Statistical physics uses probability to connect microscopic dynamics to macroscopic behavior. Its classical approaches deal largely with homogeneous systems at or near equilibrium, often in regimes where macroscopic observables become insensitive to microscopic fluctuations. However, science now faces questions about various complex systems—from living matter to social and information-processing systems—which are instead heterogeneous, strongly fluctuating, or far from equilibrium. Yet they can exhibit robust structures that transcend microscopic details, including thermodynamic constraints, scaling laws, and characteristic fluctuation statistics. Understanding when such universal properties emerge, and where they break down, is a central challenge in modern statistical physics.
We develop theoretical and computational tools to address this challenge, drawing on stochastic processes, hydrodynamics, scaling, large deviation theory, stochastic thermodynamics, information theory, Bayesian inference, and agent-based models. We combine models and data to study subjects across a broad range of scales, including active matter, heat and information engines, artificial intelligence, ecosystems, opinion dynamics, and societies.
