Human-Machine Teaming

Most AI/ML models today operate as mere tools, performing specialized tasks with superhuman speed and capability, but failing to integrate well with the nuances of human operators. All too often, the addition of AI/ML into an existing system actually increases the workload for human operators, as people strive to not only do their own jobs, but to interpret and respond to AI outputs. 

The true measure of an AI model’s value should be its ability to solve human problems and enhance human capabilities. In pursuit of this vision, Galois’s human-machine teaming research aims to create AI/ML systems that operate less like tools and more like collaborators—intuitive, responsive, and adaptable. Our approach to teaming is multifaceted, including active research into metacognition and intention modeling, approaches that could yield AI models trained to recognize changing contexts, understand collaborative goals, and adapt their model and responses to suit individual needs. 

The potential applications for these sorts of human-machine teaming systems are nearly limitless. Already, Galois has developed AI teaming tools to analyze vast amounts of oceanographic and ship data to predict and counteract illegal fishing in the Galapagos. Another project aims to introduce an AI teammate to the cockpit, enhancing fighter pilot performance in high-stress scenarios by adapting how the system communicates about potential threats based on a pilot’s unique profile and responses over time. The same teaming technology is being applied to the creation of personalized training and education. By curating AI-driven learning experiences, tailored to the individual, we can anticipate skill decay and optimize training, ultimately fostering a smarter, faster-learning, better prepared school system, work force, and military — empowered through collaboration with advanced AI teammates. 

At Galois, the interweaving of human intelligence with AI is a continuous journey towards more profound, more effective teamwork. This partnership goes beyond mere task completion; it’s a pursuit of seamless integration, where AI becomes as dynamic and responsive as the best of our human colleagues.