The information content of network dynamics across scales

Complex systems can be studied at multiple spatial and temporal scales. Although previous studies have shown that coarser levels of description can enhance emergent information, they have largely focused on the structure of complex systems, i.e. a static network topology, rather than on the dynamical processes that such networks can represent. In this seminar I will show how the information contained in network-driven dynamics transforms under systematic coarse-graining, involving concepts such as prediction, reconstruction, and causal emergence. Beyond clarifying the informational meaning of coarse-graining, the proposed framework suggests new strategies for multiscale analysis and noise filtering in complex systems, with implications ranging from neuroscience to computation.