CausalModel

CausalModel is a framework for representing causal relationships among variables, enabling reasoning about interventions and counterfactuals. It formalizes how outcomes arise from mechanisms and can be used to predict the effects of actions in a system.

The most common formalization is the structural causal model (SCM). An SCM consists of endogenous variables, exogenous variables, and a set of structural equations that specify each variable as a function of its direct causes. It can be represented by a directed acyclic graph (DAG) where edges encode direct causal influences.

Inference tasks include determining whether a causal effect is identifiable from observed data, using rules such as back-door and front-door criteria and the do-calculus. The do-operator formalizes interventions, and identifiability determines if the effect of do(X=x) on Y can be computed from data.

Learning from data involves estimating the structural equations and sometimes the graph, using observational data, randomized experiments, or hybrid designs. Counterfactual reasoning derives statements about what would have happened under alternative actions.

Causal models underpin many fields, including epidemiology, economics, and machine learning, and inform policy decisions, causal discovery, and fairness. Limitations include reliance on assumptions about the data-generating process and potential model misspecification.