This program supports innovative basic research on the fundamental principles and methodologies needed to enable intelligent machine behavior, particularly in support of mixed-initiative (i.e., human-machine teaming) systems. The overall vision of this program is that future computational systems will achieve high levels of performance, adaptation, flexibility, self-repair, and other forms of intelligent behavior in the complex, uncertain, adversarial, and highly dynamic environments faced by the U.S. Air Force. This program covers the full spectrum of computational and machine intelligence, from cognitively plausible reasoning processes that are responsible for human performance in complex problem-solving and decision-making tasks, to non-cognitive computational models of intelligence necessary to create robust intelligent systems. Robustness in this context is the ability to achieve high performance given at least some or all of the following factors: uncertainty, incompleteness or errors in knowledge; limitations on sensing; real-world complexity and dynamic change; adversarial factors; unexpected events including system faults; and out-of-scope requirements on system behavior. In the midst of this spectrum are the technologies explicitly needed to seamlessly incorporate intelligent computational systems into mixed human-machine teams. The program is divided into three sub-areas that span the full spectrum of computational and machine intelligence. They are: Computational Cognition, Human-Machine Teaming and Machine Intelligence.
The program encourages cross-disciplinary teams with collaboration including computer scientists, neuroscientists, cognitive scientists, mathematicians, statisticians, operation and management science researchers, information scientists, econometricians and game theoreticians, etc., especially when the research pertains to common issues and when collaboration is likely to generate bidirectional benefits. This program is aggressive, accepts risk, and seeks to be a pathfinder for U.S. Air Force research in this area. Proposals that may lead to breakthroughs or highly disruptive results are especially encouraged.
Basic Research Objectives:
The Computational Cognition sub-area supports innovative basic research on high-order cognitive processes that are responsible for good human performance in complex problem solving and decision-making tasks – we only want to model the things people excel at. The sub-area also seeks to support research on building computational systems that derive from and/or integrate cognitive and biological models of human and animal intelligence. The overall objective is to understand and exploit these processes to create computational models that perform as well as or better than the reasoning systems they emulate. This sub-area seeks basic research that pertains to exploiting the capabilities of the mind and brain (human or animal) for creating more intelligent machines, as well as cognitively plausible mechanisms inspired by human (or animal) reasoning. This includes computational models based on human and animal performance in perception, attention, memory, learning, reasoning, and decision making in order to improve machine performance.
This sub-area does NOT, however, support statistical approaches to machine learning (e.g., "Deep Learning"), or related variants, as fundamental science in that area is addressed by the Science of Information, Computation, Fusion and Learning program described elsewhere in this BAA.
The Machine Intelligence sub-area supports innovative basic research on fundamental principles and methodologies of computational intelligence necessary to create robust intelligent systems. These methodologies may be cognitively inspired, or non-cognitive in nature, taking full advantage of the strengths embodied in mathematical and computational systems, such as the ability to reason with complex formal logic. This sub-area encourages research enabling the creation of computational systems that embody intelligent behavior based on cognitively inspired or purely mathematical approaches. Proposals that lead to advances in the basic principles of machine intelligence for memory, reasoning, planning, scheduling, and cognitively-inspired learning (i.e., NOT "Deep Learning" or other statistical means), action, and communication are desired insofar as these contribute directly towards robustness as defined above.
The Human-Machine Teaming sub-area is primarily concerned with the machine-side of mixed human-machine decision-making, which appears at all levels of U.S. Air Force operations and pervades every stage of U.S. Air Force missions. To that end, new theoretical and empirical guidance is needed to prescribe maximally effective mixtures of human and machine decision making in environments that are becoming increasingly complex and demanding as a result of the high uncertainty, complexity, time urgency, and rapidly changing nature of military missions. This sub-area seeks new empirical and theoretical basic research that enables intelligent machines to perform as true "teammates," adapting their behavior to accommodate changes in the environment, as well as augmenting the performance of human teammates when needed. This includes basic science in collaborative human-machine teams to aid the machine-side of inference, analysis, prediction, planning, scheduling, and decision making.
You are highly encouraged to contact our Program Officer prior to developing a full proposal to briefly discuss the current state-of-the-art, how your research would advance it, the approximate cost for a three (3) to five (5) year effort, and if there are any specific submission target dates.
Collaboration ToolsFollowing are different ways you can share information with Dr. Greenwald and your colleagues.
Program Reviews & Meetings
Contact InformationDr. Hal S. Greenwald (Acting) AFOSR/RTA2 Email: Machine.Itel@us.af.mil