This project will develop a new transdisciplinary Institute on Data Science for Intelligent Systems and People Interaction referred to as DATA-INSPIRE. This institute is premised on the belief that advances in data science principles are needed to impact the emerging paradigm of intelligent machines and their convergence with human society, and in particular to further improve the performance and better explain the operation of such machines that can accomplish diverse, real-world tasks and interact effectively with people. Fundamental notions of data science that can enhance development of intelligent machines can impact pressing problems facing our planet: healthcare, transportation, urban systems, etc. DATA-INSPIRE will bring together mathematicians, statisticians, and computer scientists for transdisciplinary research projects, new educational initiatives, workshops, and other efforts designed to catalyze a new foundational data science community focused on the development of intelligent, interactive machines. It will prepare students for transdisciplinary foundational work in data science, aid curriculum development, and involve government and industrial partners in collaborations and to aid in understanding of workforce issues resulting from use of intelligent machines.
Intelligent machines, such as robots, are evolving from simple automata performing repetitive tasks in highly structured and enclosed workspaces to sophisticated, closed-loop systems capable of satisfying human specifications in dynamic environments that include people. To manage and master the operations of complex machines and their interactions with people, it is necessary to better understand and adapt the data that drive the algorithms that control them. DATA-INSPIRE will address the following challenges. (1) Failures in tasks such as autonomous driving or robotic surgery can have devastating consequences. Data-driven solutions are often opaque computational tools for which it is impossible to verify correctness or explain failures. Formal tools, integrated with data, are needed to remove ambiguity about what causes intelligent machines to perform in certain ways. (2) Data-driven solutions frequently depend critically on vast corpora of accurately labeled training instances, which can be difficult to collect for physical operations. Tools are needed to reduce machine learning methods' dependence on large amounts of task-specific supervision. (3) Most intelligent machines need to react to sensing data under critical deadlines, which, if not met, can jeopardize operations. Mathematical and statistical analyses of the dynamics of learning and control are needed to assist with more effective real-time decision making.
This project is part of the National Science Foundation's Harnessing the Data Revolution (HDR) Big Idea activity.
This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.
