WebGuide: Guiding Collaborative Learning
on the Web with Perspectives

Gerry Stahl

Center for LifeLong Learning & Design

and Institute of Cognitive Science

University of Colorado

Boulder, CO, USA

Gerry.Stahl@Colorado.edu
www.cis.drexel.edu/faculty/gerry

Abstract: We are developing a Web-based tool called WebGuide to mediate and structure collaborative learning. This software uses an innovative mechanism to define a flexible system of perspectives on a shared knowledge construction space. WebGuide provides an electronic and persistent workspace for individuals and teams to develop and share distinctive points of view on a topic. We are designing the software and associated usage practices by trying it out in a middle school classroom and an advanced graduate seminar. Our experience in these use situations has raised a range of questions concerning theoretical and practical issues, which are driving our research. This paper is a reflection on what we are learning collaboratively about how software artifacts can mediate learning and shared cognition.

Keywords: perspectives, collaborative learning, computer support, knowledge building, artifact

Commentaries: All JIME articles are published with links to a commentaries area, which may include elements from the article’s original review debate. Readers are invited to make use of this resource, and to add their own commentaries. The authors, reviewers, and anyone else who has ‘subscribed’ to this article via the website will receive e-mail copies of your postings.

1. Introductory Narrative

For some years now I have been interested in how to personalize the delivery of information from knowledge repositories to people based on their preferred perspectives on the information (Stahl, 1995; 1996). For instance, designers often critique an evolving design artifact from alternative technical points of view; different designers have different personal concerns and styles, requiring considerations based upon access to different rules of thumb, rationale, constraints, standards and other forms of domain knowledge. Computer design environments should support these important interpretive perspectives (Stahl, 1993a; 1993b). I am now primarily interested in applying similar mechanisms of perspectival computer support within contexts of collaborative learning (Stahl, 2000).

Last year, Ted Habermann – an information architect at NOAA who makes geophysical data available to school children over the Web – suggested to me that we try to develop some computer support for a project at his son’s middle school. Dan Kowal, the environmental sciences teacher at the Logan School for Creative Learning in Denver, was planning a year-long investigation of alternative perspectives on the issue of “acid mine drainage” (AMD) – the pollution of drinking water supplies by heavy metals washed out of old gold mines. The fact that Dan and I were interested in “perspectives” from different perspectives seemed to provide a basis for fruitful collaboration. Ted obtained NSF funding for the project and we all spent last summer (1998) planning the course and its perspectives-based software. Each of us brought in colleagues and worked to create a Java application (WebGuide), a set of auxiliary web pages, a group of adult mentors representing different perspectives on AMD and a course curriculum.

The class started in September and the software was deployed in October. The students in Dan’s class were aware of the experimental nature of the software they were using and were encouraged to critique it and enter their ideas into WebGuide. Feedback from these twelve-year-old students provided initial experience with the usability of WebGuide and resulted in a re-implementation of the interface and optimization of the algorithms over Christmas vacation.

In January 1999, I organized an interdisciplinary seminar of doctoral students from cognitive, educational and computational sciences to study theoretical texts that might provide insight into how to support collaborative learning with perspectives-based software. The seminar uses WebGuide as a major medium for communication and reflection, including reflection on our use of the software. This provides a second source of experience and raises a number of issues that will need to be addressed in software redesign this summer.

In this paper I would like to begin a reflection on the issues that have arisen through our WebGuide experiences because I think they are critical to the ability to support collaborative learning with computer-based environments. The potential for computer mediation of collaboration seems extraordinary, but our experience warns us that the practical barriers are also enormous. Certainly, our experiences are not unique, and similar projects at the universities of Toronto, Michigan, Berkeley, Northwestern, Vanderbilt, Georgia Tech, etc. have run into them for years. Indeed, we observed many of these issues in a seminar last year prior to the implementation of WebGuide (dePaula, 1998; Koschmann & Stahl, 1998). However, I believe that perspectives-based software addresses or transforms some of the issues and raises some of its own.

