Knowledge Acquisition for Constructive Systems

Over the past ten years, significant progress has been made in understanding how the knowledge acquisition process for classification systems can be automated. But during this period little attention has been paid to the problem of how to automate the knowledge acquisition process for systems that solve problems by constructing solutions. This paper describes SALT, a tool designed to assist with knowledge acquisition for configuration tasks. SALT assumes a problem-solving strategy involving stages of generate, test, backup, modify, and regenerate. It exploits this problem-solving strategy to guide its interrogation of domain experts and to represent the knowledge they provide in a way that insures it will be brought to bear whenever relevant.