Editorial: special issue on verification and validation

As for any software , users of knowledge-based systems (KBS) need to know that they can rely on the system to do its job properly. Assuring the reliability of knowledge-based systems has become an important issue in the development of the knowledge engineering discipline. The processes employed directly to assure the reliability of software are called verification and validation (V & V). Roughly speaking , ␷ alidation is the process of determining if a KBS meets its users' requirements ; ␷ erification is the process of determining if a KBS has been constructed to comply with certain formally-specified properties , such as consistency and irredundancy. Implicitly , validation includes verification. Verification and validation techniques for KBS have been discussed and debated in workshops at many of the predominant artificial intelligence conferences in recent years. The purpose of this special issue is to provide ''snapshots'' of the current state of the V & V area for KBS , by collecting together representative works from three of the most recent workshops : These workshops succeeded in highlighting many of the significant issues and trends within their area of concern. These issues and trends are reflected in the articles selected for this issue , the authors of which have expanded and updated their original workshop papers. The purpose of this introduction is to highlight some of the issues and trends in KBS V & V , to put this collection in its context. In recent years , the main technological theme in the KBS validation area has been the development of tools for automatic verification of knowledge bases. Within this sub-area , the dominant concern has been with the ''first generation'' type of rule-based systems , and the verification has been aimed at detecting anomalies — such as subsumed or conflicting rules—which are symptomatic of logical faults in the knowledge base. The paper by Murrel and Plant of fers a mature view of this kind of verification , wherein the rule base is modelled using graph theory , and the anomalies are defined in graph-theoretic terms .