Drafts for AAAI Spring Symposium
Artificial Intelligence in Knowledge Management Stanford University, March 24-26, 1997

Towards a Well-Founded Technology For Organizational Memories

Andreas Abecker, Ansgar Bernardi, Bernd Bachmann, Knut Hinkelmann, Otto Khn and Michael Sintek, German Research Center for Artificial Intelligence (DFKI) GmbH, P.O. Box 2080, D-67608 Kaiserslautern, Germany. email: aabecker@dfki.uni-kl.de

We report on ongoing work in the knowledge management lab at DFKI Kaiserslautern. We present a three-layered architecture for an organizational memory as a central technological prerequisite for knowledge management. Explicit representation of knowledge context and task-specific relevance are discussed. Compressed Postscript

Statement of Interest for AAAI Spring Symposium: AI in Knowledge Management

Cynthia L. Bernstein, Andersen Consulting LLP, Andersen Consulting Technology Park, 3773 Willow Road, Northbrook, IL 60062, USA. email: cynthia.l.bernstein@ac.com

I work at Andersen Consulting on knowledge navigation in the area of knowledge management. My roles include conceptualizing issues about knowledge navigation and addressing them by designing tools, interfaces, processes, and architectures to facilitate knowledge navigation. The end-users of these knowledge navigation tools are Andersen consultants, who do their work supported by our global firms collective knowledge capital. HTML

Agents in Knowledge Management

Jeffrey M. Bradshaw, Research and Technology, Boeing Information and Support Services P.O. Box 3707, M/S 7L-44, Seattle, WA 98024, USA. email: jbrad@redwood.rt.cs.boeing.com

A Model of Crisis Managment System including Mental Representation

Alain Cardon and Stephane Durand, PSI-INSARouen, LAFORIA, 9 Place Emile Blondel 76134 Mont Saint Aignant Cedex, Paris VI URA CNRS 1095, France. email: Alain.Cardon@insa-rouen.fr, Stephane.Durand@insa-rouen.fr

We present a dynamic model for Communication and Information Systems of crisis managment which takes into account the mental representation from actors. This model allows the representation of the intentions and the judgments expressed by the differrent actors when they exchange information about the situation. It mainly uses auto modifying dynamic Multi-Agent Sytems and produces a graphical desciption of judgments expressed on the phenomenon. Compressed Postscript

Constraints for Knowledge Maintenance

John Debenham, Computing Sciences, University of Technology, Sydney, PO Box 123, Broadway, NSW 2007, Australia. email: debenham@socs.uts.EDU.AU

Constraints are used to manage the integrity of knowledge. Knowledge-based systems applications are frequently constructed as a knowledge base, or rule base, component coupled with a relational database component. The most expensive of these two components to build and maintain is typically the knowledge base. Constraints, which play a key role in database management, are seldom mentioned in connection with knowledge. One class of knowledge constraints protect the integrity of knowledge bases during maintenance by preventing the introduction of update anomalies. Another class of knowledge constraints contributes to the efficiency of the maintenance procedure. The efficiency of the maintenance procedure is increased further if the knowledge in the knowledge base has been normalised. An experimental knowledge-based systems design and maintenance tool which incorporates this approach has been built and trialed in a commercial environment. Compressed Postscript

Providing User-Support in Performing Knowledge Discovery in Databases

Robert Engels, Michael Erdmann, Rainer Perkuhn, and Rudi Studer, Institut fur Angewandte Informatik und Formale Beschreibungsverfahren, University of Karlsruhe (TH), D-76128 Karlsruhe, Germany. email: {engels | erdmann | perkuhn | studer}@aifb.uni-karlsruhe.de

The statement of interest sketches a project called PUK, which aims at Providing User-Support in Performing Knowledge Discovery in Databases. KDD is - as we think - one important facet of Knowledge Management. The PUK project has two main objectives: (a) developing a methodology for supporting the user when performing an entire KDD process and (b) evaluating methodological progress with real industrial applications. The methodology should should provide a user guidance module to enable *domain* experts to specify KDD processes by (re-)using stored KDD processes, data mining algorithms etc. Thus a repository and retrieval mechanisms were needed. Compressed Postscript