Now let me describe how computer support for perspectives has evolved in WebGuide. I will first discuss the preliminary implementation as used in Dan’s middle school environmental course and explain how perspectives are supported in that version. A number of design issues led to an extended attempt to bring theory to the aid of reflection on practice. This included a graduate seminar that used a revised version of WebGuide. Finally, following this paper is a slightly condensed version of the dialog that took place between the JIME reviewers and the author, where responses from Winter 2000 and Spring 2001 bring in reflections from subsequent design iterations.

2. Practice I: Environmental Perspectives

An early implementation of WebGuide is in use in Dan’s classroom at the Logan School. For the past five years, his class of middle school students has researched the environmental damage done to mountain streams by “acid mine drainage” from deserted gold mines high in the Rocky Mountains above Denver. The students actually solved the technical problem at the source of a stream coming into Boulder from the Gamble Gulch mine site by building an artificial constructed wetlands area to filter out heavy metals. This year they are investigating the broader ramifications of their success; they are looking at the social issue of acid mine drainage from various alternative – and presumably conflicting – perspectives. The students interview adult mentors to get opinions from specific perspectives: environmental, governmental, mine-owner and local landowner. Then, working in teams corresponding to each of these perspectives, they articulate the position of their perspective on a set of shared questions.

The “Gamble Gulch” application of WebGuide serves as the medium through which the students collaboratively research these issues with their mentors and with teammates. Each student and mentor has their personal display perspective, and their display perspectives each inherit from one of the content-based team perspectives (environmental protection, governmental regulation, etc.), depending upon which intellectual perspective they are working on constructing.

Figure 1 shows one student’s (Blake) personal perspective on the class discourse. The tree of discussion threads was “seeded” with question categories, such as “Environmental Analysis Questions.” Within these categories, the teacher and I posted specific questions for the students to explore, like, “Do you believe that AMD is a serious threat to the environment?” Here, Blake has sent an email to a mentor asking for information related to this question. Email interactions happen through WebGuide and are retained as notes in its display perspectives. When replies are sent back, they are automatically posted to the discussion outline under the original email. When someone clicks on a title, the contents of that note are displayed in an HTML frame below the applet (as is the body of the student’s email in Figure 1).

Blake is working in his personal perspective, which inherits from the Class, Student team and Landowner team perspectives (see the red arrows in Figure 2). Note that the display of his personal perspective (in Figure 1) includes notes that Dan and I entered in the Student perspective to structure the work of all the students. Blake can add, edit and delete ideas in his perspective, as well as sending email in it. Because he is a member of the landowner team and the student group as well as the class, he can browse ideas in the Student comparison, the Landowner comparison and the Gamble Gulch class comparison perspectives (see list of perspectives accessible to him on the right of Figure 1).

For this application, the teacher has decided that perspective comparing and negotiation will take place in live classroom discussions, rather than in WebGuide. After a team or the whole class reaches a consensus, the teacher will enter the statements that they have agreed to into the team or class perspective.

The goal of the year-long course is not only to negotiate within teams to construct the various positions, but also to negotiate among the positions to reach consensus or to clarify differences. Dan designed this class – with its use of WebGuide – to teach students that knowledge is perspectival, that different people construct views, compilations of facts and arguments differently depending upon their social situation. He hopes that his students will not only learn to evaluate statements as deriving from different perspectives, but also learn to negotiate the intertwining of perspectives to the extent that this is possible.

3. Computer Support of Perspectives

The term “perspectives” is over-loaded with meanings; this frequently produces confusion even when it is intended to tacitly exploit aspects of the perspectives metaphor from one domain into another. It may be helpful at this point to distinguish three types of perspectives: literal, figurative and computational.

• Literal perspectives are optical or perceptual orientations: one sees objects from the specific angle or vantage point of the physical location of one’s eyes.
• Figurative perspectives take metaphorical license and refer to, for instance, different ways of conceptualizing a theme, as in adopting a skeptical view of a conversational claim.
• Computational perspectives are the result of software mechanisms that classify elements in a database for selective display. In WebGuide, for example, if I enter a note in my personal perspective then that note will be displayed whenever my perspective is displayed but not when someone else’s personal perspective is displayed.

