This exploratory project examines how techniques for automatically generating dialog contributions and managing dialog can be incorporated into the narrative structures of outdoor role-playing computer games, such as quests and mysteries. The underlying hypothesis is that the effective use of natural language and dialog technologies in interactive games will eventually lead to more compelling and engaging games, appealing to a much wider segment of the population, and usable for a much wider range of educational, assistive and entertainment applications. As a vehicle for testing the team's research ideas, a prototype novel physical-activity based social role-playing mystery game, Spy Feet is implemented, aimed at encouraging physical activity in young women and girls.
Spy Feet runs on mobile devices equipped with GPS, and the game world is mapped onto an outdoor environment consisting of paths and landmarks that are used by plot points in the game. Players must walk between landmarks where dialogs with game characters take place or where clues are located, thus covering a significant amount of a natural terrain while playing the game. Spy Feet utilizes a prototype natural language generation engine, Spy-Gen, that dynamically generates dialogue game narrative sequences whose characterizations and interaction style is targeted at young women and girls. Spy Feet uses Spy-Gen to dynamically adapt aspects of game play to the user and her environment in order to explore how such techniques lead to different game play experiences and outcomes.
One of the most important outcomes of this project was a proof of concept of a new kind of game. Building on the popularity of computer role playing games, we invented a new kind of game, an outdoor augmented reality role playing exercise game, and built a prototype exercise game called SpyFeet that runs on any Android smartphone equipped with GPS and accelerometers for tracking users locations and type of physical activity. For a project that only lasted one year, SpyFeet seemed to capture an undue share of public interest and enthusiasm. The PI was interviewed and featured in a local public radio broadcast. Several students in science writing programs sought us out and chose to do a major project writing up and describing SpyFeet. There were several newspaper articles that discussed it at length along with other UCSC games. Researchers from other universities and from games companies who attended talks by the PIs blogged about or wrote monthly news summaries that discussed SpyFeet. This served to increase public interest in science, including among the population of middle school kids, a number of whom participated in the four user studies we carried out over the year, and whose interest in science is critical for the future economic success of the U.S. The project was driven by three research goals: (1) to explore how techniques for language generation and dialogue management in the area of "natural language processing" (NLP) can be incorporated into the narrative structures of role playing games; (2) to situate NLP-enriched gameplay in the player's real-world physical environment, by linking game events to real world locations; and (3) to investigate whether this combination could potentially motivate increased physical activity for a target population of middle-school girls. There has been very little previous work using NLP in any type of game or interactive story. Thus, these were ambitious goals for a small, short-term project -- but we are pleased to report initial successes in all three areas. These goals are important for a number of reasons. First, research on preventive health has shown that despite various interventions, physical activity declines precipitously in adolescents, especially in girls, causing obesity, diabetes and cardiovascular problems. Novel interventions for motivating teenagers to exercise would thus help address a national health problem. Our results show that a highly-customized role-playing game experience, focused on narrative and character relationships, has significant potential for encouraging increased physical activity. Second, modern role-playing games are requiring larger and larger amounts of writing, while still not enabling the vast range of world states that players may expect. Significant strides in natural language generation for games could create increased possibilities for players (with significant commercial potential) while simultaneously lowering the production burden on creators -- not only helping large commercial game makers, but also enabling work by educators, non-profits, and small studios, who might like to create educational games or games for social justice, but cannot afford the resources given current tools. Our results represent significant scientific progress toward this goal.
