1995-1996 Psychoquium Speakers page.