“ITR/IM: Perspectives on Collaborative Knowledge-Building”

Project Summary

Knowledge-Building: This project proposes to develop an approach to information management grounded in an innovative theory of learning and collaboration as knowledge-building. The theory is oriented toward guiding development of technology that can better fulfill growing societal requirements for Web-based support of groups. It offers an analysis of collaborative knowledge-building activities and of Knowledge-Building Environments (KBEs) to support these activities. To do this, it brings together and synthesizes approaches and concepts from situated learning, activity theory, hermeneutics, distributed cognition, and related theories from philosophy, social science, education, and computer science.

Knowledge-Building Environments: The proposed work centers on a new way of using computers to support collaboration by integrating information management support for various activities of knowledge-building. Such KBE software prototypes go beyond superficial discussion or chat and beyond choices and decisions among fixed options to support the co-construction of deep knowledge, innovation, and shared understanding. They support learning, working, and innovation over time within groups that may be physically distributed.

Perspectives: A key innovation for KBEs in this project is the modeling of the interplay between individuals and the group in collaborative settings. This is done by providing personal and group computational perspectives: virtual workspaces whose contents are interrelated by automated inheritance mechanisms. Computational perspectives provide a new, dynamic, personalized form of on-line information management that supports the fundamental structure of collaboration. They help to manage a shared information space so that participants view information relevant to themselves and can process (edit, rearrange, reconceptualize) that information without affecting anyone else’s personal perspective. Then, through supported negotiation activities, information is migrated to sub-group and group perspectives, where it represents shared knowledge.

Research and Training Focus: The project will directly employ four students and will involve many more in seminars and class projects. Because KBE software is essentially a new form of learning technology, education students as well as computer science and other students will be involved in designing, developing, deploying, and assessing the software. This will help to create an interdisciplinary educational technology research focus that the PI and his colleagues have already begun to foster.

Software Development: The proposed work will build an infrastructure for local, national, and international collaboration on KBE software development, providing a more reliable basis for assembling KBE prototypes customized to particular deployment sites. An open source perspectives server will be released, allowing researchers to develop KBE interface components that simply call this server for database access and dynamic computation of perspective contents. A standard for data interchange with the server and for interoperability among KBE systems will facilitate a component architecture and the use of shared tools to assess KBE usage.

Study Sites: The project will assess KBEs with perspectives in realistic study sites: an academic research group, a collaborative learning seminar, a corporate training setting, and an industrial design group. Quantitative analysis of captured textual contents will be compared with results from non-integrated threaded discussion systems and other groupware. Qualitative analysis of surveys and field notes will investigate issues of deployment, adoption, social practice, utility, and effectiveness.

Project Impacts: The proposed project should result in progress in the development and assessment of KBEs, an emerging form of software with a potential to significantly extend human cognition by supporting collaborative knowledge-building activities and by providing persistent external memory of what took place during the collaboration. The release of a perspectives server with its associated standards will provide a concrete basis for catalyzing local, national, and international collaboration among KBE researchers. This will promote KBE research as an important new research focus.

Project Description

This project is guided by a theory of collaborative Knowledge-Building Environments (KBEs) that we are developing. This theory proposes the following principles:

·        Collaborative knowledge-building is a particular view of group learning that focuses on a range of activities that take place within communities, as opposed to focusing on learning as the transmission of bits of information to individual learners.

·        Collaborative knowledge-building takes place largely through the interaction among people with different understandings from multiple personal and group perspectives.

·        Such knowledge-building within groups can be helped by appropriately designed information technology (IT) that supports various knowledge-building activities and supports interaction among alternative perspectives.

