{"id":"arxiv/2610.22286","url":"https://arxiv.org/abs/2610.22286","tags":["math.DS","physics.ao-ph","physics.data-an","stat.ME"],"score":0,"title":"Causal Fingerprints Reveal Hidden Mechanisms in Partially-Observed Dynamical Systems","domain":"math.DS","source":"arxiv","authors":["Zhang Jiang","Marios Andreou","Nan Chen"],"reviews":0,"subject":"math.DS","summary":"In partially-observed dynamical systems, distinct hidden mechanisms can produce similar observable summary statistics while explaining individual events differently. We introduce causal fingerprints: realization-dependent patterns that track how information about a hidden state is revealed through subsequent evolution. Within a prescribed dynamical model, we compare the filtering posterior, based on observations up to a given time, with the smoothing posterior, obtained after adding an expanding window of future observations. We quantify this retrospective information gain using…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-10-15","reactions":{}}