SBIR Phase I: Automated Detection of Confounds and Inappropriate Context to Promote Prosocial Learning and Cognition

Otherwordly Llc, Takoma Park MD

Investigators

Abstract

The broader/commercial impact of this Small Business Innovation Research (SBIR) Phase I project is the development of Artificial Intelligence (AI)-based algorithms that generate content for word-meaning video games at an affordable cost. Word-meaning games support literacy, fluency, critical thinking, and cross-cultural understanding for players of all ages and backgrounds via adaptive vocabulary scaling systems and accessibility options for players with visual or motor difficulties. Science, Technology, Engineering, and Mathematics (STEM) literacy is supported by incorporating STEM content in a mix of entertaining and serious content. These prosocial and cognitive impacts are essential for personal and professional growth, cultural competence, and will be measured by game learning researchers. The project will also contribute to the field of natural language processing and machine learning through the addition of new benchmarks to open-source resources. Success in reducing content creation costs could lead to licensing content to other game publishers and the creation of additional word-meaning games on the market, benefiting players. This project is uniquely positioned to help retain game industry jobs in the U.S. and contribute to the growth of the industry. The technical innovation of the project is threefold: 1) development of algorithms for unrelated word content generation, 2) development of appropriateness and offensiveness filters for natural language content, and 3) evaluation of a word-meaning game’s ability to improve cognitive function and social awareness. This research and development has the potential to address a gap in the field of natural language processing on unrelatedness. Part of this effort contributes to open-source benchmarks for future research. Similarly, social bias is a prevalent and well-known issue in machine learning models, potentially offensive or inappropriate word combinations need to be detected and avoided via newly developed algorithms that explicitly detect and avoid publishing such content. To achieve both goals, a variety of machine learning techniques, including those that leverage existing large natural language models, will be employed and evaluated for accuracy. Using these algorithms as a foundation for content creation, the word-meaning game will integrate the generated content. The program will be evaluated with regard to its ability to increase social awareness and confidence with an expanding vocabulary. Specifically, the study will evaluate both brief gameplay and long-term gameplay and measure efficacy with in-game metrics and surveys. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.