The form of IT that we are interested in – collaborative Knowledge-Building Environments – represents a distinctive approach that overlaps related work in Computer-Supported Collaborative Learning (CSCL) and Computer-Supported Cooperative Work (CSCW). IT support for learning is traditionally oriented toward the transmission of information to individual students. Even where it is based on a view of student construction of knowledge, as with Intelligent Tutoring Systems (ITS) for algebra or physics, the goal is measured by testing the incorporation of pre-defined content or methods into the individual’s understanding (Wenger, 1987) . A more student-centered, constructivist approach is taken by Interactive Learning Environments (ILE), which might, for instance, allow students to create ecologies in SimLife to learn biology, or programs in Turtle Logo to explore math concepts (Papert, 1980) . In contrast, a KBE primarily supports the group process and leaves matters of content up to the participants (which may include a teacher who raises particular content issues and helps maintain focus). In this way, it applies CSCW approaches to CSCL. A review of CSCW technology for groups (Kraemer & Pinsonneault, 1990) distinguishes group communication support systems (GCSSs) from decision support (GDSSs). GCSSs are specific communication media like email and video-conferencing. In providing computational tools for group decision making, GDSSs tend to support isolated, focused activities that integrate products of individual work. In contrast, a KBE aims to support a broad spectrum of knowledge-building activities – both individual and group – in a more seamless fashion. It supports the construction of areas of knowledge through group inquiry over extended periods of time. It also supports the interplay of individual and group more comprehensively, through integrated mechanisms of "computational perspectives" and negotiation that treat the group as more than just the sum of the individuals.

Assessments of CSCL and CSCW systems have defined a number of key issues for evaluating the problems and successes of such systems. For instance, in simple threaded discussion forums common problems include: short threads (a tendency for discussions to die quickly), low participation (lack of motivation to participate), few cross-references (little convergence of ideas), and superficial content (minimal depth of investigation) (dePaula, 1998; Guzdial & Turns, 2000; Hewitt & Teplovs, 1999) . On the other hand, GDSSs and GCSSs attempt to decrease communication barriers within the group, while increasing task-oriented focus, depth of analysis, and decision quality (Connolly, 1997; Kraemer & Pinsonneault, 1990) . Social informatics studies have raised additional issues of software deployment and adoption in addition to questions of usability and utility (Kling, 1999) . These are some of the dimensions along which KBEs must be assessed within realistic learning and working social contexts.

To date, the PI and his colleagues have begun to develop KBE theory in conjunction with Web-based KBE prototypes that support many of the activities described in the theory and that have been tested informally in collaborative learning classrooms. In particular, computational support for personal and group perspectives has been developed and tried out. Support for computational perspectives was explored in the PI’s dissertation (Stahl, 1993a) and has since been refined and adapted to the Web (Stahl, 1999a) . This work has been described in relevant CSCL and CSCW conferences (see references by Stahl). The proposed project will build on existing concepts and prototypes, extending them substantially by: implementing a technical infrastructure to support data interoperability, integration of functionalities, and rapid prototyping; deploying customized KBEs in specific study sites; observing the social impacts of these IT systems in the work settings; revising the theory based on empirical findings; and fostering a community of researchers working on IT support for knowledge-building in workgroups.

1. Preparation for Proposed Work under Prior NSF Support

1.1. Organizational Memory and Organizational Learning (CSS)

“Conceptual Frameworks and Computational Support for Organizational Memories and Organizational Learning (OMOL),” PIs: Gerhard Fischer, Gerry Stahl, Jonathan Ostwald, September 1997 – August 2000, $725,000, from NSF CSS Program #IRR-9711951.

This grant prepared much of the background for the proposed work. The OMOL project started from a model of computer support for organizations as Domain-Oriented Design Environments (DODEs) in which both domain knowledge and local knowledge are stored in the form of artifact designs and associated design rationale (Fischer, 1994) . This CSCW model evolved into one of Collaborative Information Environments (CIEs), that emphasized the interactive, asynchronous, persistent discussion of concepts and issues within an organization (Stahl, 1998; Stahl, 2000a) . Gradually, interest in organizational learning aspects led to involvement in CSCL and the model of collaborative Knowledge-Building Environments (KBEs) (Fischer et al., 1999) . A number of software prototypes were developed to explore the use of the Web as a communication and collaboration medium. Of these, the most important for the proposed work are the following:

Work on this grant led to the focus on KBEs as models of computer support for organizational memory and organizational learning. In particular, it provided a number of different systems, each with useful functionality, and brought home the need to define component standards so the functionalities can be combined more flexibly. As we tested and deployed these systems, we confronted serious issues of adoption and focused our concerns increasingly on socio-technical and social informatics (Kling, 1999) issues: motivation, media competition, critical mass, social practices, seeding, management, re-seeding, convergence of ideas, peer-to-peer collaboration, deployment strategies. These issues led to a new research agenda (Stahl, 1999b) and this proposal.

1.2. WebGuide and Environmental Perspectives (NOAA)

“Collaborative Web-Based Tools for Learning to Integrate Scientific Results into Social Policy,” PIs: Ray Habermann, Gerry Stahl, November 1998 – July 1999, $89,338, NSF, #EAR-9870934.

This grant funded the initial implementation of WebGuide as an integrated Java applet KBE supporting personal and group Perspectives. It was a joint effort between the PI, a middle school teacher, and a research group at the National Oceanographic and Atmospheric Administration (NOAA) labs in Boulder. The teacher taught an environmental science class in which he wanted to spend the year having his students interview various adults and construct a set of contrasting perspectives (conservationist, regulatory, business, community) on a particular local environmental issue that the students had previously been involved in. WebGuide was used by the students to collect notes on their interviews and to formulate personal and team perspectives on the issue. Results of this software trial were analyzed and presented at conferences (Stahl, 1999a; 1999b; 1999c; Stahl & Herrmann, 1999) . These findings led to a number of revisions of WebGuide, including the separation of the Perspectives mechanism from the Web interface, and recognition of the need for software architectures, standards, and components to support flexible rapid prototyping of KBEs.

1.3. Collaboration in KBEs (CILT)

“Interoperability Among Knowledge-Building Environments,” PI: Gerry Stahl, September 1999 – August 2000, $9,124.21, from NSF-funded Center for Innovative Learning Technology (CILT), Subcontract #17-000359 under NSF grant #EIA-9720384.

This is a current seed grant whose purpose is to stimulate collaboration among KBE research groups. Part of the intention of the grant was to prepare a proposal for fuller funding, such as the present proposal and its currently pending complementary NSF proposals for “IT Support for Knowledge-Building in Workgroups” and “ITR/EWF: Collaborative Research on Knowledge-Building Environments: Growing a National and International Research Community for Distance Learning Information Technology.” This grant has already resulted in a semester-long student project involving three graduate and three undergraduate students (one collaborating virtually from Germany) creating an XML DTD that defines a data format for data imported from several different KBE prototypes and displayed in a Web browser using XSL. The grant supported a workshop entitled “Collaborating on the Design and Assessment of KBEs in the 2000's” at CSCL ’99 at Stanford. This workshop attracted over 60 participants and was preceded by an on-line discussion of 28 submitted position papers. This grant has led to the emphasis on collaboration among KBE research groups and the need to put into place some of the technical and social conditions for such collaboration (Stahl, 1999a) , as proposed here.

2. Overview of Proposed Work

2.1. Theory of Collaborative Knowledge-Building Environments

Collaborative Knowledge-Building

Information Technology (IT) is a broad field that can be conceptualized in various ways. Traditionally, the computer was thought of as a medium for storing and delivering data, that can then be used by people in their work. More recently, the computer (especially with the Web) has become a medium of communication, through which people share information and knowledge. This communication can take a variety of forms. In simple forms of e-commerce or on-line voting, people submit their decisions about a fixed list of choices. In chat and most email, people exchange greetings and opinions, generally without changing those opinions. Many systems in recent years have tried to support a particular form of communication or social interaction like brainstorming or decision-making – often with very positive results (Connolly, 1997; Vogel et al., 1987) . We are interested in a distinct but broader process of communication which we term collaborative knowledge-building. Here, groups of people construct new knowledge through interaction of their ideas and perspectives, usually eventually preserved in documents or other artifacts.