WebGuide implements a system of computational (i.e., computer-supported, automated) perspectives designed to exploit the perspective metaphor in order to support characteristics of collaboration and collaborative learning. It is unique in a number of ways that distinguish it from other software systems that may use the term “perspectives”:

• Other systems refer to different representations of information as perspectives. They might have a graphical and a textual view of the same data. In WebGuide, different data is displayed in different perspectives – using the same representation, hierarchically structured titles of textual notes.
• In WebGuide, the perspectives mechanism is neither a simple tagging of data nor a database view, but is a dynamic computation that takes into account a web of inheritance among perspectives. Thus, Blake’s perspective includes not only information that he entered in his perspective, but also information inherited from the Class, Student and Landowner perspectives.
• Furthermore, the web of perspectives can be extended by users interactively and the inheritance of information is always computed based on the current configuration of this web.
• In addition, the information in a perspective has a user-maintained structure in which each note has one or more parent notes and may have children notes, creating a web of notes within each perspective. The order of children displayed under a parent note is user-defined and maintained so that WebGuide can be used to organize ideas within outline structures.

The idea of perspectives on the Web traces its lineage to ideas like “trail blazing” (Bush, 1945), “transclusion” (Nelson, 1981), and “virtual copies” (Mittal et al., 1986) – techniques for defining and sharing alternative views on large hypertext spaces. At the University of Colorado we have been exploring this approach to computational perspectives in desktop applications for the past decade (McCall et al., 1990; Stahl, 1993b). WebGuide is our first truly Web-based version. The core of WebGuide consists of a perspectives server named POW! (Perspectives On the Web), which communicates with Java, Perl or HTML interfaces.

The computational perspectives mechanism we have been exploring incorporates the following features for a community of users (Stahl, 1993a):

• Individual community members have access to what appears to be their own information source. This is called their personal perspective. It consists of notes from a shared central information repository that are tagged for display within that particular perspective (or in any perspective inherited by that perspective).
• Notes can be created, edited, rearranged, linked together or deleted by users within their own personal perspective without affecting the work of others.
• Another student, Annie, can integrate a note from Blake’s perspective into her own personal perspective by creating a link or virtual copy of the note. If Blake modifies the original note, then it changes in Annie’s perspective as well. However, if Annie modifies the note, a new note is actually created for her, so that Blake’s perspective is not changed. This arrangement generally makes sense because Annie wants to view (or inherit) Blake’s note, even if it evolves. However, Blake should not be affected by the actions of someone who copied one of his notes.
• Alternatively, Annie can physically copy the contents of a note from Blake’s perspective. In this case, the copies are not linked to each other in any way. Since Annie and Blake are viewing physically distinct notes now, either can make changes without affecting the other’s perspective.
• There is an inheritance web of perspectives; descendants inherit the contents of their ancestor perspectives. Changes (additions, edits, deletions) in the ancestor are seen in descendent perspectives, but not vice versa. New perspectives can be created by users. Perspectives can inherit from existing perspectives. Thus, a team comparison perspective can be created that inherits and displays the contents of the perspectives of the team members. A hierarchy of team, sub-team, personal and comparison perspectives can be built to match the needs of a particular community (Figure 2).

This model of computational perspectives has the important advantage of letting team members inherit the content of their team’s perspective and other information sources without having to generate it from scratch. They can then experiment with this content on their own without worrying about affecting what others see. This is advantageous as long as one only wants to use someone else’s information to develop one’s own figurative perspective. Such “perspective-making” is important in thinking about and judging issues from particular perspectives.

However, if one wants to influence the content of other team members’ perspectives through “perspective-taking” (Boland & Tenkasi, 1995), then this approach is limited because one cannot change someone else’s content directly. Moreover, for supporting collaborative work it is important that the perspectives maintain at least a partial overlap of their contents in order to reach successful mutual understanding and coordination. The underlying subjective opinions must be intertwined to establish intersubjective understanding (Tomasello et al., 1993). We are interested in exploring how to support the intertwining of perspectives with our computational perspectives mechanisms. We will return to this issue after describing the types of perspectives used in our applications.