Acquiring, Maintaining, and Customizing Organizational Work Process Descriptions

Douglas B. Fridsma, John Gennari and Mark Musen, Section on Medical Informatics, Stanford University School of Medicine, Stanford University, Stanford, CA 94305-5479, USA. email: {fridsma, gennari, musen}@smi.stanford.edu

Models of work processes and of organizations' activities are an important part of workflow systems and simulation, and capture procedural knowledge stored within the organization. However, acquiring, maintaining, and customizing these models can be difficult. To assist the acquisition and maintenance of organizational models, we have developed a set of knowledge-acquisition tools suitable for use in the domain of medical organizations. In modeling medical organization processes, three areas of expertise must be brought together. Expertise in medical care, organizational structure, and modeling is necessary to describe accurately the process of medical care delivery for simulation or workflow management. We propose a method of modeling work processes that uses the Protege suite of tools to generate organization-specific work process descriptions. We have created a set of Protege knowledge-acquisition tools customized for medical, organization, and modeling experts, and have used a prototype system to create detailed, site-specific process descriptions. In our prototype system, we have identified a set of transformation operators that should make possible additional computer-based support. We believe this methodology will improve acquisition of an organizational model, and make it easier to maintain an accurate model in the face of changes in medical process or organizational structure. Compressed Postscript

What is knowledge that we may manage it?

Brian R. Gaines, Knowledge Science Institute, University of Calgary, Alberta, Canada T2N 1N4. email: gaines@cpsc.ucalgary.ca

Studies of the 'knowledge level' in artificial intelligence research model it as arising from social processes whereby agents need to be aware of, utilize, develop and manage the capabilities of other agents. 'Knowledge' is a state variable that we impute to an agent to account for its capabilities to perform tasks. The notion of knowledge is then extended to account for other phenomena in which an agent's capabilities are changed in some situation--for example, by access to written material, a tool, or other agents. This extension leads to the definition of two different but highly related notions of knowledge: the knowledge imputed to an agent to account for its capabilities; and the knowledge imputed to something which can change an agent's capabilites. This paper applies this framework to issues of 'knowledge management' within organisations showing how the standard approaches to managing people and resouces derive from it, and how new approaches have arisen as information technology has come to play a major role.

Case Base Engineering for Large Scale Industrial Applications

Kalyan Moy Gupta, Atlantis Aerospace Corporation, 1 Kenview Blvd., Brampton, ON, L6T 5E6, Canada. email: kgupta@atlantis.com

In the recent years, case-based reasoning (CBR) has emerged as a promising technology for industrial decision support. The effectiveness of the decisions supported by a CBR system depends, to a large extent, on the quality of knowledge contained in the cases. Case bases containing good quality problem solving information can be developed through the process of case base engineering (CBE). CBE for large scale industrial CBR applications requires a framework that brings together the methodologies from diverse fields. These include information requirement analysis, decision analysis, knowledge acquisition, knowledge representation, information presentation, and the management of the growth and evolution of case bases. In this talk, we present the processes and methodologies used in CBE. The methodologies are discussed in light of applications in the domain of plastics manufacturing equipment maintenance and aircraft maintenance. Finally, we identify research opportunities in CBE.

Consolidating Multi-Source and Multi-Media Knowledge

Roger T. Hartley, Robert L. Kelsey, Robert B. Webster, New Mexico State University and Los Alamos National Laboratory, USA. email: rth@cs.nmsu.edu, rkelsey@cs.nmsu.edu, robw@lanl.gov

Making use of multi-media knowledge and information from multiple sources is a problem of knowledge management. Even when a user knows of all the different sources of knowledge available, it is a difficult task finding pertinent knowledge and information within those sources. Finding what is important to the user may mean selecting, filtering, and distilling large quantities of knowledge. In addition, the knowledge is hidden within many different types of formats, representations, and media, including audio, video, paper, and electronic. The problem before us is to use the techniques and philosophy of knowledge management and knowledge engineering to consolidate the knowledge important to a user and deliver it to the user's desktop. Compressed Postscript

Unique Challenges of Managing Inductive Knowledge

David Jensen Experimental Knowledge Systems Lab Computer Science Department University of Massachusetts Amherst, MA 01003-4610, USA. email: jensen@cs.umass.edu

