From agnar.aamodt@ifi.unit.no Mon Oct 4 21:10:43 EDT 1993 Article: 19028 of comp.ai Xref: honeydew.srv.cs.cmu.edu comp.ai:19028 Newsgroups: comp.ai Path: honeydew.srv.cs.cmu.edu!magnesium.club.cc.cmu.edu!news.sei.cmu.edu!cis.ohio-state.edu!math.ohio-state.edu!howland.reston.ans.net!pipex!sunic!trane.uninett.no!ugle.unit.no!sophus.ifi.unit.no!user From: agnar.aamodt@ifi.unit.no (Agnar Aamodt) Subject: Seminar: Knowledge Level Modeling Message-ID: Followup-To: comp.ai Sender: news@ugle.unit.no (NetNews Administrator) Organization: University of Trondheim, Department of Informatics Date: Thu, 30 Sep 93 15:53:51 GMT Lines: 209 ================================================================= K N O W L E D G E L E V E L M O D E L I N G Seminar at the University of Trondheim, Norway 19-20 November 1993 CALL FOR PARTICIPATION AND PRESENTATIONS ================================================================= GENERAL ------- We are organizing an open seminar at the University of Trondheim to discuss the role of the knowledge level in current AI research and practice. There are three invited speakers, and some open slots for others who want to report on their work related to this issue. The aim of the seminar is to achieve a deeper understanding of the notion of the knowledge level, its current usage, and its future role for research and development in areas such as knowledge-based systems, cognitive science, and behaviour-based robotics. Starting from the definition in Newell's 1982 paper, the seminar will address the following issues: - The Knowledge Level as Newell saw it, its definition, purpose and application. Developments from Newell's 1982 paper to his 1989 book (Unified architectures of cognition), in which he reconfirms the knowledge level as a system description level, and extends its role. - Evolvement of the knowledge level from its original idea into more 'operationalized' and structured interpretations. Knowledge level modeling has become an important means to analysis and design in areas such as knowledge acquisition and intelligent architectures. This has lead to structured analysis and design methodologies such as KADS, Components of Expertise, Generic Tasks, Role Limiting Methods, etc. - Relations between the knowledge level and its realization at the symbol level. This may be generalized to relations between the knowledge level and other system description levels in general, but the focus will be on issues related to the difficult task of implementing (parts of) a conceptual and intentional knowledge level model in a computer language. - The role of the knowledge level as a scientific tool for research in AI and cognitive science. - Tool support, both for performing knowledge level analysis, and for assisting the transformation/reformulation into a symbol level model and a running target system. The focus will be on fundamental and principal issues related to the knowledge level and its role. Descriptions of specific modeling methods and techniques will not be a target in itself, but a means to exemplify the more general issues. PROGRAM ------- The three invited talks are: * Walter Van de Velde, Free University of Brussels, Belgium: "ISSUES ON KNOWLEDGE LEVEL MODELING" Since its introduction in the early 80s the notion of knowledge level has been an important catalizer of research in knowledge systems. This talk discusses how it is being turned into a useful tool for the development of knowledge systems and how the original and present interpretations can be tied together again. It shows how the knowledge level changed our views on what knowledge systems are and how the problems with first generation expert systems might be overcome. Two other issues are discussed in some more detail. The first one is the precise methodological role of the knowledge level. The second issue concerns the nature of knowledge level theories of problem solving and its implications for next generation architectures. Dr. Walter Van de Velde is a research director at the Belgian National Science Foundation and co-director of the Artificial Intelligence Laboratory of the Flemish Free University of Brussels (VUB) where he heads the knowledge based systems group. His main research interests are second generation expert systems, machine learning, and architectures for intelligent behavior. * Jon Sticklen, Michigan State University, USA: "SOFTWARE ENGINEERING, TASK SPECIFIC ARCHITECTURES, AND THE KNOWLEDGE LEVEL: THREE PERSPECTIVES ON THE SAME ELEPHANT " Software engineering as a discipline aims at the goal of effective construction and maintenance of software artifacts which accomplish specified tasks. Task specific architectures as a sub-discipline of AI aims at the goal of capturing expert problem solving in terms "natural" to the problem. The Knowledge Level Hypothesis, as a much used but little understood concept, lays a putative framework for describing a very high level analysis for problem solving systems. As originally described, the KL is not a prescription for representation or a methodology for system building, but rather is best understood as an analytic tool for understanding the operation of knowledge systems. All three - Software Engineering, TSAs, and the KL viewpoint - deal with describing problem solving behavior. In this position paper, we will argue that there is an emerging integration of