The World Wide Web is currently used primarily as a medium for broadcast, a one-to-many monologue providing limited opportunities for interactive collaboration. However, the web possesses the potential to serve as a medium for design, a collaborative activity involving the construction of shared artifacts and shared understanding among multiple stakeholders. To realize this potential, we need to reconceptualize the web as a medium and to redefine the current roles of web users (those who access web content) and web masters (those who produce web content). Communities of designers must be enabled to not only access information on the web, but also to use the web to create, share and evolve information spaces and artifacts without having to rely directly on web masters. In this paper, we first illustrate the difficulties of using the web as a medium for collaborative interaction. We then propose a new framework for reconceptualizing the web and the roles of web masters and web users, and describe two prototype systems, Dynasites and the Agentsheets Behavior Exchange, that fit within our framework.
Design, Extensible web-based structures, Dynasites, High tech scribes, communities of practice, Agentsheets Behavior Exchange
The rapid growth and acceptance of the World Wide Web has catapulted this technology into the status of our newest mass medium. Like all new technology, the web has been primarily understood in terms of familiar media [1], namely television and magazines. These media are best suited for broadcast--a one-to-many monologue that affords little or no opportunity for interactive collaboration. This paper argues that viewing the web in terms of existing broadcast-oriented media may obscure its potential as a design medium.
We understand design as a process of making sense of an initially vague problem. Design is a common activity in myriad professions [2,3]; authors design articles or novels, doctors design treatments and procedures, architects design buildings and computer programmers design software. As professions become more specialized, design activities are becoming more interdisciplinary and collaborative. Design is increasingly an argumentative process involving multiple stakeholders who must develop a shared understanding despite having different backgrounds and perspectives on a given problem [4]. For example, multimedia software development routinely involves graphic designers, programmers, analysts, users and technical writers, to name a few possible participants. Seen as a design medium that provides distributed access to a shared information space, the web holds great promise to support communities of designers to develop shared understanding through the construction, discussion, and evolution of artifacts [5].
For the web to fulfill its potential as a medium for design, however, web users need to gain considerably more independence from web masters. Just as the pen was taken out of the hands of scribes in the middle ages [6], the role of web masters (the high tech scribes of the web) should be redefined so that the owners of information have more control over it. Although millions of users access information on the web regularly, far fewer actually posses the technology and the skills necessary to create and share new information [7]. To create new content on the web, most users must employ the services of their local web master--a specially trained individual who possesses the requisite skills and tools to place new information on the web. Gaining control over their information requires that web users not only gain the ability to author new content more easily, but that once this content is created it can be modified and added to by a larger community of users. To achieve this goal, we have developed a conceptual framework and prototypes that reconceptualize the web as a medium for design rather than just for broadcast.
In this paper we describe the current limitations of the web as a design medium, illustrating our arguments from experiences gained teaching a course in which students attempted to use the web to support collaborative interaction. We then describe a framework for conceptualizing the web that redefines the roles of web users and web masters. We present two prototype applications that support design activities on the web, Dynasites and the Agentsheets Behavior Exchange, and describe them in terms of our new framework.
Broadcast media, like television and magazines, draw strict boundaries between producers of information and consumers, thus limiting the opportunity for collaborative interaction. Though these strict boundaries are appropriate for some media--it may not be desirable to have television shows altered or extended by viewers--we must recognize that computational media, such as the web, afford new opportunities for information consumers to become information producers as well. Transcending the traditional boundaries between producers and consumers of information is essential for a design medium, because design entails collaborative construction of shared understanding and shared artifacts. The web as a design medium must provide a shared and extendible repository of design artifacts that can be evolved by a community of designers.
Communities of designers already employ computational technology to enhance communication. One of the most frequent uses of the Internet is for sending and receiving email [8], and many commercially available email packages now permit the sending of attachments, including word processing files, spreadsheets, graphics, etc., so that people can share artifacts more easily. Computer supported cooperative work (CSCW) tools, such as Lotus Notes®, allow users to create, share and comment on things like schedules and documents, and are also gaining in popularity. Although the web could be used to support these and other types of activities, currently its primary use is for information broadcast.