Our theory of collaborative knowledge-building (Fischer et al., 1993/1998; Stahl, 1975; 1993a; 1993b; 1999c; 2000a; 2000b; Stahl & Herrmann, 1999) proposes a concept we call the synergistic moment; we intend to investigate the validity of this concept in the proposed project. The synergistic moment is the critical point during collaboration in which a group constructs meaning that transcends what any participant may have “in mind.” The shared understanding that is generated in this process is a subtle phenomenon: It does not mean that everyone is in complete agreement or even that each individual has the same internal cognitive representations of what is discussed. Rather, it means that a certain group view has been expressed. The unit of analysis for describing this is the group, and is manifested in the group's discourse. Individuals may agree to disagree with the group understanding, and careful investigation may reveal that individual understandings differ from the group's view (Hatano & Inagaki, 1991) . The intersubjective "sharing" is not a correspondence or overlapping of individuals' mental content, but a coordination or interaction of their participation in joint socio-cultural activity (Matusov, 1996) .The synergistic moment is an emergent property of the group dialog as a cacophony of voices (Bakhtin, 1986) . It could easily pass unnoticed as a magical fount of creativity; to more deeply understand it likely requires "thick description" (Geertz, 1973) and detailed interaction/discourse analysis (Jordan & Henderson, 1995) , and therefore presupposes that the interaction was captured in some medium. Fortunately, the literature on CSCL contains a number of incisive analyses (Roschelle, 1998) of the synergistic moment, although they do not highlight it as such.

The synergistic moment is a result of perspective-sharing (Boland & Tenkasi, 1995) , but at the group rather than the individual level. It overcomes the problem pointed out by Feltovich et al. (Feltovich et al., 1996) , that any one perspective may limit the ability to comprehend creatively the complexity of a topic under discussion. What typically happens is that one person makes a statement from her personal perspective; someone else interprets that statement from his own perspective and responds accordingly; others continue this process so that the discourse consists implicitly of reinterpretations from various perspectives. The drive to establish intersubjectivity and shared knowledge is powered by socio-cognitive conflict and contention among perspectives according to studies by Piaget and his followers (Perret-Clermont & Schubauer-Leoni, 1981) . The dialog proceeds through sequential turn-taking and attempts to repair “misunderstandings” as understood from particular perspectives and reinterpreted from others. Thanks to the human drive to impose coherent social meaning structures (Geertz, 1973) , a synergistic group understanding emerges. This shared understanding can play a central role in the further activity of the group and can be more or less adopted by individuals into their personal perspectives. Although the synergistic moment seems to the participants to emerge spontaneously, it can be understood as the result of many identifiable knowledge-building activities, as represented in our model (below).

Perspectives in Knowledge-Building

According to hermeneutics – the philosophy of interpretation – human understanding is fundamentally perspectival. We construct knowledge from our situated perspective in the world: our historical position, cultural tools, and practical interests (Gadamer, 1960/1988; Heidegger, 1927/1996; Stahl, 1975) . Computational support for knowledge-building can represent our interpretive perspectives with computational Perspectives (Boland & Tenkasi, 1995; Nygaard & Sørgaard, 1987; Winograd & Flores, 1986) . (In this proposal, Perspective–with-a-capital-P will refer to the proposed computational mechanism that mirrors human interpretive perspectives-with-a-lower-case-p.)  In this sense, Knowledge-Building Environments (KBEs) with computational Perspectives are designed to support the essential structure of collaboration. A key hypothesis of the proposed work is that KBEs benefit from an approach that represents the perspectival nature of collaboration. A goal of the project is to facilitate the incorporation of a computational Perspectives mechanism in KBEs – both in our own prototypes and in the work of other KBE research groups around the world.

