Authors: Steven H. Kim
Addresses: Laboratory for Manufacturing and Productivity, Massachusetts Institute of Technology, Cambridge, Massachusetts, USA
Abstract: An intelligent system should learn over time to improve its performance and to adapt to changes in the environment. This paper addresses the nature of learning and knowledge-generation. A rational reasoning strategy based on the popular approach to first-order predicate logic does not allow for changes in the specification of behaviour. According to the rules of inference, the nature of the deductions appears to be closed: in particular, the conclusions follow strictly from the initial knowledge base, regardless of external influences. This paper shows that reasoning is closed under the rules of predicate calculus, but that it must be open under any meaningful interpretation of a creative learning system. This is achieved by incorporating domain-dependent knowledge.
Keywords: intelligent systems; reasoning; predicate logic; inference; predicate calculus; domain-dependent logic; logical closure; domain dependence; creative learning.
International Journal of Computer Applications in Technology, 1989 Vol.2 No.4, pp.228 - 233
Published online: 11 Jun 2014 *Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article