To learn is to construct personal knowledge within a social context. As our social information space explodes global, its computerized media must support personalization and restructuring on demand to help us make sense of its decontextualized content in our own terms.
Initial attempts to use the Internet to disseminate informational artifacts already demonstrate the need to develop computational supports for people to locate, use, adapt and share items of interest to them. Results of searches and browses that people undertake should coalesce into flexibly structured personal information space. When people adapt items from the universal space to their personal tasks and styles, they should be able to share their custom versions with communities of commonality.
The central project of the L3D U&U group is WebNet, an effort to build a Web-based design environment. The idea is to allow LAN designers to communicate across the Internet by sharing designs, rationale, simulation artifacts, etc. The system will be optimized to support a rapidly changing technological context. Although the domain incorporates global standards, much of the shared knowledge will be contextualized by concrete designs.
Each community of designers (e.g., network managers at CU) will be encouraged to personalize their WebNet info space by tying in their own email repositories, importing CAD drawings of their buildings, capturing their current network topologies and defining their own terminologies in the end-user programming language. People can share their personal contributions by linking in Web pages with rationale of their reasons for their designs, their definitions of simulation behaviors in the language, and their own sample designs. Other communities can construct versions of WebNet by selecting subsets of agents, rationale and language phrases that are relevant to their needs.
The L3D group has developed simulation environments for use in the classroom and begun to distribute them across the Internet. In particular, WebQuest is an environment in which kids make quest games that incorporate researching facts on the Web. Students play their classmate's games and critique their designs. Eventually, kids will share their creations with other schools across the Web. They will share characters and behavior-defining subroutines in the end-user programming language as well as entire quests. As the mass of sharable quests, agents and behaviors accumulates, computer supports for locating, selecting and using the most desirable ones will become critical.
Quests will need to be associated with curriculum and related classroom resources to be pedagogically useful to teachers. Some form of indexing will be necessary to help teachers or students find the quest that most closely matches their needs and capabilities. For instance, some quests might take advantage of special peripheral equipment or build on special background knowledge; one does not want to download multiple quests just to find out they are not useful. Most importantly, as kids modify quests that they find on the Web and then upload their versions, it will be necessary to differentiate the versions and allow potential users to select the version best for them. I prototyped a Teacher's Curriculum Assistant to explore mechanisms for supporting the locating, using, adapting and sharing of classroom resources in digital libraries.
We have been hampered in our attempts to build design rationale systems by the difficulty of getting people to make their reasoning explicit and to enter it in a computer. GIMMe captures email to avoid this problem, but ends up with an unsystematic body of documents with uneven coverage of the issues. The institutionalization of documentation standards like ISO 9000 -- especially when combined with a decentralized decision-making structure like Total Quality Management -- requires employees to formulate comprehensive statements of policies and procedures. It is still hard to get people and work groups to articulate their working knowledge and enter it into a corporate memory. However, computer supports can elicit explicit statements of tacit knowledge, can manage the group editing process and can deliver the shared information in personally relevant formats.
The CIE demo system I developed recently explores some mechanisms for supporting the construction and use of evolving organizational memories. The system is seeded with knowledge about the ISO 9000 requirements, including templates for typical documents. Two central processes for the asynchronous development of policy documents by committee are supported: commenting on shared drafts and editing new versions. There are mechanisms for viewing comments and versions created at different levels of the organization and for merging multiple versions or promoting selected personal versions to group consensual documents. This is a way of reseeding the memory as it grows. Computational indicators warn when versions of a particular document are proliferating and consolidation is warranted.
Corporations are complexly structured organizations and documents of their policies and procedures are sensitive assets. Individual employees are not interested in all of an organization's documents, but just in those that affect their own work -- and management does not want everyone editing every document. In CIE, security concerns merge with personalization features. A perspectives mechanism governs access to the organizational memory for each employee, manager and auditor based upon their position in the corporate work-flow. Employees have read, comment and edit permissions relative to their quality work circle, department, branch, peers or roles. Perspectives allow people and groups at different levels to experiment with and share tentative versions without interfering with the official drafts.
The World Wide Web opens countless possibilities for global sharing of information to create and sustain knowledge-based communities. However, to be useful knowledge bases on the Web require sophisticated computational supports. As on-line memories grow and evolve, they need to be pruned and restructured to meet the needs of individuals with specific tasks and concerns. This cannot be done just at the global level -- although reseeding at this level is necessary as well (like Netscape frequently releasing new versions of its browser with new functionality). Everyone has their own interests and must be able to maintain their own personal information space as views onto rapidly evolving universal spaces.
The Web model of hypertext -- with simple links connecting static pages -- is too undifferentiated to support this. In my systems I try to apply a model of computational hypertext to the Web. My Hermes substrate is based on:
|fine-granularity of text so that documents can be constructed dynamically,|
|typing of links and nodes so that links and nodes can be displayed based on selective queries,|
|hierarchies of perspectives so that displays are personalized to include appropriate domains of content,|
|an end-user query language for knowledge-based computational navigation and dynamic display,|
|end-user extensibility of link/node types, the perspective hierarchy and the end-user language to support evolution.|
These mechanisms allow for the generation of pages whose contents and links are dynamically personalized. Systems with such mechanisms let people construct their own knowledge by structuring personal information spaces that provide useful and usable access to virtually limitless, constantly evolving organizational and global repositories.
Go to top of this page
Return to Gerry Stahl's Home Page
Send email to Gerry.Stahl@drexel.edu
This page last modified on January 05, 2004