Reasoning About Function in Reflective Systems

Abstract

Functional models have been extensively investigated in the context of several problem-solving tasks such as device diagnosis and design. In this paper, we view problem solvers themselves as devices, use functional models to represent how they work, and subsequently employ these models for performance-driven reflective reasoning and learning. We represent the functioning of a problem solver as a structure-behavior-function model that specifies how the knowledge and reasoning of the problem solver results in the achievement of its goals. We view performance-driven learning as the task of redesigning the knowledge and reasoning of the problem solver. We use the structure-behavior-function model of the problem solver to monitor its reasoning, reflectively assign blame when it fails, and redesign its knowledge and reasoning. This paper describes an architecture for reflective model-based reasoning that is capable of a broad range of learning tasks. It also illustrates reflective model-based learning using examples from the Autognostic system, a reflective path planner.