4. Types of Perspectives

WebGuide provides several levels of perspectives (see Figure 2) within a web of perspective inheritance to help students compile their individual and joint research:

• The class perspective is created by the teacher to start each team off with an initial structure and some suggested topics. It typically establishes a framework for classroom activities and defines a space used to instantiate the goal of collecting the products of collaborative intellectual work.
• The team perspective contains notes that have been accepted by a team. This perspective can be pivotal; it gradually collects the products of the team effort.
• The student’s personal perspective is an individual’s work space. It inherits a view of everything in the student’s team’s perspective. Thus, it displays the owner’s own work within the context of notes proposed or negotiated by the team and class – as modified by the student. Students can each modify (add, edit, delete, rearrange, link) their virtual copies of team notes in their personal perspectives. They can also create completely new material there. This computational perspective provides a personal workspace in which a student can construct his or her own figurative perspective on shared knowledge. Other people can view the student’s personal perspective, but they cannot modify it.
• The comparison perspective combines all the personal perspectives of team members and the team perspective, so that anyone can compare all the work that is going on in the team. It inherits from the personal perspective and, indirectly, from the team and class perspectives. Students can go here to get ideas and copy notes into their own personal perspective or propose items for the team perspective.

Of course, there is not really a duplication of information in the community memory. The perspectives mechanism merely displays the information differently in the different perspectival views, in accordance with the relations of inheritance.

To design software for collaborative learning in schools means to design curriculum and classroom process as well (Stahl et al., 1995a; 1995b). Computer support has to be matched with appropriate content (typically stored in WebGuide or on the Web) and with constructivist practices for knowledge-building communities (Scardamalia & Bereiter, 1991). The design of the WebGuide interface and the perspectives mechanism must be adapted to individual application situations, with appropriate seeding of content, structuring of the perspectives web and establishing of access policies.

In Logan School, for instance, students each enter notes in their personal perspectives using information available to them: from the Web, books, encyclopedia, discussions, interviews of mentors or other sources. Students can review the notes in the class perspective, their team perspective and the personal perspectives of their teammates. All of these contents are collected in comparison perspectives, where they are labeled by their perspective of origin. Students extract from the research those items that are of interest to them. Then they organize and develop the data they have collected by categorizing, summarizing, labeling and annotating. The stages of investigating, collecting and editing can be iterated as many times as desired. Team members then negotiate which notes should be promoted to the team perspective to represent their collaborative statement of their team’s perspective on acid mine drainage.

5. Issues for Perspectives

As an initial field testing of the WebGuide system, the Logan School trial is generating valuable experience in the practicalities of deploying such a sophisticated program to young students over the Web. The students are enthusiastic users of the system and offer (within WebGuide) many ideas for improvements to the interface and the functionality. Consequently, WebGuide is benefiting from rapid cycles of participatory design. The differing viewpoints, expectations and realities of the software developers, teachers and students provide a dynamic field of constraints and tensions within which the software, its goals and the understanding of the different participants co-evolve.

The first issues to hit home when we deployed WebGuide were the problems of response time and screen real estate. The student computers were slower, had smaller monitors, lacked good Internet connections and were further from the server than the computers of the developers. We were, of course, already familiar with these issues from other Web applications, but one never knows quite how things will work out and how they will be accepted until one tests them under classroom conditions.

A pre-release prototype of WebGuide used dynamic HTML pages. This meant that each time one expanded a different part of the outline of titles one had to wait for a new page to be sent across the Internet. It also greatly constrained the interface functionality. However, when we moved to a Java applet, we had to wait several minutes to download the applet code to each student computer. Furthermore, it entailed running all the perspectives computations on the slow student computer. In order to reduce the download time significantly, we first rewrote the interface using standard Java Swing classes that can be stored on the student machines. Then we split the applet into a client (the interface) and a server (the perspectives computations and database access). By downloading only the client part to the classroom, we not only reduced the download time further, but also ran the time-consuming computations on our faster server computers.

Such technical problems can be solved relatively easily, by optimizing algorithms or by adjusting tradeoffs based on local conditions. Issues of social practice are much more intransigent. There seem to be two major issues for software like WebGuide, that is, software for threaded discussions and collaborative knowledge construction:

1. Lack of convergence among the ideas developed in the supported discussions.
2. Avoidance of system use in favor of email, face-to-face conversation or inaction.

WebGuide introduces its computational perspectives mechanism as a structural feature to facilitate the articulation of convergent ideas and even incorporates email. In attempting to address the above problems, it raises a new set of issues:

3. Is the perspectives metaphor a natural one (or can it be made natural) so that people will use computational perspectives to construct their figurative perspectives?
4. Can the web of perspectives be represented in a convenient and understandable format?