Commercial tools currently available for creating web pages and sites reinforce the perception of the web as a broadcast medium. The Hyper Text Mark-up Language (HTML), and applications used to generate it (Adobe Systems PageMill®, Optima's PageSpinner®, etc.) are all geared toward the presentation of information but provide minimal support for collaboration and feedback. There is an HTML tag that, when supported by the web browser, allows email to be sent to a particular address, but that is typically the extent of feedback that most tools readily provide. More recent commercial applications provide additional features for supporting richer feedback mechanisms. For instance, Microsoft FrontPage® employs its WebBot™ technology to simplify the creation of forms and threaded discussion groups that can be embedded within web pages [9].
Web publishing tools are important in that they simplify the creation of web content, which is the first step in placing control of information into the hands of web users. However, to support design requires not only simplifying the creation of web content, but also allowing that content to be extended and evolved by others once it is created. The architecture of the web provides for this kind of collaboration, but it requires the definition of special computer programs using protocols like CGI (Common Gateway Interface) or languages such as Java.
Three models of web-based interaction (Figure 1) and the tools that are used to support them are outlined below:
In M1, web users enlist web masters to help create HTML documents that broadcast information over the web. In M2, web users and web masters create more sophisticated HTML documents that allow for indirect feedback (via email or on-line forms), but the feedback does not alter the content of the document. In M3 web users have the ability to directly alter and extend the content of web-based documents.
During the Spring of 1996, we designed and conducted an experimental independent study course at the University of Colorado with five non-computer science undergraduates interested in studying the web as a tool for learning. In teaching this course, we hoped to gain insight into whether the web was adequate as an educational medium, and if not, where it was lacking. Early in the semester, the students were asked to describe what they felt was good about the web as a medium. The two qualities students cited as the most important were the web's global audience and the fact that many different perspectives were potentially available for creating a broad and shared understanding of any given topic.
In addition to reading literature and critiquing existing educational web sites, each student was responsible for creating a site that provided learning exercises in a topic of interest to that student. The students were able to eventually learn web publishing tools sufficiently enough to produce their own HTML documents. A web master was available to assist in moving the documents to a web server, thus making them accessible to the public. At this point, the students were able to use the web for M1-type interaction.
The students came from a variety of backgrounds, and chose an eclectic group of topics for their web sites. A physicist created pages about black holes, an economist created a site describing what money was, two fine art students created an on-line gallery of computer art, and an environmental design student created a site about arctic alpine ecology.
The students were engaged in authentic, self-directed learning activities, and they wanted to deepen their knowledge of the web as a learning tool by engaging a distributed community of learners [10-12]. They wanted to achieve this by creating web sites that other interested learners could directly contribute to and evolve. Our students wanted to build web sites that would allow other users to contribute to the design. They wanted M3-type interaction, but they were thwarted from this goal because the tools they were provided were not powerful enough. It became clear that the only way to create this kind of interactivity was to write CGI scripts on the web server, and even the web master was unable to do this.
In the end, the students settled on providing several links in
their pages where other web users could email them feedback. The
feedback was then (eventually) added to their web pages. Although
the students had desired M3 interaction, they had to compromise
with an M2 model. In order to achieve M3 interaction, a new framework
for reconsidering the web and the roles of web masters and web
users needs to be developed, and new tools that fit within that
framework need to be created.
Our students' experience illustrates the need to redefine the roles of web users and web masters in order to promote collaborative design. Designers (a special kind of web user) need the ability to continually and directly evolve and refine their information space, without relying on a high tech scribe (the web master). At the same time, it is clear that designers cannot be expected to have the ability to create their own environments. For this task, the web master's specialized knowledge and access to tools is necessary.
