Gerhard Fischer
University of Colorado
Center for LifeLong Learning and Design (L3D)
Department of Computer Science and Institute of Cognitive Science
Campus Box 430
Boulder, CO 80309-0430
Phone: (303) 492-1592
Fax: (303) 492-2844
E-mail: gerhard@cs.colorado.edu
WWW: http://www.cs.colorado.edu/~gerhard/
Summary Statement for Grant IRI-9311839

Human-Centered, Intelligent Agents Supporting Communication and Collaboration in Domain-Oriented Design Environments


Project Summary
Goal. There is a growing awareness among researchers and users of computer systems that systems based on a ³tool² metaphor do not scale up to the information-rich, high-functionality systems of the future. The goals of our project are (1) the exten sion of theoretical and conceptual frameworks and (2) the design, implementation, and evaluation of prototype systems in order to accommodate the need for and to evaluate the contribution of intelligent agents for supporting communication and collaboratio n within domain-oriented design environments.

Objectives. Our research is exploring the embedding of intelligent agents into domain-oriented design environments with the goals of reducing the cognitive load on designers through active behavior and improving the quality of the designed artifact . Agents could, for instance, help designers avoid overlooking important possibilities and settling on suboptimal plateaus. Incorporating intelligent agents into design raises numerous conceptual, technical, and social issues. These issues are a consequ ence of mixed-initiative dialogs made possible by agents. Briefly, the conceptual issues we are investigating include shared context, control of initiative and intervention, and focus of attention. Technical problems include user manipulation of agents t hrough an agent editor, activation of agents in a shared context, presentation of agents, and creation of shared context through specification, construction, task representations, and interaction histories. Social issues addressed by our research include the new role distributions between humans and computational agents, namely, the embedding of agents in human-centered cooperative problem solving systems supporting communication, coordination, and collaboration. These issues are being addressed by provid ing extended task representations and accountability through explanation.

Focus and Domain. The focus of our investigation is on indirect, long-term collaboration in the design and evolution of complex artifacts. The primary domain we are using is that of local area networks. Plans for networks need to be understood fo r long-term evolution (maintenance, enhancements, and modifications) by designers who were not necessarily present during the original design and implementation. A focus on design and specific design domains provides unique opportunities for investigating agent-based architectures. First, the behavior of agents can be tuned (made ³intelligent²) by a representation of a shared context that, in design, can rely on interpretations restricted by domain-orientation, partially constructed artifacts, specificati ons, and interaction histories. Second, agent-based architectures have the potential for supporting the representation and incremental accumulation of information over the course of a design. Third, in design, agents support communication and collaborati on among humans, requiring a human-centered approach rather than a system-centered approach to information representation and delivery.

Over the last few years, we have analyzed numerous high-functionality systems and information repositories and have (almost universally) identified usage patterns reflecting ignorance with respect to information in a design domain. Based on our analyses of these high-functionality, information-rich systems, we have identified the following problems users have in accessing information related to a design task: (1) users do not have well-formed goals and plans with respect to their task and, consequently, with respect to accessing information for the task; (2) users do not know about the existence of information relevant to their task; (3) users do not know how to access information they need; (4) users do not know when to apply information they do find; and (5) users do not understand how to combine and adapt information according to the needs of their task. Based on a case study, we developed the analogy that passive information repositories are like discount stores: all the items are present but there are no knowledgeable salespersons to connect customers with the items they need and explain how to use them. Our goal is to introduce computer-based agents that provide the support analogous to experienced salespersons.

Relationship to Previous Research. Our current research project builds on the research work and the prototype system developed in our previous research project, ³Supporting Collaborative Design with Integrated Knowledge-Based Design Environments,² which led to the creation of large, passive information repositories (such as artifact memories and institutional memories). Empirical evaluations have shown that the existence of information is not good enough. In this grant we address the observed short comings by creating active repositories, by making designers aware of information relevant to the task at hand, and by complementing information access with information delivery.

Large, Passive Information Repositories. Large amounts of information are a natural consequence of design. Catalogs of previous designs, lists of design components, texts of design rationale, specifications, and codified constraints are examples of some of the types of information that accumulates. Although design information is accumulated with the intent of helping future designers either to modify a previous design or to start a new project, the right information often fails to reach the design er at the right time. As a consequence, active information stores are needed to place relevant information in the forefront, serving to remind, warn, and otherwise advise a designer working on a specific task. As early as the time of the envisioning of t he Memex system. Vannevar Bush predicted that the great rate of expansion of scientific literature would make it increasingly difficult to find the relevant information. The realization that the mere existence of information is insufficient is becoming ob vious in an increasing number of domains.

