Stanford University

News Service



CONTACT: Stanford University News Service (650) 723-2558

Virtual children's theater comes to life

STANFORD -- What child has not wished, at one time or another, that a favorite toy would come to life and play?

Thanks to the magic of the computer, this dream is finally becoming true, at least for characters on the video screen. Barbara Hayes-Roth, a senior research scientist in computer science at Stanford University, has employed some of the techniques of artificial intelligence to create a "virtual theater" in which children collaborate with animated characters to create and perform their own stories.

"There is significant science involved, but I also like the idea of creating a computer game that will interest girls," said Hayes-Roth. "Young girls are not part of the targeted market for such games. As a result, many develop an arms-length relationship with the technology. I think that the virtual theater has the potential to attract girls to computers."

Equally important, she said, the virtual theater can serve as an experience-based educational environment in which children can learn writing and computer programming skills, as well as artistic and social skills.

And directed improvisation by computer characters may have even greater potential, Hayes-Roth said. It could provide a new paradigm for human-computer interaction. At a minimum, it offers a method of communication in which people don't have to provide their computers explicitly with all the details required to perform a given task.

Ideally, computer characters could evolve into valuable assistants that anticipate many of their user's needs and find optimal ways to fulfill them: a capability that has been one of the goals of artificial intelligence research for more than 20 years.

Hayes-Roth, her colleagues Robert van Gent, Lee Brownston, Erik Sincoff, Ruth Huard and Brian Lent from the Knowledge Systems Laboratory, and Professor of English literature Larry Friedlander have developed a computer code that serves as the brains of video characters. The code controls how the characters respond to each other and to the instructions that they receive from the children. It allows the characters to perform what she calls "directed improvisation." Like actors in an improvisational theater workshop, they can perform a new work, following general directions, without detailed preparation and making use of the resources at hand.

The basis of the code is something called intelligent agent technology. Intelligent agents include programs that are smart enough to do routine tasks, such as periodically checking a person's electronic mail or locating specific information on a computer network. More sophisticated agents perform more complex tasks, such as monitoring medical patients.

"A lot of research has been done on agents. Their use for improvisation, however, requires greater variety, more flexibility and more autonomy than other applications to date," Hayes-Roth said.

Using agent technology as a base, she is developing virtual theater software that will allow children to take part in all the activities involved in creating and improvising a play: playcrafting, creating characters, casting, directing, acting, and designing sets, props, costumes and even music. Then, during a performance, the children can act as one or more of the characters, an omniscient director or a member of the audience.

To date, the dramatis personae for this virtual children's theater consists of just two characters: a big woggle and a small woggle. Woggles are egg-shaped characters from an animation system developed by Joe Bates at Carnegie Mellon University that incorporates RAL software from Production Systems Technologies of Pittsburgh. Hayes-Roth's group has given its version of woggles multiple gaits and the ability to speak lines, follow directions and improvise.

For example, a child can direct the big woggle to act curious and friendly and the small woggle to act playful. Working within these abstract directions, the characters will improvise their own behavior and interpret their partner's behavior as an episode begins.

In a typical episode, the big woggle begins alone in the environment. Acting curious, it begins to look around. When the small woggle enters, it observes the big character standing still, but does not act on the observation. Instead, acting playful, it begins to play alone for a while and then hide.

Meanwhile the big woggle observes the small woggle playing alone and concludes it is shy. Acting friendly, the big woggle decides to approach the small woggle, greet it and invite it to play. The two characters have conflicting plans. What happens depends on which acts first. If the big woggle invites its fellow character to play, then the small woggle discards its plan and accepts the invitation. But, if the small woggle hides before the big woggle issues its invitation, then the big woggle can respond by getting into the game. Or, if the small woggle notices the big woggle approaching, it drops its plan to hide and does something else.

With the help of her own children -- Aaron, 13, and Nora, 10 -- Hayes-Roth has endowed these characters with personality traits, physical behaviors and a repertoire of recorded utterances. This allows the characters to exhibit different moods and other life-like qualities.

The computer scientists plan to develop a new animation system that will permit children to create new bodies, sets and props by configuring components from a graphics library.

Children can control the characters in what Hayes-Roth calls the "animated-puppets mode." The children work side-by-side, directing the characters' moods and behavior by making choices from a graphical user interface that shows them how their puppets feel and what they are thinking about doing at each moment. In this fashion, the children can work together to create a shared story, just as they would with physical puppets. In this case, however, the video puppets also collaborate in the playcrafting by focusing the children's attention on behavioral options that make sense in the current context and by improvising within the constraints of the children's directions.

The computer scientists intend to implement a number of other modes in the near future. In the "animated-actors mode," the children act as playwriters, preparing a complete script that the characters then follow. In the "improv troupe mode," the children direct the behavior of selected actors while the others perform without direction. This allows the children to function as improvisational actors within a larger troupe.

In two other modes, the children act something like a movie director. In the "collaborative playcrafting mode," they use animated puppets to improvise stories. Then, when they see something they like, they can store the corresponding script in a "repertory" for future use. In the "collaborative scriptwriting mode," they first write a script and then, when the actors improvise something the children like, the exact set of actions can be saved and replayed in future performances.



This is an archived release.

This release is not available in any other form. Images mentioned in this release are not available online.
Stanford News Service has an extensive library of images, some of which may be available to you online. Direct your request by EMail to

© Stanford University. All Rights Reserved. Stanford, CA 94305. (650) 723-2300. Terms of Use | Copyright Complaints