Techniques for inducing knowledge from databases, often grouped under the term knowledge discovery, are becoming increasingly important to organizations in business, government, and science. However, relatively little attention has been paid to the long-term management of induced knowledge. Induced knowledge presents unique challenges, including managing statistical significance and inductive bias. These two challenges have important implications for valid and efficient knowledge management. Compressed Postscript

Knowledge Management on a Global Scale

Philip Klahr, Vice President, Customer Quality, Inference Corporation, 100 Rowland Way, Novato, California 94945 USA. email: klahr@inference.com

Over the past few years, there have been a number of companies developing knowledge-based systems in the area of customer support. Since many of these companies are global, there has been increasing interest on building and deploying knowledge bases on a global scale -- both to leverage the knowledge and expertise that is distributed around the world, and also to make case bases available to all the regional support organizations to solve customer problems. The principal AI-based technology that has been used in these customer support efforts is Case-Based Reasoning (CBR). "Cases" are used as the representational framework to capture knowledge of customer issues, problems and solutions, and general queries that customers will likely ask. Providing a CBR search engine to end users allows them to find previous solutions that may help solve a particular customer+s problem. While the issues in developing a global knowledge system are many, they are now fairly well understood. However, no two approaches are identical, since companies have differing business objectives, operational infrastructures, and technical requirements. The focus of this presentation is to specify the issues companies need to address in developing a global knowledge system, and provide examples, and case studies, of the alternative approaches companies have taken. HTML

A Concept Mapping Tool to Emulate Multiple Knowledge Representation Formalisms

Rob Kremer, Knowledge Science Institute, University of Calgary, Calgary, Alberta Canada T2N 1N4. email: kremer@cpsc.ucalgary.ca

Concept maps are used in a wide variety of disciplines because of their ability to make complex information structures explicit. Concept maps can be used informally or formally - where the graphical "syntax" of the maps is tightly controlled. Both forms are needed. Constraint Graphs is a program in which users can constrain arbitrary graphs to conform to any of a wide variety of graphical formalisms. The Graphs program is combined with a graphical user interface to yield an interactive concept mapping system, that can transcend informal concept mapping and at least several concept mapping formalisms. HTML

Knowledge Management using a Semantic-Network

Yoshitaka Kuwata and Masashi Yatsu, NTT DATA CORP., 66-2 Horikawa-cho, Saiwai-ku, Kawasaki-shi, Kanagawa, 210 JAPAN. email: kuwata@lit.rd.nttdata.co.jp

It has become very important for advanced organizations to make the best use of information gathered from database in companies and from the Internet. There are three stages in the information life cycle; Finding, Organizing and Sharing. Many technologies have been developed for the finding stage. On the other hand, no concrete organizing and sharing technologies exist to manage the information found. In this paper, we focus on the management of information sharing among group members. We propose to use a semantic-network for organizing information that has been gathered. We show an example of how to manage URL information in order to share it among small groups. As the system is under development, only a partial evaluation of the method is given. Compressed Postscript

Knowledge Management Using MODEL-ECS

Dickson Lukose, Distributed Artificial Intelligence Centre (DAIC*). Department of Mathematics, Statistics, and Computing Science, The University of New England, Armidale, 2351, N.S.W., Australia. email: lukose@peirce.une.edu.au

Knowledge is the fundamental resource that allows us to function intelligently. Similarly, organisations (i.e., enterprises) typically use different types of knowledge to enhance their performance. Task ontology modelling is the key to corporate knowledge management via modelling shareable and reusable knowledge bases. The mapping between the domain specific task ontology, and the domain independent task ontology, is crucial for realising shareable and reusable knowledge bases. This paper suggests an approach to achieving this objective, by proposing a three phase knowledge engineering approach. Task ontology is modelled using MODEL-ECS (i.e., graphically based executable conceptual modelling language). Conceptualisation of the organisation and its knowledge base is crucial to formalising the mapping functions necessary to realise corporate memory modelling. This paper describes one such approach. Compressed Postscript

CoMo-Kit: Knowledge Based Workflow Management - Summary of an ongoing development

Frank Maurer, AG Expertensystemem, Universitat Kaiserslautern, Postfach 3049, D-67653 Kaiserslautern, Germany. email: maurer@informatik.uni-kl.de

The paper briefly describes the goals and the approach of the CoMo-Kit project on knowledge- based workflow management in engineering domains. Compressed Postscript

