We now sketch the learning environment, then cast it into a larger view, one in which we see the substrate of a Glass Bead Game (Hesse, 1990) evolving.

(ThinkAlong, 1993) - TSC. The program implements a framework for representation of process-oriented knowledge and includes functions that allow for the manipulation of that knowledge in a variety of ways. A primary function is that of the automated construction of models when given a variety of objects and rules of behavior. For instance, at its simplest level, a child can type in simple sentences and view a graphical representation of a taxonomy of objects described in those sentences. Sentences that describe behaviors result in rules. Configure a group of objects and ask the program to build a model. The result will be a directed graph representation, called an envisionment, of the behaviors the group of objects can take.

A sketch of a graph constructed by sentences typed by the author's (then) 7-year old daughter

is Figure 1. Sentences that were entered were of the form

A fourteen-year-old intern spent a summer "teaching" the program about the habitat of phytoplankton and zooplankton. She added sentences that describe processes, events that would happen in that environment. The program first built a diagram of the taxonomic aspects of her environment, then it went on to build an envisionment of those processes occurring, events such as zooplankton eating phytoplankton, and so forth.

A rather more complex model has been built to model object behaviors at the molecular level (Trelease, et al, 1999). We provide a view of one of the envisionments produced in that study (Figure 2). While the diagram appears somewhat cryptic, the initial conditions for the model start at the left. Envisionment A has been annotated to reflect the semantics of each node in the graph.

We describe this model as one of a stage setting, as an application of a theatrical metaphor. Initial conditions consist of Actors on a stage, each set in some relationship with the others. Given such a stage setting, rules related to each actor on stage are then exercised. Those rules that can fire, that is, that can cause some change to the stage setting, do so. Each rule that fires creates a new node; multiple nodes connected to a single prior node indicate that multiple rules fired.

Application of a stage metaphor satisfies an intention of representing images of game play in a MUVE. Direct links between behaviors of a game engine are thus readily translated to the graphic images presented at the web site. Graph traversal over an entire envisionment can be rendered in animated scenes in the virtual environment.

Building models is itself a learning experience. The user has the task of finding a suitable way to transfer personal constructs into a form usable by the program. Transformation alone involves research and thinking. The construction of a model can result in an expectation failure, usually, a model that does not behave as expected. An expectation failure, followed by knowledge editing summarizes the primary user experience capability required of The Scholar's Companion.

A secondary capability of the program is that of performance of a sequence of mutations of knowledge supplied by the user. We tend to think of users as "teachers" since they are supplying the program with representations of their personal constructs. We view the secondary capability as one of advanced "what if" study of that knowledge. To perform such studies, the program applies an evolutionary algorithm to the mutation of objects in the closed universe supplied by the user. We discuss that algorithm next.

Coupling Bead Game players together at a graphically rich portal will provide the stimulus for players to exercise their engines off line. Learning takes place at a variety of levels, each level behaving as, perhaps, a fractal image of a larger level.

Design goals for The Scholar's Companion center on the notion of providing a user with the opportunity to explore some domain of interest from many different views. The idea is that one can hold some concept up in the air, rotate it around, and study it from many different sides. This process starts with the "teaching" process, where the user enters new knowledge into the program. Doing so with a qualitative vocabulary, as opposed to a purely mathematical vocabulary, opens the door for users at nearly all levels of expressive capability. Indeed, the program can be programmed to respond to Icons dragged onto a workspace by very young learners. An overview visualization of the design of The Scholar's Companion is Figure 3.

A variety of views, from the simple textual - sentences entered by users or rendered by the program in response to questions - to the purely mathematical view as expressed in the rendering of solutions to equations entered by the user or derived by some program. The question-answering capability of the program has been used in the conduct of classroom exercises in two high school biology courses.

In between, we find Simulation. This is the view that expands an envisionment that describes the response of process rules to a stage set by the user. What-if games can be played with The Scholar's Companion in this manner. At the same time, serious research can be conducted as well. A Ph.D. was gained by performing research using the program to help discover new process control rules for polymer curing (Abrams, 1995).

How might this environment be used? We envision an excursion into an educational game environment to be composed of several kinds of events. Consider the Grand Bead Master (Magister Ludi in the Hesse vernacular) - a web site itself, staffed by game designers. Suppose that the web site presents a first Bead, a game design complete with rules of competition. Now, consider a user, an individual who has a copy of TSC and is familiar with it. The user would be expected to begin the construction of a knowledge base, a small universe of discourse suitable for the development of a simulation that allows for off line exploration and discovery of winning strategies. This will be done off line, after the user has visited the web site. Such a user will have access to knowledge accumulated from the games of the past, archives at the portal. These links to prior art may serve as the basis for a new effort, or they may provide ideas from which the individual takes a new point of view.

The individual may run off line competitions with others, or may build a simulator that runs as an internal opponent. Analysis of the results of play may support the discovery of new heuristics that eventually lead to a winning strategy. Eventually, the individual enters the on line competition, coupling the locally trained TSC to an on line game hub, the MUVE itself. There will be others coupled to the hub and the games begin.

Conclusions

We have been developing the thread woven here with a goal to make available a program patterned after The Scholar's Companion that provides a powerful tool for learning by discovery. We believe that the development of a graphically rich MUVE environment, one that will support and animate game exercises, is an appropriate vehicle by which we might encourage learning experiences for users of all ages.

In the long run, the system we envision places content rich multiple user virtual environments firmly in the center of a web activity that encourages and contributes to the growth of knowledge in a variety of ways. It is this growth of knowledge that lays foundations for human development.