Bayesian networks and Structural Causal Models are often conflated, however the two modeling approaches are very different. This tech report explores these differences, and explains why Causal Models have far superior functionality.
Bayesian networks and Structural Causal Models are often conflated, however the two modeling approaches are very different. Bayesian networks are purely statistical models that summarize probabilistic relationships between variables. Structural Causal Models represent the underlying causal processes that are responsible for these probabilistic relationships. This gives Causal Models superior functionality, that includes the capability to design optimal interventions and to reason about counterfactual scenarios. This report unpacks the ways in which Structural Causal Models outrun standard Bayesian networks.
This paper is part of our tech reports series, in which we share some of the most significant concepts in the causality literature. Our tech reports are written primarily for a technical audience. For a less technical introduction to causal discovery, go over to our blog.