Computational Perspectives have been explored by the PI in a number of software prototypes, in his dissertation system, and in his theoretical publications (Stahl, 1993a; 1993b; 1995; 1998; Stahl & Herrmann, 1999; Stahl et al., 1995) . In a single-user system, computational Perspectives may correspond to different domains or professional viewpoints on a design problem, such as electrical, plumbing, structural, and heating concerns in architecture (Fischer et al., 1993; 1993/1998) . In a KBE to support collaboration, computational Perspectives typically provide personal or group workspaces for the development of different sets of ideas. In this way, they can model the relationships among the various personal and group interpretive perspectives at work in the construction of collaborative knowledge.

We hypothesize that computational Perspectives can support the synergistic moment in collaborative knowledge-building by providing the necessary contact among different personal Perspectives, allowing them to interact, and then locating the results in a group Perspective. By situating the traditionally ephemeral synergistic moment within an explicit structure of computational Perspectives and by doing so in a persistent way, a KBE provides new opportunities for group self-reflection.

An important complement to Perspectives is negotiation. Negotiation is a process through which divergent personal perspectives converge on a collaborative shared understanding. When Perspectives and negotiation are effectively “intertwined” in a KBE, they compensate for each other’s potential problems: Negotiation converges ideas so that everyone can benefit from the ideas of other perspectives, while personal Perspectives allow people to work on their own views while potentially time-consuming negotiations are underway (Stahl & Herrmann, 1999) .

For instance, when WebGuide – a KBE with computational Perspectives implemented by the PI – was used in a middle school environmental science classroom, students each had their own personal Perspective in which to develop their own responses to questions posed by the teacher. The teacher’s questions to the whole class were posed in the class’ group Perspective. From there they were automatically inherited into the team Perspectives. The content of the team Perspectives was, in turn, inherited into the personal Perspectives of team members. Gradually, students migrated their ideas to team Perspectives that represented either conservationist, governmental, corporate, or citizen perspectives on the ecological controversy – depending on which perspective team the student was part of. Then they could work with the ideas of their team-mates and negotiate their team position. In the end, the different teams negotiated to spell out agreements and disagreements (Stahl, 1999c) .

The analysis of the synergistic moment suggests that negotiation need not take the explicit, rationalist forms typical of GDSSs, such as voting. Group results may emerge naturally out of the intertwining of Perspectives in group discussion. A challenge of the proposed work will be to develop software support for capturing such results and migrating them un-intrusively to group Perspectives.

The Potential of IT Support for Knowledge-Building

IT support has the potential of transforming the activities underlying the synergistic moment. For one thing, it would make those activities publicly accessible. The group could then reflect upon the emergence of its shared understanding by looking over the persistent record of its dialog. Such reflection might prove especially useful in contentious situations or for newcomers who were not part of the original dialog and are motivated to re-open the issue – as illustrated by Matusov (1996) . Furthermore, computer support of perspectives could make explicit the interplay of different personal Perspectives and the migration of ideas and their interpretations between personal and group Perspectives. Ironically, perhaps, the “asynchronous” medium of the Web would allow group members to interact simultaneously – without waiting for sequential turns – thereby overcoming what Peters (1998) characterizes as “the hardest argument against democracy: the ability of only one person to speak and be heard at a time” (p. 261). Of course, as we have already discovered with the Web in general, the increased flood of ideas raises complex information management issues. We do not yet understand the full social impact of the envisioned KBEs – we will only know how they are used once they have been implemented, deployed in naturalistic settings, and observed.

