Understanding how biological learning processes give rise to high-level cognition—such as reasoning, abstraction, and planning—remains a central challenge in neuroscience. At the same time, statistical physics offers powerful tools for describing how complex, organized behaviors can emerge from simple, local interactions in large systems. This program will bring together researchers across neurobiology, statistical physics, machine learning, and cognitive science to explore how concepts from these fields can be combined to illuminate the dynamics of learning, the formation of internal representations, and the emergence of flexible, structured cognition.

Key questions the program aims to address are: (1) How do large neural systems self-organize to support abstraction, reasoning, and decision-making? (2) What statistical principles govern learning beyond the scale of individual neurons and synapses? (3) How does biological learning leverage noise, redundancy, and variability to produce stable and robust cognition? These questions highlight the main themes which include criticality and phase transitions in learning systems, the emergence of compositional and hierarchical representations, and statistical mechanics approaches to plasticity, memory, and generalization. Overall, the program aims to further our understanding of how meso-scale cognitive processes emerge through the combination and repetition of very many micro-processes.

The program will foster interdisciplinary collaborations to develop new frameworks that bridge biophysics, theory, and behavior, and help identify unifying principles underlying emergent high-level cognitive functions across species and architectures.