CHI 99 Online Communities Workshop


Learning How To Build and Use

Shared Virtual Worlds for Education


J. Michael Moshell, Charles E. Hughes

School of Computer Science
University of Central Florida
Orlando, FL 32816
moshell@cs.ucf.edu, ceh@cs.ucf.edu

Research Summary

The authors have built a series of 2.5-dimensional and 3d shared graphical virtual worlds for educational use within a formal learning model called the Virtual Academy (VA). VA is based around a progression of roles from Guest to Cast Member to World Builder to Tool Builder, and has been tested with third graders through college undergraduates. This paper reports the principle successes and problems that experiments have revealed in the VA model.

Research Methodology

Students work with teachers to select an educational theme and develop a story line. They then implement the virtual world by constructing graphical models and programming behaviors (as few as possible). The emphasis is on puppeteer (role-playing) control of props and activities, rather than on automated behaviors. The 2.5-dimensional ExploreNet system was used to build and test worlds built around stories about African American history (Zora's World), ecology (DinoLand) and economics (AutoLand.) The 3 dimensional Parkview system was used to build Caracol Time Travel, an archaeological project about the Classic Maya.

Insights about Guests and Cast Members. The maturity level of Cast Members is crucial; younger learners do not manage others' experiences well. Sixth graders can generally succeed as Cast Members. With third graders, an in-person buddy ("shoulder surfer") with prior experience works better than a cyberspace Cast Member/mentor. This is partly due to the lack of typing skills of this age children. It may also be due to our failure to develop worlds that motivate younger students to want to play a nurturing role.

Insights about World Building. Teacher input is essential if you want the world actually used. A hard problem is to balance the needs of the story with the actual lesson to be taught. (One story team had the players cast artifacts into the fire to release the spirits; a big no-no in archaeology. After brainstorming, they tossed in a drawing instead.) Another hard problem is to get the right mix of storytelling, graphical, auditory and enactment talent on your world building team. As with a small church choir, you have to find useful roles for all comers.

Techies and dreamers, boys and girls. In our middle school DinoLand experiment we found that boys were more interested in programming behaviors, while girls were more interested in graphic design. At the college level there was less gender specificity but more job specialization. People with skills for "early" jobs such as initial story-line development are in danger of becoming uninvolved later, so we made them team leaders. This worked well by empowering the humanists, vis-a-vis the technologists.

Future Directions

Communication Substructure. Shared worlds require a responsive communication and coordination infrastructure so participants can interact in a common environment. Our earlier efforts have been based on variants of the dead reckoning protocols developed in the distributed interactive community, with a mixture of UDP and TCP/IP used for the actual message transportation. Our current efforts are intended to raise the level of abstraction by developing within the conceptual framework of tuple spaces, and the specific contexts of JavaSpace and TSpace. This approach raises challenging issues of scalability, while providing convenient and simple solutions to retrospective analyses of user interactions.

Authoring Behaviors and Worlds. Since our typical users are teachers and students between fourth and twelfth grades, we must have a way to support the easy creation of worlds with a minimum of effort in authoring new behaviors. To some degree our basic philosophy of puppeteering helps to reduce the need to program behaviors, but not to eliminate it. We are interested in others' solutions to this problem.

Peer Mentoring. In our experiments to date, we have either not used mentors, randomly assigned them, or had teachers make the assignments. In a different context (web-delivered college courses) we have been able to show the effectiveness of automated matching for peer mentoring. How effectively can we do this in the open context of web-based shared virtual worlds? What features can be provided to world builders that assist them in developing worlds in which such peer relationships make sense?

Other Important Issues for the Field

Autonomous and Semi-Autonomous Characters. Puppeteering depends on the availability of participants (called "cast members") who are "in on" the secrets of the world and who help to carry the story line forward. There are occasions where an autonomous character might be better able to play this role (e. g. when cast members are not available.) We would like to work with developers of agent technology to explore this area.

Selected Publications