Knowledge Management for the Applied Sciences

Craig Mcdonald, Daniel Pun and John Weckert, Knowledge Management Group, School of Information Studies, Charles Sturt University, P.O. Box 588, Wagga Wagga, N.S.W. 2678, Australia. email: cmcdonald@csu.edu.au, dpun@csu.edu.au, jweckert@csu.edu.au

Knowledge of how the natural world works (science) and how humans can interfere for their own purposes (technology) is created by scientific research and is also being built into computer-based systems for technology transfer. First generation expert systems are built from knowledge acquired from human experts who have domain-specific knowledge in a specific field. This study argues that systems, which give advice, should not only capture expert's thoughts and experiences, but also take advantage of objective sources of knowledge. Such sources include published empirical data, text books, research reports, etc. This approach differs from that used in first generation expert systems in that it uses specific published research to augment expert opinion. A prototype Knowledge Management System (KMS) is being built in the irrigation of grapevines as a means of evaluating the KMS approach. Compressed Postscript

Issues for Knowledge Management from Experiences in Supporting Group Knowledge Elicitation & Creation in Ill-defined, Emerging Situations

John T. Nosek and Michael D. McNeese, Computer & Information Sciences Dept, Temple University, Philadelphia, PA19122, USA and Armstrong Laboratory / Human Engineering Division Wright-Patterson Air Force Base, OH, USA. email: nosek@thunder.ocis.temple.edu

While organizational environments become more complex and dynamic, there exists limited time to exploit an opportunity or solve a problem. Real time decision makers under stress, for example, combat fighter pilots, come across a fast unfolding situation and must quickly make an acceptable decision. They focus on assessing the situation and taking acceptable actions that present themselves - optimization is not really possible nor sought. Because of increasing time pressures to act quickly, organizational decision makers are beginning to find themselves in a similar position to real time decision makers -- situation assessment is critical and acceptable actions are sought. We are not trying to say that fighter pilots and organizational decision makers work within the same split second time frames, however, the changes for fighter warriors and organizational decision makers can inform our understanding of group sensemaking. For example, with the increasing use of Òreal time information in the cockpitÓ, pilots will be able to view advanced imagery and communicate/collaborate using highly sophisticated technologies. As these advancements become prevalent, the view of the pilot as individual warfighter becomes obsolete, and barriers between real time and organization decision making breakdown. Within this complex, broad bandwidth decision space, there are many possible actor-to-actor or actor-to-agent couplings that underlie group sensemaking. Compressed Postscript

KR, Diagrams For Models, and Computation

Cyrus F. Nourani, Project METAAI, USA. email: 73244.377@compuserve.com

We present the method of knowledge representation with G-diagrams and applications to define computable models for AI reasoning . G-diagrams are diagrams defined from a minimal set of function symbols that can inductively define a model. G-diagrams are devised for various KR and AI reasoning applications as a minimal efficient computable method of representing knowledge and AI worlds. We show how computable AI world knowledge is representable and how various G-diagrams are applied towards KR from planning with nondeterminism and planning with free proof trees to partial deduction with abductive diagrams presented by this paper. Practical diagram computation is by 0-1 assignments to Boolean predicates. A correspondence between intial models definable by G-diagrams, theire computations, and Boolean algebras are presented by defining a correspondence between inital models and Boolean algebras. Computability problems for KR are presented in brief viewed from recent diagram computing theories. Compressed Postscript

Clustering Algorithm for Large-Scale Databases

Yuiko Ohta, Seishi Okamoto, and Nobuhiro Yugami, FUJITSU LABORATORIES LTD. NetMedia Laboratory, Japan. email: yuiko@flab.fujitsu.co.jp, seishi@flab.fujitsu.co.jp, yugami@flab.fujitsu.co.jp

This paper proposes a new clustering algorithm to create hierarchical concepts. Clustering system is very helpful to organize databases, since it extracts useful information from databases. There are many incremental and non-incremental algorithms for clustering. Incremental methods are efficient, but have the problem depending on the input order of instances. On the other hand, the performance of non-incremental methods has no relation to the input order of instances, but non-incremental methods need much computational time. Here, we present an non-incremental algorithm to generate hierarchical clusterings within reasonable time. Compressed Postscript

Position paper for the AAAI Symposium on "AI in Knowledge Management"

