Supporting Incremental Knowledge Elicitation in Decision-Theoretic Systems

Interactive and Mixed-Initiative Decision-Theoretic Systems

Volume

Papers from the 1998 AAAI Spring Symposium

Contents

Knowledge elicitation continues to be a bottleneck to constructing decision-theoretic systems. Most knowledge representations for these systems require complete knowledge of the domain before the systems become useable. Methodologies and techniques for incremental elicitation of knowledge in support of users’ current goals is desirable. A primary goal of our research is to develop a comprehensive software engineering, knowledge engineering, and knowledge elicitation methodology for Symbiotic Information Reasoning and Decision Support. To that end, in this position paper we briefly discuss Bayesian knowledge bases, a probabilistic knowledge representation allowing for incomplete specification of knowledge. We describe how Bayesian knowledge bases along with an intelligent interface agent are used in an expert system shell called PESKI to support incremental knowledge elicitation.