Explainable expert systems: A research program in information processing

摘要

Our work in Explainable Expert Systems (EES) had two goals: to extend and enhance the range of explanations that expert systems can offer, and to ease their maintenance and evolution. As suggested in our proposal, these goals are complementary because they place similar demands on the underlying architecture of the expert system: they both require the knowledge contained in a system to be explicitly represented, in a high-level declarative language and in a modular fashion. With these two goals in mind, the Explainable Expert Systems (EES) framework was designed to remedy limitations to explainability and evolvability that stem from related fundamental flaws in the underlying architecture of current expert systems.