Theories of human cognitive development emphasize the important role of external memories to extend short-term and long-term human memory (Donald, 1991; Norman, 1993) . They also stress that individual cognition is a social product, highly mediated by social symbol systems, cultural artifacts, processes of structuration, and group collaboration (Bourdieu, 1972/1995; Geertz, 1973; Giddens, 1984; Hutchins, 1996; Vygotsky, 1930/1978) . This suggests that computer support for collaboration has the potential to significantly advance the power of human cognition. In addition to maintaining a persistent external memory, IT can help people to be more reflective and creative – as has been demonstrated in computer support for brainstorming and decision-making (Connolly, 1997; Vogel et al., 1987) . However, as our research to date indicates, despite the fact that the Web seems to offer a promising technological base for such a development, computer-supported collaboration is a complex process that requires a sophisticated body of knowledge that we are just beginning to assemble. Moreover, the potential is beyond the reach of any single research group.

We believe that IT support for collaborative knowledge-building has not yet been developed to near its potential. KBE research has been carried on for over a decade now, starting with the CSILE system and continuing with KIE, CoVis, etc. (Cuthbert, 1999; Pea, 1993; Scardamalia & Bereiter, 1991; 1996) . Recently, commercial systems like KnowledgeForum, WebCT, and LearningSpace are catching on. However, as yet there has been no systematic attempt to support the variety of activities that are involved in knowledge-building. There is no general theory of collaborative knowledge-building as a social process. Existing research tends to target specific contexts like middle school science with specialized closed systems, rather than developing interchangeable, open source components that can be applied in a full range of contexts. While networks of KBE researchers are coming together in other countries, there is little organized effort to collaborate in the US. The proposed project aims to change this situation. However, collaboration across institutions cannot be started by just wishing for it. This project tries to put some of the necessary conditions in place by developing technical infrastructure (standards, a Perspectives Server, interface components) and initial results that can be used to stimulate discussion and collaboration among KBE researchers locally, nationally, and internationally. Interoperability and collaboration will allow isolated advances to be exchanged, new functionalities to be shared, and test data to be compared.

A Model of Collaborative Knowledge-Building

One approach to better understanding how to design computer support for collaborative knowledge-building in social settings is to conceptualize the various constituent activities involved in individual and social knowledge-building. The diagram below from (Stahl, 2000b) provides a starting point for this, combining aspects of activity theory, situated learning, hermeneutic philosophy, and distributed cognition theory (Chaiklin & Lave, 1993; Cole, 1996; Engeström et al., 1999; Gadamer, 1960/1988; Hutchins, 1996; Lave & Wenger, 1991; Nardi, 1996) .

The idea of this diagram is that knowledge-building can proceed through many different activities. The sequential structure of the model is only illustrative of an ideal conceptualization. We understand that these activities complexly overlap in practice. The possible relationships among the individual activities – and particularly the interactions between the personal and social – can be complex and varied. The purpose of the diagram is to suggest a number of distinct activities that could be supported by a KBE with multiple functionality. The sequential labeling of these activities corresponds to proposed KBE components listed in Table 1 below, and it is not intended to imply a necessary order to the activities.

Figure 1. A model of personal understanding and social knowledge-building.

A set of seminal books and articles in Computer-Supported Collaborative Learning (CSCL) has formulated a view of learning as a social process of collaborative knowledge-building within communities of practice (Brown & Campione, 1994; Brown & Duguid, 1991; Lave, 1991; Lave & Wenger, 1991; Pea, 1993; Scardamalia & Bereiter, 1996; Wenger, 1998) . However, these texts do not make the set of cognitive and social activities that underlie such a view explicit in the manner attempted in our KBE theory.

Starting in the lower left corner, Figure 1 shows a cycle of personal understanding. The rest of the diagram depicts how personal beliefs can be articulated in language and become part of social interaction. Note that the results of social knowledge-building eventually feed into personal understanding, providing the evolving toolkit of culturally-based individual cognitive capabilities. The depicted knowledge-building activities are discussed briefly below in the context of proposed computer support.

IT Support for Knowledge-Building Activities

Each of the activities of social knowledge-building pictured in Figure 1 can be supported computationally. Table 1 lists an illustrative form of support for each. It also lists corresponding prototypes that we have developed. Support for each activity is briefly discussed following the table.

