Paul Fodor, associate professor of practice in Stony Brook University’s Department of Computer Science, is part of a team awarded a contract by the Defense Advanced Research Projects Agency (DARPA) to improve how artificial intelligence systems interpret complex regulatory language.
The project, led by RTX BBN Technologies, addresses a critical challenge: current AI systems struggle to accurately interpret regulations, which are often filled with exceptions, nuances and complex legal terminology. Fodor joins researchers from Johns Hopkins University and RTX BBN Technologies in developing DONATELLO (DeONtic AssistanT Enabled by Large Language Models), a solution designed to translate intricate regulatory texts into actionable logic.
The initiative supports DARPA’s Human-AI Communication for Deontic Reasoning Devops (CODORD) program, which aims to enable AI systems to effectively reason about obligations, permissions and prohibitions. Even advanced AI systems can get things wrong when rules are ambiguous or packed with exceptions.
In military and government operations, personnel often lack immediate access to legal experts who can provide guidance on regulatory compliance. DONATELLO employs deontic logic—a framework focused on permissibility, prohibition, and obligation—to enable AI tools to interpret regulations with explainable reasoning.
Beyond military applications, the technology could transform regulatory compliance in healthcare, financial services, and government administration. Work will be performed across Cambridge, Massachusetts; Baltimore, Maryland; and Stony Brook, New York.
Fodor brings extensive expertise in artificial intelligence, natural language processing and logic programming to the collaboration. His research background includes semantic reasoning, database systems and knowledge representation.
By Yuganshu Jain
Add comment