On the Use of Reflection Principles in英文文献.pdfVIP

On the Use of Reflection Principles in Modelling Complex Reasoning Jan Treur* Programming Research Group, Expert Systems Section, University of Amsterdam, 1098 SJ Amsterdam, The Netherlands Complex patterns of reasoning as involved in diagnosis, process and robot control, and design can be formally described by interacting basic inference processes. Reflection principles are used to give a formal description of some types of interactions between basic inference processes. Thus, complex reasoning is modelled in a modular form. I. INTRODUCTION Complex tasks for expert systems, such as diagnosis or process and robot control or design, often show a complex pattern of reasoning involving a num- ber of more primitive patterns of reasoning. For example, certain diagnostic tasks can be viewed as an alternation of two kinds of reasoning: on the one hand the reasoning to provide a choice of hypothesis and on the other hand the reasoning to test this hypothesis. Continuing this analysis of splitting a complex task into more primitive inference processes provides a number of processes which are atomic, that is, which cannot be split in a natural manner any further. We call these atomic inference processes basic. Basic inference processes show a uniform kind of reasoning, in a sense explained below. Testing a hy- pothesis is such a basic inference process. A logical description of testing a hypothesis as a basic inference process in diagnostic reasoning is given in Treur. This description can be extended to other basic inference processes involved in diagnostic reasoning as is shown below. Our logical description of diagnostic tasks can be extended as well to other complex tasks, such as process and robot control and design. These areas are being investigated; some first reports are given in T r e ~ r .The