Knowledge-Based Validation Method for Validating Intelligent Systems

May 2002

Proceedings of the Fifteenth International Florida Artificial Intelligence Research Society Conference (FLAIRS 2002)

Volume

Proceedings of the Fifteenth International Florida Artificial Intelligence Research Society Conference (FLAIRS 2002)

All Papers

To diminish the validation load on busy experts, a bi-directional, many-sided explanation typed multi-step validation method (MMBV) was proposed, which can decrease the validation tasks of experts (Tsuruta et al., 2000). This paper proposes the extension of MMBV. In this method, through the validation results and comments of experts, validation knowledge can be automatically acquired as case-based knowledge. Utilizing this knowledge, computers and KEs can validate, mostly without experts. Consequently, this method is effective for solving the bottleneck in acquiring validation knowledge from experts who are busy, and this further reduces validation load of such experts.