Rashmi Pandya, Motorola ECID, 16 Euroway, Blagrove, Swindon SN5 8YQ, UK. email: pandyar@ecid.cig.mot.com

Knowledge based systems can be widely used to add strategic value. Regardless of the industry, there are opportunities to maximise results using KBS technology in products, services and institutionalised learning. In the telecommunications industry, KBS's are already being used in the marketplace, in wire-based telephone networks, data networks and other related areas. There now seems to be an interest in the possibility of extending the application of knowledge based systems to the vast and rapidly growing area of mobile communications. However, this interest is perplexed by the various novel methodologies and representations unique to these systems. Reasoning systems, expert systems, planning and learning systems may all be suitably applied in the field of mobile communications and each paradigm emphasises the need for suitable knowledge acquisition, representation and management. Motorola is particularly interested in the assessment and survey of sound and practical AI techniques to knowledge acquisition and management with the view of possible application.

A data-processing environment to revise knowledge of a researcher in applied sciences

Michel Sala, LIRMM, UMR 9928 CNRS-Montpellier II 161 rue Ada, 34392 Montpellier Cedex 5, France. email: sala@lirmm.lirmm.fr

The objective of our work is to enable a researcher in applied sciences to revise his knowledge on his field. The basic idea is both to help the confrontation of this knowledge with data of experimentation or results of computationnal tools and to provide a whole of explanations. In order to install this environment, we firstly described the knowledge of the researcher in the form of a set of constraints. We also built two modules :

This work gave rise to system SIGALE (Sequence-Ig Alignment Learning Environnement) dedicated to the classification of the immunoglobulin genetic sequences. In this article, we will present the architecture of our environment, its different components and we will conclude with an example of the cycle acquisition / revision into immunogenetics. Compressed Postscript

Negotiating Multidisciplinary Integration: From Collaborative Argumentation to Organisational Memory

Simon Buckingham Shum, Knowledge Media Institute, The Open University, U.K. email: S.Buckingham.Shum@open.ac.uk

are defined, project constraints shift, and teams reconcile competing agendas. Graphical argumentation provides a shared working memory in meetings by focusing discussion. Secondly, the product of using such a tool to conduct discussions is a shared long term memory of the intellectual investment, thus resisting 'organisational amnesia.' Hypermedia groupware provides a way to link informal, socially embedded knowledge with other work artifacts such as reports, sketches and simulations. Examples of this approach's application are surveyed, looking particularly at the use of semiformal argumentation to integrate modelling techniques from the computing and cognitive sciences. This is followed by considering the cognitive, group and organisational dynamics that can support or obstruct the introduction of such an approach. The concluding discussion seeks to open up a broader debate about organisational knowledge systems, and proposes several key issues which any organisational knowledge approach should reflect on. Compressed Postscript

Balancing Formality with Informality: Some Requirements for Organisational Memory Technologies

Simon Buckingham Shum, Knowledge Media Institute, The Open University, U.K. email: S.Buckingham.Shum@open.ac.uk

Numerous disciplines are now trying to analyse and represent the processes and products of organisational memory, in order to clarify what tangible representations future knowledge managers will work with. This short paper begins by reflecting briefly on the nature of systematic representations, as a reminder of the commitments that are made in any classification process. It is argued that there are important political dimensions to such classification, with implications for knowledge modelling. The paper then identifies four processes by which organisational expertise is shared. These processes may represent both a challenge and an opportunity for knowledge modelling approaches. The closing discussion pinpoints formalisation as a particularly important process in knowledge management, considers technological developments that support incremental formalisation as holding particular promise, and proposes the principles that only stable, sanctioned knowledge should be formalised, in order to avoid the many problems caused by premature formalisation of organisational knowledge. Compressed Postscript

Caring About Knowledge: The Importance of the Link Between Knowledge and Values

Sue P. Stafford, Department of Philosophy, Simmons College, Boston, MA, USA. email: SSTAFFORD@VMSVAX.SIMMONS.EDU

This paper presents five major claims. One business "case study" will be used to illustrate each of them.