SE, TSAs, and the KL viewpoint. We will argue that the confluence of SE and the KL are in fact converging on a (strongly augmented) TSA view. Dr. Jon Sticklen is Associate Professor of Computer Science and Director of the Intelligent Systems Laboratory, Michigan State University. Following his Ph.D. studies at the Ohio State University, Dr. Sticklen joined the faculty at MSU where he initiated an active research program in knowledge-based systems. Sticklen's major research foci are in integrative approaches to large grain task specific problem solving, function-based reasoning, and the theory of knowledge-based systems. Sticklen and his associates are currently engaged in a wide variety of domain projects including the design and fabrication of polymer composite materials, the modeling of landscape level ecological systems, and the development of decision support software for managing production agriculture in the lower Nile valley. * Carl Gustaf Jansson, Stockholm University/KTH, Sweden: "ONTOLOGICAL VARIETY AND SHIFTS, AND ITS IMPACT ON KNOWLEDGE LEVEL DESIGN " For many applications, ontological issues are not considered. It is assumed that representation and problem solving procedures in a system are based upon conceptions of the domain which are shared among all users of the system. This may be a realistic simplification in some cases, but for many applications where aspects of help, explanation and tutoring have an important role, a system must be able to cope with a variety of ontologies and ontological shifts. The systematic handling of this kind of situation will have a great impact on methods for knowledge level design. The purpose of the talk is to argue for this thesis and illustrate the problems. Examples will be taken from cosmology, heat transfer and programming methodology. The research problems described are focused in a recently started European Science Foundation program on Learning in Humans and Machines. Dr. Carl Gustaf Jansson is acting professor and deputy head of department for Computer and Systems Sciences, Stockholm University. He is chairman of the board for the Centre for Information Technology and Cognitive Science, and scientific advisor at the Swedish Institute for Computer Science (SICS). He currently heads a research group in Machine Learning and supervises research students in knowledge-based systems design, intelligent interfaces, and natural language processing. We will also present ongoing research in our department related to the seminar topic: * Agnar Aamodt, University of Trondheim, Norway: "KNOWLEDGE LEVEL MODELING AND THE SYMBOL LEVEL GAP" Over the last few years we have seen significant advances in methods for constructing conceptual, knowledge level models of system behaviour, as well as in techniques for symbol level realizations of knowledge-based systems. This has lead to increased focus on the transition between these two modeling levels. Current methods tend to be strongly biased by a top-down approach to knowledge modeling, without addressing iterative aspects of modeling through sustained learning from experience. This talk will outline the major problems involved, present a framework for combining top-down knowledge modeling with bottom-up machine learning, and describe initial attempts to analyse the problems involved using a knowledge-level analysis tool (Krest, based on the Components of Expertise framework). It will further discuss the role of case-based reasoning as a means to integrate initial knowledge modeling with sustained learning. Dr. Agnar Aamodt is a senior researcher in the Department of Informatics, University of Trondheim. His research is directed towards improved methods for construction and maintenance of knowledge-based systems. It includes work on knowledge acquisition and modeling, integrated problem solving and learning architectures, and case-based reasoning. OPEN INVITATION AND PRACTICALITIES ---------------------------------- We hereby invite others who also would like to present their work to submit a 1-2 page abstract. There will be room for another 5-6 presentations of 15-30 minutes length. We are planning to publish a book on knowledge level modeling based on presentations at the seminar. Papers for the book should be submitted after the seminar (date to be decided). The seminar will be held at the University of Trondheim, College of Arts and Science. The main part of the seminar will take place on Friday 19th of November. On Saturday 20th we will have in-depth discussions of some of Friday's topics. The seminar is informal, open to those interested, and free to attend. If you want to contribute or attend, please send an email message stating so, indicate whether you want to make a presentation, and include a few lines stating your particular interest and/or work on issues related to knowledge level modeling. Agnar Aamodt Seminar Organizer --------------------------------------------------------------------------- Address etc.: Agnar Aamodt | Email: agnar@ifi.unit.no Department of Informatics | Fax: +47-7-591733 College of Arts and Science | Phone: (office): +47-7-591838 University of Trondheim | Phone: (secr.): +47-7-591840 N-7055 Dragvoll, NORWAY | Phone: (home): +47-7-977365 --------------------------------------------------------------------------- Article 4857 of news.announce.conferences: Xref: crabapple.srv.cs.cmu.edu