Integrated Information in the Active Inference Framework

Watch a 19-minute conference talk exploring the intersection of active inference framework and Integrated Information Theory in modeling consciousness. Learn how perception and action selection are unified through internal generative models that help agents infer future observations and select actions. Discover how integrated information measures can be applied to assess consciousness levels in active inference agents through a newly defined approach incorporating structural assumptions. Examine experimental results demonstrating how modeling choices affect integrated information values, and explore additional information-theoretic measures including Morphological Computation that quantifies agent-environment interactions.