SBIR/STTR Award attributes
The objective of the proposed work is to create an analysis software that judge advocates can use to find discrepancies and areas of legal concern quickly and accurately within targeting packages during use-of-force decision-making. The AI Legal Assistant (AILA) will be trained to scan targeting packages to identify discrepancies between images and their human-generated captions for specific Air Force use cases. Customization possibilities for Phase II include ensuring objects (e.g., planes, vehicles, etc.) match in both detail (i.e., color, model, etc.) and use (e.g., civilian vs. military); detecting human bias and assumptions in labeling (e.g., “combatant” vs. “military-aged male” or definitions like “military outpost” which have clear legal definitions); civilian damage estimation / proportionality (e.g., an image labeled as a “field” has different civilian implications if it is an agricultural field vs. a sports field); and collateral estimation (e.g., confirming whether potential dwellings are in the strike zone). If successful, AILA will increase the speed and accuracy of legal review and ultimately improve U.S. targeting outcomes.
