This project will develop advanced methods for automatically generating scenarios that are intrinsically motivating, responsive to the user's behavior, and potentially beneficial in many spheres of the economy and culture. As numerous existing communication media are merging, computer-based narratives have become increasingly complex, significant, and influential. However, at present they are inflexible and ignore the characteristics and goals of the individual user. For many of our most pressing societal needs - from more effective education to addressing climatology challenges - a key component of any approach is motivation. Intrinsic motivation, which involves performing an activity because it is inherently interesting, is often associated with deep learning and creativity. This project employs an approach that combines types of reasoning drawn from computational creativity and logic programming to meet this challenge.
The research will demonstrate the first successful scenario generator, dynamically combining narrative and simulated action, and showing that generation can be guided by domain models, laying the foundation for projects to leverage generated scenarios for intrinsically motivating learning and other activities. In so doing, it will execute a novel evaluation plan that will produce a more refined understanding of the relationship between intrinsic motivation and learning, including the interrelated categories of engagement, agency, and valuation of outcomes by comparing player experience of integrated scenario systems with and without allowing the user's choices to influence the challenges and narrative. It is hoped that this research will demonstrate the utility of heterogeneous architectures for enabling previously-impossible experiences for users in a wide range of interactive experiences. A specific focus is creation of a computer-based educational experience focused on meteorological shifts, made possible by this project's technical advances and informed by its findings related to user motivation and learning.
While the research community has had some success in generating both narrative and simulated action, generating both together presents novel research questions. Further, to meet social needs, this generation must be capable of being guided by pedagogical and other explicitly-represented goals. This research seeks to achieve two fundamental advances. First, the project will demonstrate a successful scenario generator, showing that generation can be guided by domain models, which will enable future projects to leverage generated scenarios for intrinsically motivating learning and other activities. Second, user studies will support the first empirically grounded understanding of the effects of narrative and action responsiveness and variation on user experience, particularly engagement, motivation, and learning. By demonstrating successful scenario generation, this project will turn attention to the scenario as a potential fundamental unit for educational software, with the potential to reach underrepresented groups and produce significant economic benefit.
