Axiomatization of Interventional Probability Distributions

Explore a 58-minute seminar lecture from the Stochastic Systems for Anomalous Diffusion series, delivered by Dr. Kayvan Sadeghi from University College London at the Isaac Newton Institute. Delve into the fundamental concepts of causal intervention and its axiomatization in causal inference. Learn about new approaches to axiomatizing families of probability distributions for different types of interventional distributions, offering a simplified theory of causality that overcomes traditional limitations. Discover how this innovative framework operates without structural causal model assumptions, focuses on single-variable interventions, accommodates latent variables and causal cycles, and derives causal graphs as a natural consequence rather than a prerequisite. Understand how the intervened distributions maintain Markovian properties in relation to intervened causal graphs, ensuring compatibility with existing causal inference theory while providing a more natural approach than conventional structural causal models.