A framework specifying the redefinition of these roles is grounded by an understanding of which tasks are best left to designers, and which tasks are best delegated to web masters. To delegate is to entrust someone with the responsibility for performing some task--to give ownership of a task to someone else. Delegation of design tasks is not desirable, because design has at least as much to do with defining a problem as with solving a problem [3]. Design proceeds through incremental steps of creating artifacts and reflecting upon the results--a "reflective conversation" between the designer and the materials of the design situation. In the case of collaborative design, design becomes a conversation in the more literal sense because designers use the design medium not only to facilitate a conversation with the design situation, but also with other designers. In collaborative design, the artifacts and the conversation about the artifacts evolve as the design process unfolds. The evolution of artifacts and conversation must be performed directly by designers, rather than be delegated to web masters.
We are not calling for an end to the existence of web masters. Reducing the reliance on these high tech scribes does not mean denying the value of specialization and division of labor. Professional expertise has its place--if it is used properly. There is evidence from other design disciplines (e.g., architecture [13]) that design methods neglecting expertise possessed by well-informed and skilled specialists will lead to inferior products. What we are calling for is an awareness of the necessity for designers to be in charge of their own media in collaborative design. However, designers will never be completely independent of web masters because designers are typically not able to create their own information spaces from scratch. Designers must delegate this task to web masters. These specialists should provide evolvable, information spaces, but not specific content.
The Seeding, Evolutionary Growth, Reseeding (SER) framework [14] (Figure 3) describes a distribution of roles between web masters and designers in collaborative design on the web. The framework is motivated by the hypothesis that design communities must be empowered to directly manage their information spaces. At the same time, the SER model acknowledges that sustained usefulness and usability of information spaces requires a continual relation between designers and web masters. Web masters are necessary to perform tasks that require development knowledge, such as building tools for the designers and structuring information spaces.
Information spaces supporting design are regarded as seeds because they are designed to evolve. In the seeding phase, designers and web masters work together to create an information space seeded with domain knowledge. Collaboration between designers and web masters is essential in the seeding phase because designers understand their needs, while web masters understand the technology available for constructing tools.
In the evolutionary growth phase, designers work directly with the information space and with each other to create artifacts and to do design, adding information to the information space in the process.
In the reseeding phase, web masters help designers to reorganize and reformulate domain knowledge so it can be reused to support future design tasks. As in the seeding phase, web masters are required to assist designers in restructuring their environment and to implement new tools to manage it.
We have developed two prototypes that illustrate web-based information spaces that are seeded by web masters and designers, but that are evolved by designers.
Dynasites [15] is a substrate for developing evolvable web-based information spaces. Dynasites is used by web masters, in collaboration with designers, to create information space seeds, which are then evolved directly by the design community.
Most sites on the web today are implemented as HTML files which are displayed in a web browser, such as Netscape Navigator®. This model is well-suited for broadcasting information over the web. However, evolvable web-based information spaces are difficult to implement within this model because they require that the HTML source files are modified by users. Not only does this require HTML knowledge and access to the source files, but it also poses security risks from trouble makers as well as from users who might inadvertently alter important information.
Dynasites implements a different model of web-based information spaces. Rather than storing information in files, Dynasites stores information content and hypertext links within a database. Pieces from the database are then put together dynamically using the Frontier® scripting environment to create HTML pages. This model provides new possibilities for designers to evolve the information they see in their browser. Designers can add to the information space using forms that require no HTML knowledge. Information spaces created using Dynasites are similar to on-line, threaded discussions in that they allow users to communicate directly, and by doing so, evolve a shared information space; however, Dynasites information spaces differ in that they support the collaborative creation and evolution of artifacts through which communication can take place, rather than supporting communication as an end in itself.
Dynasites has been used to create an evolvable glossary of terms to support a group of authors (an instance of a design community) who wish to collaborate on a book (Figure 4). The authors all shared an interest in the topic of the book (organizational memories and organizational learning), but they had different perspectives and used different vocabularies. They wanted to understand how their respective vocabularies and perspectives overlap and where they differ, so that the book would be coherent and yet present a clear view of differing positions on controversial issues. The authors decided to employ an extensible glossary of terms as a means for developing this shared understanding.
