![]() |
|||
![]()
|
![]() |
![]() |
![]() |
5. RESEARCH ISSUESMost of the current research issues related to logic and its use in expert and knowledge-based systems focus on either methodologies for using logic methods for induction (learning) of relevant rules (e.g., Muggleton), or on the development of logical reasoning methods that incorporate probability (e.g., Poole, Bacchus, BGHK:AIJ). Regarding the role of logic in learning, the relevant issue is the long-standing concern with what has been labeled the "knowledge acquisition bottleneck" (Feigenbaum and Buchanan). The capture of expert knowledge has always been resource intensive, and learning technology (e.g., Quinlan) has similarly held the promise for easing the bottleneck. The relatively recent development of logic-based inductive techniques (e.g., Muggletonderaedt) has begun to provide results (e.g., Muggleton). The work on the application of logic to learning has been accelerated by simple, robust, and relatively efficient methods of induction. As the acquisition of knowledge has always been a challenge, the rewards for improved automatic acquisition will be significant. Similarly, the role of uncertainty in expert systems has been a fundamental issue at least since the development of MYCIN (Buchanan and Shortliffe). While there have been significant developments in the theory and practice of applied probability in artificial intelligence (e.g., Pearl), more recently the role of probability within a logical framework has provided new promise for exploiting that combination for applications like expert systems (e.g., Poole). Regarding the cominbation of logic and probability, these two previously separate areas now hold promise for providing a foundation for a new generation of knowledge-based systems, where the principled use of statistics and probability is possible within the more abstract reasoning framework of logic. 6. FUTURE TRENDS AND SUMMARYThe future role of logic in expert systems is really a subset of the general role of logic within computing science. Perhaps the simplest analysis is that, despite the relative success of nonlogical methods like neural networks and genetic programming, logic provides the advantage of relatively transparent semantic analysis of knowledge. As the technological basis for the development of complex knowledge-based systems evolves, the importance of verification and validation will grow, and therefore the need for strong tools for formal analysis based on semantics. The general picture of logic portrayed here is a significant component of that future. ACKNOWLEDGMENTSBruce Buchanan graciously provided guidance on the initial threads seeking to connect research in philosophy of science and applied reasoning with the fundamental reasoning structures of expert systems. In this regard, his early work on fundamentals is sometimes forgotten amid the excitement of expert and knowledge-based system performance.
|
![]() |
|
Use of this site is subject certain Terms & Conditions. Copyright (c) 1996-1999 EarthWeb, Inc.. All rights reserved. Reproduction in whole or in part in any form or medium without express written permission of EarthWeb is prohibited. Please read our privacy policy for details. |