Grants and Contracts Details
Description
Virtual environments are increasingly used for entertainment, education, training, and therapy. Many are
interactive narratives that invite the player to collaborate with a cast of virtual characters and tell a story
together with the system. Players want agency, the freedom to take meaningful action and see the results of
their choices. Designers want to ensure the story meets certain aesthetic or pedagogic constraints. Providing
both player agency and author structure has traditionally required writing branching stories by hand, a
brittle tactic that gets exponentially more expensive with each choice offered to the player. To achieve the
richness and scope of non-digital interactive narratives, like the roleplaying exercises used to train nurses,
firefighters, and police, digital environments need to generate narratives at run time. Strong autonomy (or
simulation-based) approaches provide realistic virtual characters but struggle to control the story’s trajectory.
Systems with constraints on the content and quality of the narrative have long sought to leverage strong story
approaches which centralize reasoning about the characters, player, and author to anticipate possible stories
in advance and choose one which balances player freedom and author structure.
This project will improve the intelligence and speed of strong story narrative planning algorithms so
they can generate realistic interactive stories at run time in spaces too large for an author to anticipate
by hand at design time. Once they scale past this point, designing these environments presents a unique
authoring challenge, so we will also provide intelligent tools to help non-expert designers understand,
constrain, and test the interactive spaces they are authoring. Finally, we will evaluate principles for effective
automatic experience management to ensure the quality of an interactive narrative remains robust in the face
of unpredictable player behavior.
Intellectual Merit
Narrative is a fundamental cognitive sense-making tool and one of our most natural methods for expression,
communication, and learning. Improving the capacity of intelligent systems to reason about and generate
narratives automatically will be impactful for human/computer interaction. The goal of this project is to
enable a fundamentally new generation of interactive narrative environments that features the player freedom
of an open world, the narrative structure of a hand-written story, and the believable virtual character behavior
of a strong autonomy simulation.
Plan-based computational models of narrative are especially attractive for this purpose because they offer
a formal, generative model of goal-directed action. They can procedurally and intelligently combine small,
highly reusable action templates into large narratives according to models of realistic character behavior
and narrative structure. The key feature of narrative planning being studied in all phases of this project is
anticipation--the ability to reason intelligently about thousands of hypothetical possible futures. Efficient
anticipation is a unique computational challenge. It is valuable to the player at run time because it enables
high-agency interactions with believable characters, but it is also valuable to authors at design time because
it can help them understand the interactive spaces they are defining.
Broader Impacts
The research proposed here is domain-independent and can be adapted to a wide variety of virtual
environments, but in order to provide a proof-of-concept and a place to evaluate our methods, we will create
an intelligent virtual reality police training simulation in partnership with our local police department. It will
impact the local, state, and national efforts to reform use-of-force training. More broadly, the technology
that enables this simulation will demonstrate, both to researchers and the public, that decades of research
on strong story systems is ready to transform virtual environments that teach, train, treat, and entertain by
allowing authors to create high-agency, structured interactive narratives.
This project will enhance the infrastructure of scientific education by training two doctoral graduate
research assistants, one female and one a person of color, in a highly interdisciplinary field that attracts
1
Status | Finished |
---|---|
Effective start/end date | 10/1/19 → 9/30/24 |
Funding
- National Science Foundation: $493,256.00
Fingerprint
Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.