Summary
We present conceptual relationships between Aphasiology and Artificial Intelligence, which are relevant for understanding our cognitive functions. On one hand, language pathology provides indeed an irreplaceable source of data on linguistic cognitive functions (normally hidden); modeling these functions helps in developing "intelligent" A.I. programs, e.g. for handwritten recognition. On the other hand, such modeling leads to revisit both pattern recognition and language understanding classical approaches, and to explore new answers to the fundamental question: what does it mean to recognize or to understand?
Introduction
Neuropsychology (and more precisely Aphasiology, the study of the language pathologies occurring after brain lesions) is able to provide some insight into inthe properties of our normal cognitive system. Brain lesions could indeed be viewed as experimentation in vivo on this system, enabling namely --- namely, out of patients linguistic behaviors -- detection of to detectspecific functions, normally embedded in the complexity of the whole redundant cognitive system, and therefore indecipherable.
On the other hand, Artificial Intelligence (A. I.) approaches may also provide some insight intoon the functional properties of this complex system. The pertinence of using computational concepts to study cognitive linguistic mechanisms derives from the fact that to deal with natural language, both automatic processing and cognitive mechanisms must handle a set of common problems -- those inherent to our language. Therefore, an A.I. system dealing with natural language represents somehow a kind of experimentation in vitro on the functional characteristics of language.
Return to the content of this issue
Return to the Cybernetics and Human Knowing Homepage