VI NTIS Section

notion of irrelevance, either by automatically detecting irrelevance or by being given knowledge about irrelevance, a formal treatment of the notion is required. In this paper we present a general framework for analyzing irrelevance. We discuss several properties of irrelevance and show how they vary in a space of definitions outlined by the framework. We show how irrelevance claims can be used to justify the creation of abstractions thereby suggesting a new view on the work on abstraction. Descriptors: *Artificial Intelligence; *Expert Systems; *Problem Solving; Knowledge Based Systems Identifiers: NTISNASA Section Headings: 62GE (Computers, Control, and Informa- tion Theory-General)

Domain and Specification Models for Software Engi-

Iscoe, N.; Liu, Z.; Feng, G. neering

Microelectronics and Computer Technology Corp., Aus- tin, TX. EDS Research

Corp. Source Codes: 888888888; MX838459 Sponsor: National Aeronautics and Space Administration,

Washington, DC 1 Apr 92 6p Languages: English Journal Announcement: GRA19707; STAR3412 In Microelectronics and Computer Technology Corp.,

Proceedings of theworkshop on Change of Representa- tion and Problem Reformulation pp. 106-110.

NTIS Prices: (Order as Nl9960047147, PC A09/MF A02)

Country of Publication: United States This paper discusses our approach to representing applica-

tion domain knowledge for specific software engineering tasks. Application domain knowledge is embodied in a domain model. Domain models are used to assist in the creation of specification models. Although many different specification models can be created from any particular domain model, each specification model is consistent and correct with respect to the domain model. One aspect of the system-hierarchical organization is described in detail. Descriptors: *Expert Systems; *Knowledge Representation; *Software Engineering; Computer Programming Identifiers: NTISNASA Section Headings: 62GE (Computers, Control, and Informa- tion Theory--General)

Learning Impasses in Problem Solving Hodgson, J. P. E. Saint Joseph’s Univ., Philadelphia, PA. Dept. of Mathe-

matics and Computer Science Corp. Source Codes: 888888888; SA818756 Sponsor: National Aeronautics and Space Administration,

Washington, DC 1 Apr 92 6p Languages: English Journal Announcement: GRA19707; STAR3412 In Saint Joseph’s Univ., Proceedings of the Workshop on

Change of Representation and Problem Reformulation pp. 94-99

NTIS Prices: (Order as N19960047147, PC A09/MF A02)

Country of Publication: United States Problem Solving systems customarily use backtracking to

deal with obstacles that they encounter in the course of trying to solve a problem. This paper outlines an approach in which the possible obstacles are investigated prior to the search for a solution. This provides a solution strategy that avoids backtracking. Descriptors: *Machine Learning; *Problem Solving; Expert Systems Identifiers: NTISNASA Section Headings: 62GE (Computers, Control, and Informa- tion Theory--General)

Scare Tactics: Evaluating Problem Decompositions Using Failure Scenarios

Helm, B. R.; Fickas, S.

Corp. Source Codes: 888888888; 00684086 Sponsor: National Aeronautics and Space Administration,

Oregon Univ., Portland, OR. Dept. of Computer and

Washington, DC 1 Apr 92 9p Languages: English

Information Science

Journal Announcement: GRA19707; STAR3412 In Oregon Univ., Proceedings of the Workshop on Change

of Representation and Problem Reformulation pp. 85-93

NTIS Prices: (Order as Nl9960047147, PC A09/MF A02)

Country of Publication: United States Contract No.: NSF CCR-88-00485 Our interest is in the design of multi-agent problem-

solving systems, which we refer to as composite systems. We have proposed an approach to composite system design by decomposition of problem statements. An automated assistant called Critter provides a library of reusable design trans- formations which allow a human analyst to search the space of decompositions for a problem. In this paper we describe a method for evaluating and critiquing problem decompositions generated by this search process. The method uses knowledge stored in the form of failure decompositions attached to design transformations. We suggest the benefits of our critiquing method by showing how it could re-derive steps of a published development example. We then identify several open issues for the method. Descriptors: *Decomposition; *Knowledge Based Systems; *Libraries; *Problem Solving; Data Structures; Expert Sys- tems; Systems Engineering Identillers: NTISNASA Section Headings: 62GE (Computers, Control, and Informa- tion Theory--General)

Reformulating Non-Monotonic Theories for Inference and Updating

Grosof, B. N. IBM Thomas J. Watson Research Center, Yorktown

Heights, NY Corp. Source Codes: 0125 10000; IA057203