This SBIR Phase I project proposes to create a pro-social simulation to reduce aggressive behaviors in children, using a novel platform for simulated role-playing from crowdsourced data. Through a browser or mobile app, participants enter a 3D virtual school environment where they play the roles of a bullying victim and bystander, and interact with characters controlled by artificial intelligence (AI). By walking in another's shoes, users take the perspective of others and empathize with them -- capacities proven to foster improved relationships. The effectiveness of the simulation lies in allowing users to interact in the virtual world as they would in their real-world school. Virtual characters respond authentically by drawing from a massive database of recorded human dialogue, providing individualized responses tailored to unique user input. The project makes three important intellectual merit contributions. In furthering computer science, it integrates advanced computing techniques -- AI, natural language processing, and crowdsourcing -- in a novel way. Second, it makes theoretical contributions to social science by enhancing understanding of how social psychology principles function in the virtual world. Third, in the field of analytics, software tools will mine patterns of communication strategies, generating knowledge about the impact of specific phrases on relationships. The broader/commercial impact of reducing bullying and improving relationships between children through unscripted, psychologically vivid, safely-controlled interactions in virtual environments is enormous. An extensive and deep body of research shows that bullying is pervasive in schools and has significant acute and chronic effects on learning and health outcomes. Despite the tremendous societal cost of bullying -- estimated around $25 billion -- and legal mandates to address the problem, a majority of schools do not have anti-bullying interventions in place. Existing interventions are expensive, difficult to implement, and focus on building awareness, rather than changing behaviors. This project will create a cost-effective and rapidly-scalable technology that mimics the complexities of human interaction to increase social perspective taking and empathy in children and adolescents. Dozens of middle and high schools have already agreed to implement the technology once developed. Improvement in student relationships and learning in these schools will be immediate. The technology developed has tremendous commercial potential for any training that relies on social and communication skills, such as corporate management, doctor-patient communication, autism therapy, and English language learning. Applications in these and other areas will generate dozens of new jobs for educators, engineers, developers, and data analysts.
