HALLOW- Open-world Machine Learning with Human-Aligned Trust
Abstract
The USAF-USSF missions face highly uncertain and dynamic environments due to hostile activity or forces of nature. Under such occurrences, intelligent systems need to be aware of uncertainty, and dynamically adapt to continuously operate in the best possible manner. Over the past 50 years, machine learning has achieved remarkable success in known contexts for which they are trained. However, machine learning still lacks the rigor, agility, and flexibility necessary to operate in complex, unknown, and contested open-world environments, where out-of-distribution (OOD) data can naturally arise. The PIs plan for this research is to develop a novel framework called Human-Aligned Learning in the Open World (HALLOW). That is, an intelligent system that can perform automatic decision making that is aligned with trusted humans in difficult and unanticipated situations, and humans can also guide the intelligent agent to improve and adapt over time.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Feb 29, 2024
- Source ID
- FA95502310184
Entities
People
- Sharon Li
Organizations
- Air Force Office of Scientific Research
- United States Air Force
- University of Wisconsin System