news.announce.conferences:4857 Newsgroups: news.announce.conferences Path: crabapple.srv.cs.cmu.edu!fs7.ece.cmu.edu!news.sei.cmu.edu!cis.ohio-state.edu!magnus.acs.ohio-state.edu!math.ohio-state.edu!howland.reston.ans.net!pipex!uunet!sparky!rick From: agnar.aamodt@ifi.unit.no (Agnar Aamodt) Subject: Knowledge Level Modeling - Trondheim Seminar Message-ID: <1993Oct3.190146.29820@sparky.sterling.com> Sender: rick@sparky.sterling.com (Richard Ohnemus) Organization: University of Trondheim, Department of Informatics Date: Sun, 3 Oct 1993 19:01:46 GMT Approved: rick@sparky.sterling.com Expires: Sun, 21 Nov 1993 08:00:00 GMT Lines: 208 X-Md4-Signature: fa2b9f57fe9b89e866da3c7862d42cb6 ================================================================= K N O W L E D G E L E V E L M O D E L I N G Seminar at the University of Trondheim, Norway 19-20 November 1993 CALL FOR PARTICIPATION AND PRESENTATIONS ================================================================= GENERAL ------- We are organizing an open seminar at the University of Trondheim to discuss the role of the knowledge level in current AI research and practice. There are three invited speakers, and some open slots for others who want to report on their work related to this issue. The aim of the seminar is to achieve a deeper understanding of the notion of the knowledge level, its current usage, and its future role for research and development in areas such as knowledge-based systems, cognitive science, and behaviour-based robotics. Starting from the definition in Newell's 1982 paper, the seminar will address the following issues: - The Knowledge Level as Newell saw it, its definition, purpose and application. Developments from Newell's 1982 paper to his 1989 book (Unified architectures of cognition), in which he reconfirms the knowledge level as a system description level, and extends its role. - Evolvement of the knowledge level from its original idea into more 'operationalized' and structured interpretations. Knowledge level modeling has become an important means to analysis and design in areas such as knowledge acquisition and intelligent architectures. This has lead to structured analysis and design methodologies such as KADS, Components of Expertise, Generic Tasks, Role Limiting Methods, etc. - Relations between the knowledge level and its realization at the symbol level. This may be generalized to relations between the knowledge level and other system description levels in general, but the focus will be on issues related to the difficult task of implementing (parts of) a conceptual and intentional knowledge level model in a computer language. - The role of the knowledge level as a scientific tool for research in AI and cognitive science. - Tool support, both for performing knowledge level analysis, and for assisting the transformation/reformulation into a symbol level model and a running target system. The focus will be on fundamental and principal issues related to the knowledge level and its role. Descriptions of specific modeling methods and techniques will not be a target in itself, but a means to exemplify the more general issues. PROGRAM ------- The three invited talks are: * Walter Van de Velde, Free University of Brussels, Belgium: "ISSUES ON KNOWLEDGE LEVEL MODELING" Since its introduction in the early 80s the notion of knowledge level has been an important catalizer of research in knowledge systems. This talk discusses how it is being turned into a useful tool for the development of knowledge systems and how the original and present interpretations can be tied together again. It shows how the knowledge level changed our views on what knowledge systems are and how the problems with first generation expert systems might be overcome. Two other issues are discussed in some more detail. The first one is the precise methodological role of the knowledge level. The second issue concerns the nature of knowledge level theories of problem solving and its implications for next generation architectures. Dr. Walter Van de Velde is a research director at the Belgian National Science Foundation and co-director of the Artificial Intelligence Laboratory of the Flemish Free University of Brussels (VUB) where he heads the knowledge based systems group. His main research interests are second generation expert systems, machine learning, and architectures for intelligent behavior. * Jon Sticklen, Michigan State University, USA: "SOFTWARE ENGINEERING, TASK SPECIFIC ARCHITECTURES, AND THE KNOWLEDGE LEVEL: THREE PERSPECTIVES ON THE SAME ELEPHANT " Software engineering as a discipline aims at the goal of effective construction and maintenance of software artifacts which accomplish specified tasks. Task specific architectures as a sub-discipline of AI aims at the goal of capturing expert problem solving in terms "natural" to the problem. The Knowledge Level Hypothesis, as a much used but little understood concept, lays a putative framework for describing a very high level analysis for problem solving systems. As originally described, the KL is not a prescription for representation or a methodology for system building, but rather is best understood as an analytic tool for understanding the operation of knowledge systems. All three - Software Engineering, TSAs, and the KL viewpoint - deal with describing problem solving behavior. In this position