The authors were not web specialists, so they did not have the
ability to implement an extensible glossary. However, they were
knowledgeable about their domain, and had specific ideas of what
terms should be in the glossary. The authors collaborated with
web masters to build an extensible glossary seed using the Dynasites
substrate. The web masters built the data structures and mechanisms
for displaying, annotating and adding new terms, and the authors
supplied the initial content. The glossary seed contained many
terms-some of which were not defined initially. This incompleteness
was acceptable because the authors expect the glossary to evolve
as it is used.
Once the glossary was seeded, the authors could evolve it directly, without relying on the web masters. In this sense the authors are in control of their own information space. By annotating glossary terms, they could hold on-line "discussions" about the meanings of terms. New terms can be added to the seed by the authors, allowing them to extend the scope of the seed. Terms can be redefined to reflect an evolving understanding of the book's subject.
On the index page of the glossary (not shown), terms that have been annotated or redefined are displayed with a number of trailing dots that indicate the number of annotations and new definitions that have been added. This method of indicating the modification a term has experienced is intended to help the authors to notice when reseeding activities are necessary.
While Dynasites is a substrate for creating textual, evolvable web-based information spaces, the Agentsheets Behavior Exchange [16,17] extends these evolvable spaces to include not only textual artifacts but computational artifacts including simulations and agents. The Agentsheets Behavior Exchange is a web site that supports Agentsheets designers.
Agentsheets [18] is a software environment for creating SimCity®-like simulations, games and domain-oriented visual programming languages. Agentsheets applications include a collection of autonomous computational processes, called agents, that are comprised of a look, their on screen representation, and a programmed behavior. Agents in Agentsheets are programmed in a programming language suitable for novice programmers called Visual AgenTalk [16].
In our previous work on the Remote Explorium [19] we supported designers of Agentsheets applications by allowing the exchange of entire Agentsheets applications via the web. Our experience with the Remote Explorium indicated that there was a need to exchange application components of a finer grain, including individual agents. In response, the Behavior Exchange serves as a web-based forum for Agentsheets designers, ranging from middle school children to professionals, to discuss and exchange individual agents and agent-based simulations.
Through the use of the Behavior Exchange the design of computational artifacts becomes a social activity:
The ability to exchange computational entities was also explored in the Buttons system [21]. The agents in Agentsheets, however, can be created by novice programmers and can be shared and archived publicly via the web.
The Behavior Exchange features a growing collection of agent classes. Under the heading of Special Effect Agents one can find a simple Flash agent that can be used to create explosions which are handy for a number of game-like applications. Agentsheets designers browsing the Behavior Exchange get a first impression of the behavior through an animated representation of the agent in the web page. The Fire agent in Figure 5(B) is flickering and the Flash agent is exploding. Each agent is completely embedded within the HTML as a binary stream. That is, when designers find interesting behaviors they will simply drag the agents out of the web page into the Agentsheets application. This process transfers the agent including all of its resources (icons providing that agent's look, and a Visual AgenTalk program providing that agent's behavior) into the designer's Agentsheets application running locally.
Agents are transferred as glass boxes allowing other designers
to change their original look and behavior. As illustrated in
the Visual AgenTalk program in Figure 5 (A), the designer of the
Flash agent had to create a large number of icons and define a
timing sequence that caused the agent to switch between icons
and play an explosion sound at the right time. In Figure 5 (B)
a designer has dragged the Flash agent into a space game in which
the flash is used to denote an explosion resulting from laser
beams hitting a target. Currently the Behavior Exchange features
agents in categories such as Transportation: Roads 
,
traffic lights 
and cars 
;
Special Effects (see below), and Musical Instruments, including
pianos 
, and trumpets 
.
Though the Behavior Exchange was seeded by web masters in conjunction with Agentsheets designers, it is continually evolved by Agentsheets designers without the intervention of web masters. At some point, the web masters working with the Agentsheets designers will have to reseed the Behavior Exchange, when it comes time to establish new agent categories or provide new functionality.
