Abstract: "The problem of controlling inference is a common one in AI systems. It is particularly acute in decision support systems, which to be effective must not only produce the right answer, but must also produce it at the right time. This problem is compounded in those situations in which...

Full description

Bibliographic Details

Main Authors:	Logan, Brian (Author), Millington, Karl (Author), Smithers, Tim (Author)
Format:	Book
Language:	English
Published:	Edinburgh 1990
Series:	University <Edinburgh> / Department of Artificial Intelligence: DAI research paper 494
Subjects:	Künstliche Intelligenz Artificial intelligence Inference
Summary:	Abstract: "The problem of controlling inference is a common one in AI systems. It is particularly acute in decision support systems, which to be effective must not only produce the right answer, but must also produce it at the right time. This problem is compounded in those situations in which a number of inconsistent alternative solutions must be explored in parallel. In this paper we discuss these problems in the context of a particular implementation, the Edinburgh Designer System (EDS), a design support system for mechanical design problems. We present an approach to the control problem which utilises the context of a task to determine which inferences the system should perform We describe a hybrid architecture, an assumption-based truth maintained blackboard system and its associated context management system which partitions the blackboard into a series of views containing the information relating to a particular design or task. The resulting system integrates truth maintenance and the control of inference within the framework of the ATMS.
Physical Description:	13 S.