Unified Theories of Cognition: Chapter 3 Human Cognitive Architecture

 

Allen Newell can safely be called one of the founders of both artificial intelligence and cognitive science. He won the National Medal of Science shortly before his death in 1992 and won numerous other awards throughout his career. Much of his most famous work was done in collaboration with Nobel prize winner Herb Simon.

In this course we are going to run the spectrum of cognition from the firings of neurons which occur in milliseconds, to thoughtfully executed plans which can function on the order of days. This chapter is largely about that time spectrum and what they can show us about human cognition. Combining the properties of neurons, for example, with what we know about the speed of object recognition, significantly constrains the types of algorithms that the brain might be using. Newell argues that the longer the time span, the fewer constraints there are on cognition. This can either be seen as a strong argument for studying neuroscience (as it has more constraints) or a strong argument against it (the constraints don’t matter much when it comes to "interesting" cognition).

Do some thinking about section 3.1 "The Human is a Symbol System." One of the major splits in cognitive science is between symbolists like Newell and "subsymbolists" like many connectionist modelers. In this course we will look at alternatives to both of these positions which attempt to bridge the gap between them. In this context a symbol is a stand-in for something else, much like a variable in a computer program. For example, when I look at a cat my brain might create a "cat" symbol. I can use the "cat" symbol as a stand-in for real world cats in order to do things like answering questions about cats. Symbolists like Newell believe that the brain can process symbols arbitrarily in the same way a computers program can act on variables. Newell’s argument basically hinges upon the fact that the human brain seems to show arbitrary adaptiveness. From Newell’s point of view thinking is just the manipulation of symbols and a cognitive architecture is the mechanism which does the manipulation. Try to think about what an alternative to a symbol system might be like.