In our trials of WebGuide we have tried to create learning situations that would encourage the use of the software, yet we have observed low levels of usage and under-utilization of the system’s full functionality. This raises the following additional issues:

5. How can learning situations be structured to take better advantage of the presumed advantages of the software?
6. How can the system’s various capabilities be distinguished, such as its support for threaded discussions and for perspective-making?

In order to answer questions of this magnitude it was necessary to gather more experience, to be more closely involved in the daily usage of the system and to develop a deeper theoretical understanding of collaborative learning and of computer mediation. Having defined these goals, I announced a seminar on the topic of “computer mediation of collaborative learning,” open to interested researchers from a number of disciplines – primarily education, cognitive psychology and computer science. The goal of the seminar was explicitly stated to be an experiment in the use of WebGuide to construct knowledge collaboratively, based on careful reading of selected texts. The texts traced the notion of computer mediation (Boland & Tenkasi, 1995; Caron, 1998; Hewitt et al., 1998; Scardamalia & Bereiter, 1996; Stahl, 1999) back to situated learning theory (Bruner, 1990; Cole, 1996; Lave, 1991; Lave, 1996; Lave & Wenger, 1991) – and from there back to the notion of mediated consciousness in Vygotsky (1930/1978) and its roots in Hegel (Habermas, 1971; Hegel, 1807/1967; Koyeve, 1947/1969) and Marx (1844/1967; Marx, 1845/1967; Marx, 1867/1976).

In Section 8 of this paper I will comment on our current understanding of the six issues listed above. But first it is necessary to describe the ways in which the seminar attempts to make use of WebGuide and the conceptualization of the theory of computer mediation that is arising in the seminar.

6. Practice II: Theoretical Perspectives

The seminar on computer mediation of collaborative learning is designed to use WebGuide in several ways:

• As the primary communication medium for internal collaboration. The seminar takes place largely on-line. Limited class time is used for people to get to know each other, to motivate the readings, to introduce themes that will be followed up on-line, and to discuss how to use WebGuide within the seminar.
• As an example collaboration support system to analyze. Highly theoretical readings on mediation and collaboration are made more concrete by discussing them in terms of what they mean in a system like WebGuide. The advantage of using a locally-developed prototype like WebGuide as our example is that we not only know how it works in detail, but we can modify its functionality or appearance to try out suggestions that arise in the seminar.
• As an electronic workspace for members to construct their individual and shared ideas. Ideas entered into WebGuide persist there, where they can be revisited and annotated at any time. Ideas that arise early in the seminar will still be available in full detail later so that they can be related to new readings and insights. The record of discussions over a semester or a year will document how perspectives developed and interacted.
• As a glossary and reference library. This application of WebGuide is seeded with a list of terms that are likely to prove important to the seminar and with the titles of seminar readings. Seminar members can develop their own definitions of these terms, modifying them based on successive readings in which the terms recur in different contexts and based on definitions offered by other members. Similarly, the different readings are discussed extensively within WebGuide. This includes people giving their summaries of important points and asking for help interpreting obscure passages. People can comment on each other’s entries and also revise their own. Of course, new terms and references can be added easily by anyone.
• As a brainstorming arena for papers. The application has already been seeded with themes that might make interesting research papers drawing on seminar readings and goals. WebGuide allows people to link notes from anywhere in the information environment to these themes and to organize notes under the themes. Thus, both individuals and groups can use this to compile, structure and refine ideas that may grow into publishable papers. Collaborative writing is a notoriously difficult process that generally ends up being dominated by one participant’s perspective or being divided up into loosely connected sections, each representing somewhat different perspectives. WebGuide may facilitate a more truly collaborative approach to organizing ideas on a coherent theme.
• As a bug report mechanism or feature request facility. Seminar participants can communicate problems they find in the software as well as propose ideas they have for new features. By having these reports and proposals shared within the WebGuide medium, they are communicated to other seminar participants, who can then be aware of the bugs (and their fixes) and can join the discussion of suggestions.

The seminar version of WebGuide incorporates a built-in permissions system that structures the social practices surrounding the use of the system. Seminar participant