Teaching AI to understand ethical rules

Los Angeles CA (SPX) Oct 04, 2024 - As artificial intelligence (AI) becomes increasingly integrated into daily activities, from creating documents to helping with research, ensuring that AI systems respond accurately and ethically has become a key challenge. The Defense Advanced Research Projects Agency (DARPA) aims to address this issue by launching the Human-AI Communications for Deontic Reasoning Devops program, known as CODORD. This initiative focuses on improving AI's ability to understand and process human intent, laws, policies, and norms in logical terms.

CODORD's goal is to translate deontic knowledge-information about obligations, permissions, and prohibitions-into logical programming languages that AI can interpret. By bridging the gap between human language and AI systems, this project seeks to automate the transfer of complex knowledge, making it both faster and more cost-efficient.

Benjamin Grosof, program manager in DARPA's Defense Sciences Office, explained the current challenge: "The current process for transmitting deontic knowledge stored in someone's mind or in written documents into a logical language is very expensive and slow, because it requires specially skilled knowledge engineers trained in logic to work with experts in particular application domains." He emphasized that overcoming this "knowledge acquisition bottleneck" is the key objective of CODORD, which aims to create methods for automatic translation of human language into AI-readable formats.

This development could revolutionize both military and civilian sectors, enabling AI to handle decision-making tasks with high assurance in areas such as operational planning, compliance with regulations, contracts, and intelligence analysis. AI tools powered by CODORD could also support decision-making in healthcare, finance, and defense by offering automated, ethical reasoning.

For instance, AI systems could evaluate banking transactions for compliance with regulations, providing detailed, easy-to-understand explanations of their reasoning. "A trader or bank compliance officer could ask the AI whether a proposed transaction of a certain dollar amount between two banks would be legally permitted versus prohibited," said Grosof. "The AI then provided an answer together with a fully detailed explanation in natural language of how it arrived at that decision."

This capability is especially critical for military use, where conveying a commander's intent accurately is essential. U.S. Marine Corps Col. Robert Gerbracht highlighted this, stating, "I need to have the assurance that an artificial intelligence would be able to pass my intent on, in the spirit in which it was given and within the ethical, legal, and moral guidelines that I, as a commander, issued that guidance."

A CODORD Proposers Day for potential participants will take place on Oct. 8, 2024. More details are available on SAM.gov, where DARPA will post the full solicitation soon.