Validation of knowledge-based systems: a reassessment of the field

• Published in: The Artificial intelligence review Vol. 43; no. 4; pp. 485 - 500
• Main Authors: Batarseh, Feras A., Gonzalez, Avelino J.
• Format: Journal Article
• Language: English
• Published: Dordrecht Springer Netherlands 01.04.2015; Springer; Springer Nature B.V
• Subjects: Artificial Intelligence; Automation; Computer programs; Computer Science; Expert systems; Failure; Knowledge; Knowledge base; Knowledge based systems; Methodology; R&D; Research & development; Research and development; Software; Validation; Validation; Software testing; Knowledge-based systems; Test cases; Lifecycle model
• ISSN: 0269-2821; 1573-7462
• DOI: 10.1007/s10462-013-9396-9

The subject of validation and verification (V&V) of knowledge-based systems (KBS) has been one of decreasing importance in the last decade. Research and development in the field reduced drastically. One of the main reasons is the persistent software challenges and failures. These failures have been categorized in different ways. One initiative however, which most researchers in the field agree upon, is that the only way to eliminate these problems is by rigorously performing V&V. Although there have been vast improvements in the field of V&V methodology, studies indicate that KBS industry still lacks rigorous validation methods. In this paper, we review the most important validation paradigms described in literature for KBS during the years of their fame. Additionally, this article studies the significant methods, aims to reassess these methods in light of recent advances, and propose new future directions for validation of KBS.