Michael D. Rychener
Progress in building systems that acquire knowledge from a variety of sources depends on determining certain functional requirements and ways for them to be met. Experiments have been performed with learning systems having a variety of functional components. The results of these experiments have brought to light deficiencies of various sorts, in systems with various degrees of effectiveness. The components considered here are: interaction language; organization of procedural elements; explanation of system behavior; accommodation to new knowledge; connection of goals with system capabilities; reformulation (mapping) of knowledge; evaluation of behavior; and compilation to achieve efficiency and automaticity. A number of approaches to knowledge acquisition tried within the Instructable Production System (IPS) Project are sketched.