Federal Grant Award Summary

Dartmouth College received a $297,131 Project Grant from the National Science Foundation's (NSF) Division of Computing and Communication Foundations under the Computer and Information Science and Engineering (CISE) program (CFDA 47.070) to develop human-aligned video understanding systems. The five-year award, effective July 1, 2026 through June 30, 2031, will produce benchmark datasets, evaluation tools, and privacy-preserving generative models that enable artificial intelligence systems to interpret emotional and social context in videos while maintaining data privacy and providing explainable decision-making. The research introduces a unified closed-loop learning framework integrating empathy-driven modeling, reasoning-aware evaluation, and privacy-preserving data generation to advance socially responsible AI applications.

The project will deliver tangible research outputs and educational resources designed to improve real-world applications such as accessibility tools, including audio descriptions for blind and low-vision audiences. Dartmouth will integrate the developed benchmark datasets, evaluation protocols, and privacy-preserving models into education and outreach activities to broaden participation in artificial intelligence research and development. The work directly addresses limitations in current AI systems that struggle to explain their decisions and operate on data containing sensitive personal information, positioning the outcomes to support responsible artificial intelligence deployment across multiple sectors.