CoVar Receives DARPA Contract to Develop AI System for Medical Decision-Making Enhancement
核心洞察
CoVar (搜索) has been awarded a Direct-to-Phase II DARPA (搜索) SBIR contract to develop VectoredAI (搜索), a dynamic human-AI teaming model for improving decision-making in high-uncertainty scenarios.
The PPADM program aims to advance scientific understanding of how individual decision-maker characteristics influence behavior, with applications in mission-critical and safety-critical systems.
VectoredAI (搜索) leverages probabilistic key decision-maker attributes and latent semantic steering to align AI solutions with critical decision makers' personalities and considerations.
CoVar (搜索), a small business specializing in advanced artificial intelligence and machine learning solutions for the U.S. Department of Defense, has been awarded a Direct-to-Phase II Small Business Innovation Research (SBIR) contract by the Defense Advanced Research Projects Agency (DARPA (搜索)). The award, announced on April 9, 2026, supports DARPA's Predictive Psychological Architectures for Decision-Making (PPADM) program, which has significant implications for healthcare AI applications.
Revolutionary AI-Human Teaming Technology
The PPADM program aims to advance the scientific understanding of how individual decision-maker characteristics influence behavior, with the goal of improving both human and algorithmic decision-making, delegation, and trust. The program focuses on decision-making in difficult domains with high uncertainty, time and resource constrained scenarios, and circumstances in which decision-makers disagree or have conflicting values - conditions commonly found in medical settings.
CoVar (搜索) will develop VectoredAI (搜索), a dynamic human-AI teaming model that leverages probabilistic key decision-maker attributes (KDMAs) and latent semantic steering. VectoredAI is designed to provide a plug-in capability that helps align AI solutions with the personalities and considerations of critical decision makers.
Technical Innovation and Collaboration
The company has partnered with Dr. Jim Neidhoefer to bring complementary expertise to the VectoredAI (搜索) development. CoVar (搜索)'s existing GEARS software, developed under DARPA (搜索)'s ASIMOV program, enables users to design world scenarios and evaluate dilemmas in terms of tensions and conflicts, supporting automated testing of both algorithmic and human decision making.
Dr. Neidhoefer has developed statistical methods for quantifying and aggregating hard-to-measure subjective factors, including KDMAs, in addition to a methodology for aligning human and algorithmic decision making. This combination of technologies addresses a critical gap in current AI systems deployed in high-stakes environments.
Addressing Trust Barriers in Critical Systems
"Barriers to AI adoption in mission critical and safety critical systems remains high due to lack of human trust," said Dr. Pete Torrione, CTO at CoVar (搜索). "Our proposed VectoredAI (搜索) solution to PPADM will expand on what is already known about human decision making to identify key trust-building factors that reduce friction in human-AI interactions and humans' willingness to delegate difficult decisions to AI."
The Direct-to-Phase II award represents CoVar (搜索)'s continued leadership in building trusted and responsible AI/ML solutions for the Department of Defense, ultimately leading to more rapid adoption of AI for mission critical systems. This technology has potential applications across healthcare, where trust between medical professionals and AI systems is crucial for patient safety and treatment efficacy.