Guidelines for the Verification and Validation of AI Systems

Lance Miller

A comprehensive set of V&V guidelines are provided for 28 different AI system development situations. Of these, 27 are based on the combinations of 3 three-valued factors; the phases of the developmental lifecycle (requirements, design, and implementation) the types syste~n components (knowledge structures, high reusable components, and others), and levels of stringent V&V (classes i, 2, and 3, based complexity and integrity factors). A single last guideline is provided for the maintenance phase. For each situation a set of recommended V&V techniques are organized into a suggested V&V plan given for each technique and the target of the technique, the measurements to be taken, the past fail criteria, and literature references. A total of 31 techniques are recommended overall. Detailed guideline procedures are provided for those 18 judged most important.


This page is copyrighted by AAAI. All rights reserved. Your use of this site constitutes acceptance of all of AAAI's terms and conditions and privacy policy.