Table 1. Forms of computer support for knowledge building activities.

(a)    Computer support should facilitate the process of articulating ideas and preserving them in convenient forms. Most KBEs, including discussion forums like DynaClass, provide an editor for articulating ideas. Some KBEs have tried to introduce procedural facilitation, scaffolding, or prompting to encourage someone to articulate an appropriate expression (Slotta & Linn, 2000) . Other approaches would be to provide an outline editor or a brainstorming area.

(b)    Public statements by one person confront those of other people. Computer support can represent the different perspectives from which these statements emerge. Perspectives are more general than representations of individuals themselves, because one person can offer statements from multiple perspectives and several people can agree on a common perspective. Perspectives can be related to one another, for instance deriving from a common perspective that they share. Computational representations of perspectives in a KBE like WebGuide (Stahl, 2000a) make explicit the important relationships among personal and group perspectives, as well as providing means for individuals and collaborative teams to articulate their own perspectives.

(c)    A KBE with support for Perspectives should provide comparison Perspectives, in which one can view and contrast alternative Perspectives and adopt or adapt ideas from other people's Perspectives. Comparison Perspectives in WebGuide aggregate ideas from various individual and/or group Perspectives and allow for comparison of them (Boland & Tenkasi, 1995; Stahl, 1999c) . Other systems like D³E (Sumner & Buckingham Shum, 1998b) facilitate commentary on documents by other people, such as reviews of journal articles.

(d)    The most common element in current KBEs is the discussion forum. This is an asynchronous, interactive communication system like DynaClass that allows people to respond to notes posted by one another. Typically, there is a thread of responses to entered notes, with a tree of divergent opinions. A KBE should go beyond superficial undirected discussion to converge on shared understandings (dePaula, 1998; Guzdial & Turns, 2000; Hewitt & Teplovs, 1999) .

(e)    Although every note in a discussion forum is a response to another note, the discussion may have a more complex implicit structure. One note might argue for or against another or provide evidence to back up the claim of another note, for instance. Such an argumentation structure can be made explicit and formalized in a representation of the argumentation graph. A component like InfoMap that displays the structure of notes graphically can contribute to participants' meta-level comprehension of their knowledge-building activity, pointing out where additional evidence is needed or where alternatives have not been explored (Buckingham Shum & Hammond, 1994; Donath et al., 1999; Suthers, 1999) .

(f)    An important requirement for constructing group knowledge is the establishment of shared understanding. This can be fostered by clarifying the meaning of important terms used in various competing claims. A glossary discussion can make explicit how different participants understand the terms they use, as in DynaGloss or DocReview (Hendricksen, 1999) .

(g)    The glossary discussion should result in a group glossary of the agreed upon definitions of important terms. Such a glossary already represents a form of group knowledge. The glossary is, of course, subject to future debate and emendation; it may make sense to define the glossary as a particular display of information from the glossary discussion (Stahl & Herrmann, 1999) .

(h)    Perhaps the most delicate phase of knowledge-building is negotiation. Computer support of negotiation tends by nature to make explicit the factors entering into the negotiation process. This can be extremely harmful to the subtle processes of persuasion if not done sensitively. On the other hand, negotiation is critical to helping multiple perspectives to converge on shared knowledge. Computer support can provide a useful tool – as long as it is carefully integrated with other social activities that allow for implicit, culturally established interpersonal interactions (Stahl, 1999b) . Group Decision Support Systems (GDSSs) have traditionally been independent systems, not integrated with the broader context of knowledge-building (Kraemer & Pinsonneault, 1990; Vogel et al., 1987) .

(i)     The accumulation of negotiated shared knowledge results in the establishment of a group perspective. Like the alternative individual and team (or subgroup) perspectives, the group perspective may be represented in a KBE. In WebGuide, the content of the group Perspective is inherited into the individual and team Perspectives, because it has been accepted by the group. Individuals can then build on this shared knowledge within their own Perspective and even begin to critique it and start the whole cycle over (Stahl, 1999a) .