  1. To manage knowledge, we must first understand that knowledge and values (care-abouts) are intimately connected.
  2. Knowledge modeling will facilitate knowledge management if knowledge and care-abouts are modeled.
  3. The sharing of knowledge models creates a new corporate reality of understanding.
  4. Understanding which encompasses both knowledge and care-abouts will result in active doing.
  5. Technology can be used to facilitate the modeling of knowledge HTML

AI-Style Knowledge Meets Web Documents: Aspects and Techniques for Hybrid KM

Doug Skuce, Department of Computer Science, University of Ottawa, Ottawa, Canada. email: doug@csi.UOttawa.CA

Knowledge is a key resource that we still do not have many new ideas how to handle. Most (online) knowledge is currently kept in conventional documents that are hard to structure, classify, browse, search, and even find. Organizations are drowning in masses of such documents of hundreds of formats. But classical AI has ignored this real and serious problem. Information retrieval research has tackled some of the problems, but is totally at odds with how AI tries to deal with knowledge problems. Cooperative work systems, particularly Lotus Notes, are beginning to tackle another aspect. Hence we need to combine ideas from these and other sources to improve the management of most kinds of knowledge. This paper describes an attack on the knowledge management problem by introducing a new hybrid notion of document cum knowledge base. We also discuss our system, IKARUS, that processes such entities. HTML

Beyond full-text search: AI-based technology to support the knowledge cycle

David M. Steier, Scott B. Huffman, Douglas I. Kalish, Price Waterhouse World Technology Centre, 68 Willow Road, Menlo Park, CA, USA. email: {steier,huffman,kalish}@tc.pw.com

From the mounds of raw information available electronically today, what professionals really need are targeted, timely nuggets of knowledge that can guide the solution to business problems. Todays common information tools Web full-text search engines and the like do not fully support this conversion of raw information into knowledge. In examining the common knowledge management problems faced by Price Waterhouse professionals, we have found that converting information to knowledge requires not only finding raw information, but also filtering through it for relevance, formatting it appropriately for the knowledge task at hand, and forwarding it to the right people. A fifth stage, feedback from the users, can allow the effectiveness of each stage to increase with time. In this paper, we describe each stage of this knowledge cycle and discuss the potential role that AI-based technology can play in its automation. We illustrate the possibilities through case studies of deployed knowledge management tools we have built at Price Waterhouse. These tools demonstrate that for targeted business tasks, AI-based technology can potentially facilitate much of the knowledge cycle, providing users with useful business knowledge that provides competitive advantage. Compressed Postscript

AI-Techniques and the Knowledge Pump

G. van Heijst, M. Hofman, E. Kruizinga, R. van der Spek, Kenniscentrum CIBIT, Arthur van Schendelstraat 570, 3500 AN Utrecht, the Netherlands. {gvheijst,mhofman,ekruizinga,rvdspek}@cibit.hvu.nl

Kenniscentrum CIBIT is a Dutch non-profit organization which, amongst other activities, provides consulting services in the areas of knowledge management and knowledge technology. In our work, we make a distinction between three viewpoints on knowledge management:

Each viewpoint has its own set of typical problems and its own set of tools and techniques to solve these problems. In this position paper, we will articulate our approach with respect to problems associated with the learning viewpoint. Compressed Postscript

Building up and Making Use of Corporate Knowledge Repositories

Gian Piero Zarri, Centre National de la Recherche Scientifique, EHESS - CAMS, 54, boulevard Raspail, 75270 PARIS Cedex 06, France. email: zarri@cams.msh-paris.fr

In this paper, we present a methodology (and some concrete experiments) for the construction and use of corporate knowledge repositories. They can be defined as on-line, computer-based storehouses of expertise, knowledge, experience and documentation about particular aspects of a corporation. We consider here only the "textual component" of corporate knowledge - i.e., all sort of economically valuable, natural language (NL) documents like news stories, telex reports, internal documentation (memos, policy statements, reports and minutes), normative texts, intelligence messages, etc. In this case, the construction of effectively usable corporate knowledge repositories can be achieved with the translation of the original documents into some type of conceptual format. The "metadocuments" obtained in this way can then be stored into a knowledge repository (knowledge base) and, given their role of advanced document models, all the traditional functions of information retrieval, e.g., searching, retrieving and producing an answer (and other functions like the intelligent navigation inside the repository) can be directly executed on them. We then illustrate here, in some details, the architecture of a prototypical system designed to exploit a knowledge repository of metadocuments. Compressed Postscript


Brian Gaines, 24-Nov-96