LIBRA - A HIGH-FIDELITY KNOWLEDGE-BASED SYSTEM FOR RECOMMENDING SCHEDULING PROGRAMS

Contemporary knowledge systems are usually neither easy to use or unde rstand. We propose a new class of knowledge systems, which we call hig h fidelity systems, that use representations and processes that closel y resemble the objects and tasks with which people are already familia r. To illustrate the concept of high fidelity, we describe Libra, one such system for recommending scheduling programs. Our initial experien ce with Libra and its usage suggests that high fidelity not only enhan ces the user acceptability of knowledge-based systems, but it also fac ilities high speed design and implementation. We chose to replicate th e knowledge in Libra because it is already published, and because we w anted to show that a Libra user can have an interface similar to the d iagrams drawn in the journal paper. The latter was used to describe th e knowledge-based framework for selecting scheduling systems. If we ca n replicate them in the user interface, we would have successfully cre ated a high fidelity system - that is, the Libra user interface would be at least as able to communicate the framework to the user as the jo urnal paper. (C) 1997 Elsevier Science Ltd.