Causal Emergence of Consciousness through Learned Multiscale Neural Dynamics in Mice
arXiv Preprint Archive September 13, 2025 Zhipeng Wang, Yingqi Rong, Kaiwei Liu et al.
A machine learning framework infers causal variables and their dynamics across multiple scales from near-cellular-resolution calcium imaging of the mouse dorsal cortex. At lower levels, variables aggregate input-driven information; at higher levels they realize causality through metastable or saddle-point dynamics during wakefulness, collapsing into localized, stochastic dynamics under anesthesia. A one-dimensional top-level conscious variable captures most causal power, but variables across other scales also contribute substantially, yielding high emergent complexity in the conscious state. These findings link neural activity to conscious states through a multiscale causal framework.