paper, we will argue that there is an emerging integration of SE, TSAs, and the KL viewpoint. We will argue that the confluence of SE and the KL are in fact converging on a (strongly augmented) TSA view. Dr. Jon Sticklen is Associate Professor of Computer Science and Director of the Intelligent Systems Laboratory, Michigan State University. Following his Ph.D. studies at the Ohio State University, Dr. Sticklen joined the faculty at MSU where he initiated an active research program in knowledge-based systems. Sticklen's major research foci are in integrative approaches to large grain task specific problem solving, function-based reasoning, and the theory of knowledge-based systems. Sticklen and his associates are currently engaged in a wide variety of domain projects including the design and fabrication of polymer composite materials, the modeling of landscape level ecological systems, and the development of decision support software for managing production agriculture in the lower Nile valley. * Carl Gustaf Jansson, Stockholm University/KTH, Sweden: "ONTOLOGICAL VARIETY AND SHIFTS, AND ITS IMPACT ON KNOWLEDGE LEVEL DESIGN " For many applications, ontological issues are not considered. It is assumed that representation and problem solving procedures in a system are based upon conceptions of the domain which are shared among all users of the system. This may be a realistic simplification in some cases, but for many applications where aspects of help, explanation and tutoring have an important role, a system must be able to cope with a variety of ontologies and ontological shifts. The systematic handling of this kind of situation will have a great impact on methods for knowledge level design. The purpose of the talk is to argue for this thesis and illustrate the problems. Examples will be taken from cosmology, heat transfer and programming methodology. The research problems described are focused in a recently started European Science Foundation program on Learning in Humans and Machines. Dr. Carl Gustaf Jansson is acting professor and deputy head of department for Computer and Systems Sciences, Stockholm University. He is chairman of the board for the Centre for Information Technology and Cognitive Science, and scientific advisor at the Swedish Institute for Computer Science (SICS). He currently heads a research group in Machine Learning and supervises research students in knowledge-based systems design, intelligent interfaces, and natural language processing. We will also present ongoing research in our department related to the seminar topic: * Agnar Aamodt, University of Trondheim, Norway: "KNOWLEDGE LEVEL MODELING AND THE SYMBOL LEVEL GAP" Over the last few years we have seen significant advances in methods for constructing conceptual, knowledge level models of system behaviour, as well as in techniques for symbol level realizations of knowledge-based systems. This has lead to increased focus on the transition between these two modeling levels. Current methods tend to be strongly biased by a top-down approach to knowledge modeling, without addressing iterative aspects of modeling through sustained learning from experience. This talk will outline the major problems involved, present a framework for combining top-down knowledge modeling with bottom-up machine learning, and describe initial attempts to analyse the problems involved using a knowledge-level analysis tool (Krest, based on the Components of Expertise framework). It will further discuss the role of case-based reasoning as a means to integrate initial knowledge modeling with sustained learning. Dr. Agnar Aamodt is a senior researcher in the Department of Informatics, University of Trondheim. His research is directed towards improved methods for construction and maintenance of knowledge-based systems. It includes work on knowledge acquisition and modeling, integrated problem solving and learning architectures, and case-based reasoning. OPEN INVITATION AND PRACTICALITIES ---------------------------------- We hereby invite others who also would like to present their work to submit a 1-2 page abstract. There will be room for another 5-6 presentations of 15-30 minutes length. We are planning to publish a book on knowledge level modeling based on presentations at the seminar. Papers for the book should be submitted after the seminar (date to be decided). The seminar will be held at the University of Trondheim, College of Arts and Science. The main part of the seminar will take place on Friday 19th of November. On Saturday 20th we will have in-depth discussions of some of Friday's topics. The seminar is informal, open to those interested, and free to attend. If you want to contribute or attend, please send an email message stating so, indicate whether you want to make a presentation, and include a few lines stating your particular interest and/or work on issues related to knowledge level modeling. Agnar Aamodt Seminar Organizer --------------------------------------------------------------------------- Address etc.: Agnar Aamodt | Email: agnar@ifi.unit.no Department of Informatics | Fax: +47-7-591733 College of Arts and Science | Phone: (office): +47-7-591838 University of Trondheim | Phone: (secr.): +47-7-591840 N-7055 Dragvoll, NORWAY | Phone: (home): +47-7-977365 ---------------------------------------------------------------------------