We have identified three models of web-based communication and characterized these categories in terms of tools that support them, the role distribution between web users and web masters, and the kinds of artifacts each model produces (Table 1). The Broadcast model (M1) is the result of seeing the web in terms of familiar media, such as television and magazine publishing. This model strictly separates information producers from consumers, and is reinforced by tools that support only the production of web content.
The Broadcast with feedback model (M2) is an emerging trend on the web, brought on by users' desire to utilize the web for interaction and also by new tools that enable feedback from other users, mainly in the form of email. In this model, feedback must be collected by either a web master or a user, and then manually added to the information space. Here the information space evolves, but the evolution is mediated by a person.
| M1 : Broadcast of information | PageMill, PageSpinner, etc. | Web masters produce content, users access information | Static HTML pages |
| M2: Broadcast with feedback | FrontPage, etc. | Web masters produce content. Viewers provide feedback to web pages via email, which is collected by either web master or end-user. | Web sites evolve through feedback, which is mediated by either web master or user, |
| M3: Collaborative Design | Dynasites and the Agentsheets Behavior Exchange. | Web master and designers help to create seeds, which are then evolved by design communities. | Evolvable text-based and computational artifacts |
The Collaborative Design model (M3) is proposed in this paper as a new way of conceptualizing the web, motivated by the need for communities of designers to directly evolve their information spaces (without mediation). This model allows informed participation [22]; it supports both being informed and participating in the world at the same time. In the context of M3, designers work with web masters to create an information space seed. Once the seed is created, the designers evolve it directly as part of the design process.
The challenge in making the collaborative design model a reality on the web is to allow designers to create more interactive web pages that can serve as forums for a new kind of communication including the joint construction of artifacts. This step requires that we reconceptualize the web as a medium with respect to the type of interactions it can afford. The evolvable glossary (created using Dynasites) illustrates how designers can interact directly with a web-based artifact, and in the process, communicate with each other to build a shared understanding.
Equally important to reconceptualizing the web with respect to the type of interaction it should afford, are the possibilities of using the web as a shared place to construct new kinds of artifacts. The Behavior Exchange is a forum for social construction of behaviors (distributed constructionism [23]), and points a way toward finding new uses for the web as a design medium.
We wish to thank the members of the Center for LifeLong Learning and Design for their thoughtful discussions and compelling arguments. We'd also like to thank Corrina Perrone for her timely advice. This work has been supported by the Advanced Research Projects Agency under Cooperative Agreement Number CDA-940860 and award number N66001-94-C-6038, and the National Science Foundation under grant number RED925-3425.
[1] McLuhan, M. Understanding Media: The Extensions of Man, The MIT Press, Cambridge, Massachusetts, 1964.
[2] Simon, H. A. The Sciences of the Artificial. second ed., MIT Press, Cambridge, Massachusetts, 1981.
[3] Schoen, D. A. The Reflective Practitioner: How Professionals Think in Action, Basic Books, New York, 1983.
[4] Bannon, L. J. "The Politics of Design: Representing Work." Communications of the ACM, 1995, 38:9, pp. 66-68.
[5] Scardamalia, M. and Bereiter, C. "Higher Levels of Agency for Children in Knowledge Building: A Challenge to the Design of New Media." Journal of the Learning Sciences, 1991, 1:1, pp. 37-68.
[6] Fischer, G. "Putting the Owners of Problems in Charge with Domain-Oriented Design EnvironmentsUser-Centered Requrements for Software Engineering Environments." In User-Centred Requirements for Software Engineering Environments. D. Gilmore, R. Winder and F. Detienne (eds.), Springer Verlag, Heidelberg, 1994, pp. 297-306.
[7] Hoffman, D., Kalsbeek, W. and Novak, T. "Internet and Web Use in the U.S." Communications of the ACM, 1996, 39:12, pp. 36-46.
[8] Kr