(j)     Shared knowledge can be further formalized. It can be represented in another symbolic system or combined into a more comprehensive system of knowledge (Stahl, 1999c) . For instance, in academic research knowledge is incorporated in new classroom lectures, conference presentations, journal articles, and books. The discussion of knowledge that has been compiled into publications can be carried out in a bibliography discussion component of a KBE.

(k)    Finally, representations of the new shared knowledge in publications and other cultural artifacts are themselves accepted as part of the established paradigm. Although still subject to occasional criticism, ideas in this form more generally provide part of the accepted base for building future knowledge. In academic circles, an annotated bibliography like Sources might provide a useful KBE component to support this knowledge building activity (Sumner & Buckingham Shum, 1998a) .

A KBE goes beyond a single-purpose system – like a simple discussion forum – and supports more than one collaborative knowledge-building activity (Muukkonen et al., 1999) . It retains a record of the knowledge that was incrementally collected – unlike common chat, newsgroup, and listserv systems that erase contributions after a short period of time. We hypothesize that it should help people to express their beliefs, to discuss them with others, to differentiate their own perspectives and adopt those of other people, clarify disagreements or misunderstandings, critique and explicate claims, negotiate shared understandings or agreements, and formulate knowledge in a lasting representation. Because KBEs are computational, they can provide facilities like searching, browsing, filtering, tailoring, and linking in order to group related ideas together automatically. KBEs can interface with other agents and software utilities – for instance sending emails to notify collaborators when important knowledge-building events have taken place (McLean, 1999) . They can also dynamically format sets of notes in convenient displays for different purposes.

2.2. An Infrastructure for KBEs

Computational Perspectives

Computational Perspectives provide a new, dynamic, personalized form of on-line information management (Stahl, 1995) . A Perspective defines an electronic workspace in which a person or group can develop ideas and manage information that belongs together – for instance because it represents the beliefs and viewpoint of a particular person, group, domain, or intellectual position. Perspectives structure a shared information space so that special coherent views can be built up and displayed. Although the mechanism of computational Perspectives is very general and flexible, the simplest way to use it in a small group is to define a personal Perspective for each member, one team Perspective for agreed upon ideas, and a comparison Perspective that collects the ideas from all the personal Perspectives.

The design philosophy behind computational Perspectives as implemented by the PI in WebGuide is that users have complete control over the content in their personal Perspectives. Thus, if my personal Perspective inherits conflicting ideas from different team Perspectives that I belong to, I can delete, edit, and rearrange those ideas at will. Other users can view the contents of my personal Perspective (except for content that I have designated as private) and they can copy items, link to them, initiate public discussions of them, and propose them for incorporation in team Perspectives – but none of this affects how the content of my Perspective is displayed to me. This allows me to build my own Perspective on the topics that are under consideration by the group. I can see what knowledge others are building, incorporate that knowledge into my Perspective, or join in with others to share, discuss, and negotiate. The same design philosophy applies of course to team Perspectives: team members jointly (through negotiation processes) have complete control over the content of their team Perspective.

Inheritance is a central defining mechanism of computational Perspectives as used in this proposal. The ability to define arbitrarily complex networks of Perspectives with multiple layers of sub-groups between the group Perspective and the individual personal Perspectives, and to have the automatic inheritance of content through the network distinguishes this approach from all other systems of “views” and “perspectives.” Inheritance in this sense is not class inheritance, but “content inheritance.” A given Perspective can inherit content from multiple other Perspectives. This content is aggregated (logical union) in the given Perspective, where it can be over-ridden with edits, deletions, rearrangements, and additions. The inheritance mechanism is derived from efficient approaches explored in hypermedia, including “delta memory” and “transclusion” (Boborow & Goldstein, 1980; McCall et al., 1990; Mittal et al., 1986; Nelson, 1981; Nelson, 1995)