rapaport@CS.BUFFALO.EDU (William J. Rapaport) (03/08/88)
STATE UNIVERSITY OF NEW YORK AT BUFFALO BUFFALO LOGIC COLLOQUIUM RANDALL R. DIPERT Department of Philosophy SUNY Fredonia THE INADEQUACY OF THE TURING TEST AND ALTERNATIVES AS CRITERIA OF MACHINE UNDERSTANDING: Reflections on the Logic of the Confirmation of Mental States In this paper, I address the question of how we would confirm a machine's, or any entity's, "understanding". I argue that knowledge of the internal properties of an entity--as opposed to "external" proper- ties and relations, such as to a linguistic or social community, or to abstract entities such as propositions--may not be sufficient for the justified attribution of understanding. I also argue that our knowledge of the internal construction or of the origin of an artificial system may serve as defeating conditions in the analogical reasoning that oth- erwise supports the claim of a system's understanding. (That is, the logic of the confirmation of understanding is itself non-monotonic!) These issues are discussed within an analysis of the complex fabric of analogical reasoning in which, for example, the Turing Test and Searle's Chinese Room counterexample are merely examples of larger issues. No previous contact with the logic of analogy, artificial intelligence, or the philosophy of mind (other than having one) is assumed. [Shorter summary: Will we (ever) be able justifiably to say that an artificial system has "understanding"? Probably not.] Tuesday, March 15, 1988 4:00 P.M. Fronczak 454, Amherst Campus For further information, contact John Corcoran, (716) 636-2438.