Limitations of Information Access. Browsing and query-oriented schemes have long served as the principal techniques for helping people retrieve information in many applications, including systems for providing on-line help and design rationale. Ho wever, these conventional retrieval techniques do not scale up to large information stores. More innovative schemes such as query by reformulation and information filtering have been proposed. Unfortunately, the problem remains that users simply will no t actively search for information when they are unaware that they need the information or that relevant information even exists. Thus, to assist users in making full use of large information repositories, information access methods need to be complemente d by information delivery methods.

Achievements
Specific work is being carried out in the areas of the development of a conceptual framework, system building, and assessment studies‹this work is centered around two systems:

  1. 1. BASAR, a framework for Building Agents Supporting Adaptive Retrieval. In this approach, we have adopted indirect management as the fundamental model of interaction, since it integrates active software compo nents (agents) supporting users in doing their tasks at hand: both humans and the computer initiate communication, monitor events, and perform tasks instead of unidirectional interaction via commands and/or direct manipulation. Based on this model, we hav e developed a taxonomy for software agents and a conceptual framework that integrates these agents in the World-Wide Web. For details see: http://zeus.gmd.de/projects/basar.html/

  2. 2. GIMMe, a Group Interactive Memory Manager system. This system aims at collecting the distributed knowledge of a community of practice (e.g. project team), making this knowledge available to all group members, and in turn ³activating² new knowledge through the application of the stored knowledge to new problem situations. Intelligent agents automatically provide the user with pointers to relevant knowledge in the group memory. For details see: http://www-l3d.cs.colorado.edu/~stefanie/GM_Info.html

The major new aspect of our research is the understanding of a need to move from a primarily individual perspective to an organizational perspective. We are beginning efforts to develop and study a form of organizational memory based on our model of domai n-oriented design environments.

In an effort aimed at gaining a deeper and broader understanding of the research issues associated with this shift, we organized a research symposium in May 1996 entitled ³Computational Support for Continually Evolving Organizational Knowledge Bases,² whi ch brought together a dozen of the leading researchers in organizational memory and organizational learning (for details, see: http://www.cs.colorado.edu/-~ostwald/symposium/symposium .html) and we participated in a workshop at the CSCW¹96 conference entitled "CSCW and Organizational Learning" [Lindstädt 1996a; Lindstädt 1996b].


Major Publications

  1. C. Thomas and G. Fischer, "Using Agents to Personalize the Web," Proceedings of the 1997 International Conference on Intelligent User Interfaces (Orlando, Florida), ACM, New York, NY, 1997, pp. 53-60.

  2. G. Fischer, S. Lindstaedt, J. Ostwald, M. Stolze, T. Sumner, B. Zimmermann, "From Domain Modeling to Collaborative Domain Construction," In Proceedings of DIS¹95, Symposium on Designing Interactive Systems, Ann Arbor, MI, 1995, pp. 75-85.

  3. G. Fischer, S. Lindstaedt, J. Ostwald, K. Schneider, and J. Smith, "Informing System Design Through Organizational Learning," Proceedings of the Second International Conference on the Learning Sciences, July 1996, Northwestern University, Evans ton/Chicago, published by: Association for the Advancement of Computing in Education, pp. 52-59.

  4. G. Fischer, "Seeding, Evolutionary Growth and Reseeding: Constructing, Capturing and Evolving Knowledge in Domain-Oriented Design Environments," Proceedings of IFIP WG 8.1/13.2 Joint Working Conference, In A. Sutcliffe, D. Benyon, F. van Assche (eds.), Domain Knowledge for Interactive System Design, IFIP Series, Chapman & Hall, London, Geneva, Switzerland, May 1996, pp. 1-16.

  5. G. Fischer, "Distributed Cognition, Learning Webs and Domain-Oriented Design Environments," Proceedings of the Conference on Computer Supported Collaborative Learning (CSCL¹95), Indiana University, October 1995, pp. 125-129.

  6. G. Fischer, R. McCall, J. Ostwald, B. Reeves, F. Shipman. "Seeding, Evolutionary Growth, and Reseeding: Supporting Incremental Development of Design Environments,"Human Factors in Computing Systems, CHI'94 Conference Proceedings (Boston, MA), 1 994, pp. 292-298.

  7. G. Fischer, A. C. Lemke, R. McCall, A. Morch. ³Making Argumentation Serve Design," in Moran/Carroll Book

  8. G. Fischer, K. Nakakoji, J. Ostwald, G. Stahl, T. Sumner. "Embedding Critics in Design Environments," The Knowledge Engineering Review Journal, Vol. 8, No. 4, December 1993, pp. 285-307.



Return to ITO Workshop Abstracts

Return to ITO Workshop Home Page