This $339,302 Project Grant, awarded by the National Science Foundation (NSF) under the Computer and Information Science and Engineering (CFDA 47.070) program, supports a collaborative research effort led by the University of Notre Dame to develop personalized language models that can tailor their responses based on the unique characteristics and preferences of individual users.
The project aims to first construct a diverse dataset to understand how people from different backgrounds and experiences prefer to receive information from artificial intelligence (AI) systems. Using this data, the researchers will then develop techniques to enable large language models to detect when they should provide responses sensitive to a user's unique needs, knowledge, and background. This will involve creating adversarial prompt and response pairs and fine-tuning the language models on a small curated dataset to improve their ability to generate acceptable answers for a diverse population. The resulting AI systems and accompanying resources, such as datasets, interfaces, and code, will help enable more personalized and equitable AI-powered technologies.
Generated 3/4/25, 4:39 AM